RFC 4782 - Quick-Start for TCP and IP
(Formats: TXT)
Network Working Group S. Floyd
Request for Comments: 4782 M. Allman
Category: Experimental ICIR
A. Jain
F5 Networks
P. Sarolahti
Nokia Research Center
January 2007
|
Quick-Start for TCP and IP
Status of This Memo
This memo defines an Experimental Protocol for the Internet
community. It does not specify an Internet standard of any kind.
Discussion and suggestions for improvement are requested.
Distribution of this memo is unlimited.
Copyright Notice
Copyright (C) The IETF Trust (2007).
Abstract
This document specifies an optional Quick-Start mechanism for
transport protocols, in cooperation with routers, to determine an
allowed sending rate at the start and, at times, in the middle of a
data transfer (e.g., after an idle period). While Quick-Start is
designed to be used by a range of transport protocols, in this
document we only specify its use with TCP. Quick-Start is designed
to allow connections to use higher sending rates when there is
significant unused bandwidth along the path, and the sender and all
of the routers along the path approve the Quick-Start Request.
This document describes many paths where Quick-Start Requests would
not be approved. These paths include all paths containing routers,
IP tunnels, MPLS paths, and the like that do not support Quick-
Start. These paths also include paths with routers or middleboxes
that drop packets containing IP options. Quick-Start Requests could
be difficult to approve over paths that include multi-access layer-
two networks. This document also describes environments where the
Quick-Start process could fail with false positives, with the sender
incorrectly assuming that the Quick-Start Request had been approved
by all of the routers along the path. As a result of these concerns,
and as a result of the difficulties and seeming absence of motivation
for routers, such as core routers to deploy Quick-Start, Quick-Start
is being proposed as a mechanism that could be of use in controlled
Floyd, et al. Experimental [Page 1]
RFC 4782 Quick-Start for TCP and IP January 2007
environments, and not as a mechanism that would be intended or
appropriate for ubiquitous deployment in the global Internet.
Table of Contents
1. Introduction ....................................................4
1.1. Conventions and Terminology ................................5
2. Assumptions and General Principles ..............................6
2.1. Overview of Quick-Start ....................................7
3. The Quick-Start Option in IP ...................................10
3.1. The Quick-Start Option for IPv4 ...........................10
3.2. The Quick-Start Option for IPv6 ...........................13
3.3. Processing the Quick-Start Request at Routers .............14
3.3.1. Processing the Report of Approved Rate .............15
3.4. The QS Nonce ..............................................16
4. The Quick-Start Mechanisms in TCP ..............................18
4.1. Sending the Quick-Start Request ...........................19
4.2. The Quick-Start Response Option in the TCP header .........20
4.3. TCP: Sending the Quick-Start Response .....................21
4.4. TCP: Receiving and Using the Quick-Start Response Packet ..22
4.5. TCP: Controlling Acknowledgement Traffic on the
Reverse Path ..............................................24
4.6. TCP: Responding to a Loss of a Quick-Start Packet .........26
4.7. TCP: A Quick-Start Request for a Larger Initial Window ....26
4.7.1. Interactions with Path MTU Discovery ...............26
4.7.2. Quick-Start Request Packets that are
Discarded by Routers or Middleboxes ................27
4.8. TCP: A Quick-Start Request in the Middle of a Connection ..28
4.9. An Example Quick-Start Scenario with TCP ..................29
5. Quick-Start and IPsec AH .......................................30
6. Quick-Start in IP Tunnels and MPLS .............................31
6.1. Simple Tunnels that Are Compatible with Quick-Start .......33
6.1.1. Simple Tunnels that Are Aware of Quick-Start .......33
6.2. Simple Tunnels that Are Not Compatible with Quick-Start ...34
6.3. Tunnels That Support Quick-Start ..........................35
6.4. Quick-Start and MPLS ......................................35
7. The Quick-Start Mechanism in Other Transport Protocols .........36
8. Using Quick-Start ..............................................37
8.1. Determining the Rate to Request ...........................37
8.2. Deciding the Permitted Rate Request at a Router ...........37
9. Evaluation of Quick-Start ......................................38
9.1. Benefits of Quick-Start ...................................38
9.2. Costs of Quick-Start ......................................39
9.3. Quick-Start with QoS-Enabled Traffic ......................41
9.4. Protection against Misbehaving Nodes ......................41
9.4.1. Misbehaving Senders ................................41
9.4.2. Receivers Lying about Whether the Request
was Approved .......................................43
Floyd, et al. Experimental [Page 2]
RFC 4782 Quick-Start for TCP and IP January 2007
9.4.3. Receivers Lying about the Approved Rate ............43
9.4.4. Collusion between Misbehaving Routers ..............44
9.5. Misbehaving Middleboxes and the IP TTL ....................46
9.6. Attacks on Quick-Start ....................................46
9.7. Simulations with Quick-Start ..............................47
10. Implementation and Deployment Issues ..........................47
10.1. Implementation Issues for Sending Quick-Start Requests ...47
10.2. Implementation Issues for Processing Quick-Start
Requests .................................................48
10.3. Possible Deployment Scenarios ............................48
10.4. A Comparison with the Deployment Problems of ECN .........50
11. Security Considerations .......................................50
12. IANA Considerations ...........................................52
12.1. IP Option ................................................52
12.2. TCP Option ...............................................52
13. Conclusions ...................................................53
14. Acknowledgements ..............................................53
Appendix A. Related Work ..........................................54
A.1. Fast Start-Ups without Explicit Information from Routers ..54
A.2. Optimistic Sending without Explicit Information from
Routers ...................................................56
A.3. Fast Start-Ups with Other Information from Routers ........56
A.4. Fast Start-Ups with More Fine-Grained Feedback from
Routers ...................................................57
A.5. Fast Start-ups with Lower-Than-Best-Effort Service ........58
Appendix B. Design Decisions ......................................59
B.1. Alternate Mechanisms for the Quick-Start Request:
ICMP and RSVP .............................................59
B.1.1. ICMP ...............................................59
B.1.2. RSVP ...............................................60
B.2. Alternate Encoding Functions ..............................61
B.3. The Quick-Start Request: Packets or Bytes? ................63
B.4. Quick-Start Semantics: Total Rate or Additional Rate? .....64
B.5. Alternate Responses to the Loss of a Quick-Start Packet ...65
B.6. Why Not Include More Functionality? .......................66
B.7. Alternate Implementations for a Quick-Start Nonce .........69
B.7.1. An Alternate Proposal for the Quick-Start Nonce ....69
B.7.2. The Earlier Request-Approved Quick-Start Nonce .....69
Appendix C. Quick-Start with DCCP .................................70
Appendix D. Possible Router Algorithm .............................72
Appendix E. Possible Additional Uses for the Quick-Start ..........74
Normative References ..............................................75
Informative References ............................................75
Floyd, et al. Experimental [Page 3]
RFC 4782 Quick-Start for TCP and IP January 2007
1. Introduction
Each connection begins with a question: "What is the appropriate
sending rate for the current network path?" The question is not
answered explicitly, but each TCP connection determines the sending
rate by probing the network path and altering the congestion window
(cwnd) based on perceived congestion. Each TCP connection starts
with a pre-configured initial congestion window (ICW). Currently,
TCP allows an initial window of between one and four segments of
maximum segment size (MSS) ([RFC2581], [RFC3390]). The TCP
connection then probes the network for available bandwidth using the
slow-start procedure ([Jac88], [RFC2581]), doubling cwnd during each
congestion-free round-trip time (RTT).
The slow-start algorithm can be time-consuming --- especially over
networks with large bandwidth or long delays. It may take a number
of RTTs in slow-start before the TCP connection begins to fully use
the available bandwidth of the network. For instance, it takes
log_2(N) - 2 round-trip times to build cwnd up to N segments,
assuming an initial congestion window of 4 segments. This time in
slow-start is not a problem for large file transfers, where the
slow-start stage is only a fraction of the total transfer time.
However, in the case of moderate-sized transfers, the connection
might carry out its entire transfer in the slow-start phase, taking
many round-trip times, where one or two RTTs might have been
sufficient when using the currently available bandwidth along the
path.
A fair amount of work has already been done to address the issue of
choosing the initial congestion window for TCP, with RFC 3390
allowing an initial window of up to four segments based on the MSS
used by the connection [RFC3390]. Our underlying premise is that
explicit feedback from all the routers along the path would be
required, in the current architecture, for best-effort connections to
use initial windows significantly larger than those allowed by
[RFC3390], in the absence of other information about the path.
In using Quick-Start, a TCP host (say, host A) would indicate its
desired sending rate in bytes per second, using a Quick-Start Option
in the IP header of a TCP packet. Each router along the path could,
in turn, either approve the requested rate, reduce the requested
rate, or indicate that the Quick-Start Request is not approved. (We
note that the `routers' referred to in this document also include the
IP-layer processing of the Quick-Start Request at the sender.) In
approving a Quick-Start Request, a router does not give preferential
treatment to subsequent packets from that connection; the router is
only asserting that it is currently underutilized and believes there
is sufficient available bandwidth to accommodate the sender's
Floyd, et al. Experimental [Page 4]
RFC 4782 Quick-Start for TCP and IP January 2007
requested rate. The Quick-Start mechanism can determine if there are
routers along the path that do not understand the Quick-Start Option,
or have not agreed to the Quick-Start rate request. TCP host B
communicates the final rate request to TCP host A in a transport-
level Quick-Start Response in an answering TCP packet.
If the Quick-Start Request is approved by all routers along the path,
then the TCP host can send at up to the approved rate for a window of
data. Subsequent transmissions will be governed by the default TCP
congestion control mechanisms of that connection. If the Quick-Start
Request is not approved, then the sender would use the default
congestion control mechanisms.
Quick-Start would not be the first mechanism for explicit
communication from routers to transport protocols about sending
rates. Explicit Congestion Notification (ECN) gives explicit
congestion control feedback from routers to transport protocols,
based on the router detecting congestion before buffer overflow
[RFC3168]. In contrast, routers would not use Quick-Start to give
congestion information, but instead would use Quick-Start as an
optional mechanism to give permission to transport protocols to use
higher sending rates, based on the ability of all the routers along
the path to determine if their respective output links are
significantly underutilized.
Section 2 gives an overview of Quick-Start. The formal
specifications for Quick-Start are contained in Sections 3, 4, 6.1.1,
and 6.3. In particular, Quick-Start is specified for IPv4 and for
IPv6 in Section 3, and is specified for TCP in Section 4. Section 6
consists mostly of a non-normative discussion of interactions of
Quick-Start with IP tunnels and MPLS; however, Section 6.1.1 and 6.3
specify behavior for IP tunnels that are aware of Quick-Start.
The rest of the document is non-normative, as follows. Section 5
shows that Quick-Start is compatible with IPsec AH (Authentication
Header). Section 7 gives a non-normative set of guidelines for
specifying Quick-Start in other transport protocols, and Section 8
discusses using Quick-Start in transport end-nodes and routers.
Section 9 gives an evaluation of the costs and benefits of Quick-
Start, and Section 10 discusses implementation and deployment issues.
The appendices discuss related work, Quick-Start design decisions,
and possible router algorithms.
1.1. Conventions and Terminology
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in [RFC2119].
Floyd, et al. Experimental [Page 5]
RFC 4782 Quick-Start for TCP and IP January 2007
2. Assumptions and General Principles
This section describes the assumptions and general principles behind
the design of the Quick-Start mechanism.
Assumptions:
* The data transfer in the two directions of a connection traverses
different queues, and possibly even different routers. Thus, any
mechanism for determining the allowed sending rate would have to be
used independently for each direction.
* The path between the two endpoints is relatively stable, such that
the path used by the Quick-Start Request is generally the same path
used by the Quick-Start packets one round-trip time later.
[ZDPS01] shows this assumption should be generally valid. However,
[RFC3819] discusses a range of Bandwidth on Demand subnets that
could cause the characteristics of the path to change over time.
* Any new mechanism must be incrementally deployable and might not be
supported by all the routers and/or end-hosts. Thus, any new
mechanism must be able to accommodate non-supporting routers or
end-hosts without disturbing the current Internet semantics. We
note that, while Quick-Start is incrementally deployable in this
sense, a Quick-Start Request cannot be approved for a particular
connection unless both end-nodes and all the routers along the path
have been configured to support Quick-Start.
General Principles:
* Our underlying premise is that explicit feedback from all the
routers along the path would be required, in the current
architecture, for best-effort connections to use initial windows
significantly larger than those allowed by [RFC3390], in the
absence of other information about the path.
* A router should only approve a Quick-Start Request if the output
link is underutilized. Any other approach will result in either
per-flow state at the router, or the possibility of a (possibly
transient) queue at the router.
* No per-flow state should be required at the router. Note that,
while per-flow state is not required, we also do not preclude a
router from storing per-flow state for making Quick-Start decisions
or for checking for misbehaving nodes.
Floyd, et al. Experimental [Page 6]
RFC 4782 Quick-Start for TCP and IP January 2007
2.1. Overview of Quick-Start
In this section, we give an overview of the use of Quick-Start with
TCP to request a higher congestion window. The description in this
section is non-normative; the normative description of Quick-Start
with IP and TCP follows in Sections 3 and 4. Quick-Start could be
used in the middle of a connection, e.g., after an idle or
underutilized period, as well as for the initial sending rate; these
uses of Quick-Start are discussed later in the document.
Quick-Start requires end-points and routers to work together, with
end-points requesting a higher sending rate in the Quick-Start (QS)
option in IP, and routers along the path approving, modifying,
discarding, or ignoring (and therefore disallowing) the Quick-Start
Request. The receiver uses reliable, transport-level mechanisms to
inform the sender of the status of the Quick-Start Request. For
example, when TCP is used, the TCP receiver sends feedback to the
sender using a Quick-Start Response option in the TCP header. In
addition, Quick-Start assumes a unicast, congestion-controlled
transport protocol; we do not consider the use of Quick-Start for
multicast traffic.
When sent as a request, the Quick-Start Option includes a request for
a sending rate in bits per second, and a Quick-Start Time to Live (QS
TTL) to be decremented by every router along the path that
understands the option and approves the request. The Quick-Start TTL
is initialized by the sender to a random value. The transport
receiver returns the rate, information about the TTL, and the Quick-
Start Nonce to the sender using transport-level mechanisms; for TCP,
the receiver sends this information in the Quick-Start Response in
the TCP header. In particular, the receiver computes the difference
between the Quick-Start TTL and the IP TTL (the TTL in the IP header)
of the Quick-Start Request packet, and returns this in the Quick-
Start Response. The sender uses the TTL difference to determine if
all the routers along the path decremented the Quick-Start TTL,
approving the Quick-Start Request.
If the request is approved by all the routers along the path, then
the TCP sender combines this allowed rate with the measurement of the
round-trip time, and ends up with an allowed TCP congestion window.
This window is sent rate-paced over the next round-trip time, or
until an ACK packet is received.
Figure 1 shows a successful use of Quick-Start, with the sender's IP
layer and both routers along the path approving the Quick-Start
Request, and the TCP receiver using the Quick-Start Response to
return information to the TCP sender. In this example, Quick-Start
is used by TCP to establish the initial congestion window.
Floyd, et al. Experimental [Page 7]
RFC 4782 Quick-Start for TCP and IP January 2007
Sender Router 1 Router 2 Receiver
------ -------- -------- --------
| <IP TTL: 63>
| <QS TTL: 91>
| <TTL Diff: 28>
| Quick-Start Request
| in SYN or SYN/ACK.
| IP: Decrement QS TTL
| to approve request -->
|
| Decrement
| QS TTL
| to approve
| request -->
|
| Decrement
| QS TTL
| to approve
| request -->
|
| <IP TTL: 60>
| <QS TTL: 88>
| <TTL Diff: 28>
| Return Quick-Start
| info to sender in
| Quick-Start Response
| <-- in TCP ACK packet.
|
| <TTL Diff: 28>
| Quick-Start approved,
| translate to cwnd.
| Report Approved Rate.
V Send cwnd paced over one RTT. -->
Figure 1: A Successful Quick-Start Request.
Figure 2 shows an unsuccessful use of Quick-Start, with one of the
routers along the path not approving the Quick-Start Request. If the
Quick-Start Request is not approved, then the sender uses the default
congestion control mechanisms for that transport protocol, including
the default initial congestion window, response to idle periods, etc.
Floyd, et al. Experimental [Page 8]
RFC 4782 Quick-Start for TCP and IP January 2007
Sender Router 1 Router 2 Receiver
------ -------- -------- --------
| <IP TTL: 63>
| <QS TTL: 91>
| <TTL Diff: 28>
| Quick-Start Request
| in SYN or SYN/ACK.
| IP: Decrement QS TTL
| to approve request -->
|
| Decrement
| QS TTL
| to approve
| request -->
|
| Forward packet
| without modifying
| Quick-Start Option. -->
|
| <IP TTL: 60>
| <QS TTL: 89>
| <TTL Diff: 29>
| Return Quick-Start
| info to sender in
| Quick-Start Response
| <-- in TCP ACK packet.
|
| <TTL Diff: 29>
| Quick-Start not approved.
| Report approved rate.
V Use default initial cwnd. -->
Figure 2: An Unsuccessful Quick-Start Request.
Floyd, et al. Experimental [Page 9]
RFC 4782 Quick-Start for TCP and IP January 2007
3. The Quick-Start Option in IP
3.1. The Quick-Start Option for IPv4
The Quick-Start Request for IPv4 is defined as follows:
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Option | Length=8 | Func. | Rate | QS TTL |
| | | 0000 |Request| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| QS Nonce | R |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 3: The Quick-Start Option for IPv4.
A Quick-Start Request.
The first byte contains the option field, which includes the one-bit
copy flag, the 2-bit class field, and the 5-bit option number.
The second byte contains the length field, indicating an option
length of eight bytes.
The third byte includes a four-bit Function field. If the Function
field is set to "0000", then the Quick-Start Option is a Rate
Request. In this case, the second half of the third byte is a four-
bit Rate Request field.
For a Rate Request, the fourth byte contains the Quick-Start TTL (QS
TTL) field. The sender MUST set the QS TTL field to a random value.
Routers that approve the Quick-Start Request decrement the QS TTL
(mod 256) by the same amount that they decrement the IP TTL. (As
elsewhere in this document, we use the term `router' imprecisely to
also include the Quick-Start functionality at the IP layer at the
sender.) The QS TTL is used by the sender to detect if all the
routers along the path understood and approved the Quick-Start
option.
For a Rate Request, the transport sender MUST calculate and store the
TTL Diff, the difference between the IP TTL value, and the QS TTL
value in the Quick-Start Request packet, as follows:
TTL Diff = ( IP TTL - QS TTL ) mod 256 (1)
Floyd, et al. Experimental [Page 10]
RFC 4782 Quick-Start for TCP and IP January 2007
For a Rate Request, bytes 5-8 contain a 30-bit QS Nonce, discussed in
Section 3.4, and a two-bit Reserved field. The sender SHOULD set the
reserved field to zero, and routers and receivers SHOULD ignore the
reserved field. The sender SHOULD set the 30-bit QS Nonce to a
random value.
The sender initializes the Rate Request to the desired sending rate,
including an estimate of the transport and IP header overhead. The
encoding function for the Rate Request sets the request rate to K*2^N
bps (bits per second), for N the value in the Rate Request field, and
for K set to 40,000. For N=0, the rate request would be set to zero,
regardless of the encoding function. This is illustrated in Table 1
below. For the four-bit Rate Request field, the request range is
from 80 Kbps to 1.3 Gbps. Alternate encodings that were considered
for the Rate Request are given in Appendix B.2.
N Rate Request (in Kbps)
--- ----------------------
0 0
1 80
2 160
3 320
4 640
5 1,280
6 2,560
7 5,120
8 10,240
9 20,480
10 40,960
11 81,920
12 163,840
13 327,680
14 655,360
15 1,310,720
Table 1: Mapping from Rate Request Field to Rate Request in Kbps.
Routers can approve the Quick-Start Request for a lower rate by
decreasing the Rate Request in the Quick-Start Request. Section 4.2
discusses the Quick-Start Response from the TCP receiver to the TCP
sender, and Section 4.4 discusses the TCP sender's mechanism for
determining if a Quick-Start Request has been approved.
Floyd, et al. Experimental [Page 11]
RFC 4782 Quick-Start for TCP and IP January 2007
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Option | Length=8 | Func. | Rate | Not Used |
| | | 1000 | Report| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| QS Nonce | R |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 4: The Quick-Start Option for IPv4.
Report of Approved Rate.
If the Function field in the third byte of the Quick-Start Option is
set to "1000", then the Quick-Start Option is a Report of Approved
Rate. In this case, the second four bits in the third byte are the
Rate Report field, formatted exactly as in the Rate Request field in
a Rate Request. For a Report of Approved Rate, the fourth byte of
the Quick-Start Option is not used. Bytes 5-8 contain a 30-bit QS
Nonce and a 2-bit Reserved field.
After an approved Rate Request, the sender MUST report the Approved
Rate, using a Quick-Start Option configured as a Report of Approved
Rate with the Rate Report field set to the approved rate, and the QS
Nonce set to the QS Nonce sent in the Quick-Start Request. The
packet containing the Report of Approved Rate MUST be either a
control packet sent before the first Quick-Start data packet, or a
Quick-Start Option in the first data packet itself. The Report of
Approved Rate does not have to be sent reliably; for example, if the
approved rate is reported in a separate control packet, the sender
does not necessarily know if the control packet has been dropped in
the network. If the packet containing the Quick-Start Request is
acknowledged, but the acknowledgement packet does not contain a
Quick-Start Response, then the sender MUST assume that the Quick-
Start Request was denied, and set a Report of Approved Rate with a
rate of zero. Routers may choose to ignore the Report of Approved
Rate, or to use the Report of Approved Rate but ignore the QS Nonce.
Alternately, some routers that use the Report of Approved Rate may
choose to match the QS Nonce, masked by the approved rate, with the
QS Nonce seen in the original request.
If the Rate Request is denied, the sender MUST send a Report of
Approved Rate with the Rate Report field set to zero.
We note that, unlike a Quick-Start Request sent at the beginning of a
connection, when a Quick-Start Request is sent in the middle of a
connection, the connection could already have an established
congestion window or sending rate. The Rate Request is the requested
total rate for the connection, including the current rate of the
Floyd, et al. Experimental [Page 12]
RFC 4782 Quick-Start for TCP and IP January 2007
connection; the Rate Request is *not* a request for an additional
sending rate over and above the current sending rate. If the Rate
Request is denied, or lowered to a value below the connection's
current sending rate, then the sender ignores the request, and
reverts to the default congestion control mechanisms of the transport
protocol.
The use of the Quick-Start Option does not require the additional use
of the Router Alert Option [RFC2113].
We note that in IPv4, a change in IP options at routers requires
recalculating the IP header checksum.
3.2. The Quick-Start Option for IPv6
The Quick-Start Option for IPv6 is placed in the Hop-by-Hop Options
extension header that is processed at every network node along the
communication path [RFC2460]. The option format following the
generic Hop-by-Hop Options header is identical to the IPv4 format,
with the exception that the Length field should exclude the common
type and length fields in the option format and be set to 6 bytes
instead of 8 bytes.
For a Quick-Start Request, the transport receiver compares the
Quick-Start TTL with the IPv6 Hop Limit field to calculate the TTL
Diff. (The Hop Limit in IPv6 is the equivalent of the TTL in IPv4.)
That is, TTL Diff MUST be calculated and stored as follows:
TTL Diff = ( IPv6 Hop Limit - QS TTL ) mod 256 (2)
Unlike IPv4, modifying or deleting the Quick-Start IPv6 Option does
not require checksum re-calculation, because the IPv6 header does not
have a checksum field, and modifying the Quick-Start Request in the
IPv6 Hop-by-Hop options header does not affect the IPv6 pseudo-
header checksum used in upper-layer checksum calculations.
Appendix A of RFC 2460 requires that all packets with the same flow
label must be originated with the same hop-by-hop header contents,
which would be incompatible with Quick-Start. However, a later IPv6
flow label specification [RFC3697] updates the use of flow labels in
IPv6 and removes this restriction. Therefore, Quick-Start is
compatible with the current IPv6 specifications.
Floyd, et al. Experimental [Page 13]
RFC 4782 Quick-Start for TCP and IP January 2007
3.3. Processing the Quick-Start Request at Routers
The Quick-Start Request does not report the current sending rate of
the connection sending the request; in the default case of a router
(or IP-layer implementation at an end-node) that does not maintain
per-flow state, a router makes the conservative assumption that the
flow's current sending rate is zero. Each participating router can
either terminate or approve the Quick-Start Request. A router MUST
only approve a Quick-Start Request if the output link is
underutilized, and if the router judges that the output link will
continue to be underutilized if this and earlier approved requests
are used by the senders. Otherwise, the router reduces or terminates
the Quick-Start Request.
While the paragraph above defines the *semantics* of approving a
Quick-Start Request, this document does not specify the specific
algorithms to be used by routers in processing Quick-Start Requests
or Reports. This is similar to RFC 3168, which specifics the
semantics of the ECN codepoints in the IP header, but does not
specify specific algorithms for routers to use in deciding when to
drop or mark packets before buffer overflow.
A router that wishes to terminate the Quick-Start Request SHOULD
either delete the Quick-Start Request from the IP header or zero the
QS TTL, QS Nonce, and Rate Request fields. Deleting the Quick-Start
Request saves resources because downstream routers will have no
option to process, but zeroing the Rate Request field may be more
efficient for routers to implement. As suggested in [B05], future
additions to Quick-Start could define error codes for routers to
insert into the QS Nonce field to report back to the sender the
reason that the Quick-Start Request was denied, e.g., that the router
is denying all Quick-Start Requests at this time, or that this
router, as a matter of policy, does not grant Quick-Start requests.
A router that doesn't understand the Quick-Start Option will simply
forward the packet with the Quick-Start Request unchanged (ensuring
that the TTL Diff will not match and Quick-Start will not be used).
If the participating router has decided to approve the Quick-Start
Request, it does the following:
* The router MUST decrement the QS TTL by the amount the IP TTL is
decremented (usually one).
* If the router is only willing to approve a Rate Request less than
that in the Quick-Start Request, then the router replaces the Rate
Request with a smaller value. The router MUST NOT increase the
Rate Request in the Quick-Start Request. If the router decreases
Floyd, et al. Experimental [Page 14]
RFC 4782 Quick-Start for TCP and IP January 2007
the Rate Request, the router MUST also modify the QS Nonce, as
described in Section 3.4.
* In IPv4, the router MUST update the IP header checksum.
If the router approves the Quick-Start Request, this approval SHOULD
be taken into account in the router's decision to accept or reject
subsequent Quick-Start Requests (e.g., using a variable that tracks
the recent aggregate of accepted Quick-Start Requests). This
consideration of earlier approved Quick-Start Requests is necessary
to ensure that the router only approves a Quick-Start Request when
the router judges that the output link will remain underutilized if
this and earlier Quick-Start Requests are used by the senders.
In addition, the approval of a Quick-Start Request SHOULD NOT be used
by the router to affect the treatment of the data packets that arrive
from this connection in the next few round-trip times. That is, the
approval by the router of a Quick-Start Request does not give
differential treatment for Quick-Start data packets at that router;
it is only a statement from the router that the router believes that
the subsequent Quick-Start data packets from this connection will not
change the current underutilized state of the router.
A non-participating router forwards the Quick-Start Request
unchanged, without decrementing the QS TTL. The non-participating
router still decrements the TTL field in the IP header, as is
required for all routers [RFC1812]. As a result, the sender will be
able to detect that the Quick-Start Request had not been understood
or approved by all of the routers along the path.
A router that uses multipath routing for packets within a single
connection MUST NOT approve a Quick-Start Request. This is discussed
in more detail in Section 9.2.
3.3.1. Processing the Report of Approved Rate
If the Quick-Start Option has the Function field set to "1000", then
this is a Report of Approved Rate, rather than a Rate Request. The
router MAY ignore such an option, and, in any case, it MUST NOT
modify the contents of the option for a Report of Approved Rate.
However, the router MAY use the Approved Rate report to check that
the sender is not lying about the approved rate. If the reported
Approved Rate is higher than the rate that the router actually
approved for this connection in the previous round-trip time, then
the router may implement some policy for cheaters. For instance, the
router MAY decide to deny future Quick-Start Requests from this
sender, including, if desired, deleting Quick-Start Requests from
future packets from this sender. Section 9.4.1 discusses misbehaving
Floyd, et al. Experimental [Page 15]
RFC 4782 Quick-Start for TCP and IP January 2007
senders in more detail. From the Report of Approved Rate, the router
can also learn if some of the downstream routers have approved the
Quick-Start Request for a smaller rate or denied the use of Quick-
Start, and adjust its bandwidth allocations accordingly.
3.4. The QS Nonce
The QS Nonce gives the Quick-Start sender some protection against
receivers lying about the value of the received Rate Request. This
is particularly important if the receiver knows the original value of
the Rate Request (e.g., when the sender always requests the same
value, and the receiver has a long history of communication with that
sender). Without the QS Nonce, there is nothing to prevent the
receiver from reporting back to the sender a Rate Request of K, when
the received Rate Request was, in fact, less than K.
Table 2 gives the format for the 30-bit QS Nonce.
Bits Purpose
--------- ------------------
Bits 0-1: Rate 15 -> Rate 14
Bits 2-3: Rate 14 -> Rate 13
Bits 4-5: Rate 13 -> Rate 12
Bits 6-7: Rate 12 -> Rate 11
Bits 8-9: Rate 11 -> Rate 10
Bits 10-11: Rate 10 -> Rate 9
Bits 12-13: Rate 9 -> Rate 8
Bits 14-15: Rate 8 -> Rate 7
Bits 16-17: Rate 7 -> Rate 6
Bits 18-19: Rate 6 -> Rate 5
Bits 20-21: Rate 5 -> Rate 4
Bits 22-23: Rate 4 -> Rate 3
Bits 24-25: Rate 3 -> Rate 2
Bits 26-27: Rate 2 -> Rate 1
Bits 28-29: Rate 1 -> Rate 0
Table 2: The QS Nonce.
The transport sender MUST initialize the QS Nonce to a random value.
If the router reduces the Rate Request from rate K to rate K-1, then
the router MUST set the field in the QS Nonce for "Rate K -> Rate
K-1" to a new random value. Similarly, if the router reduces the
Rate Request by N steps, the router MUST set the 2N bits in the
relevant fields in the QS Nonce to a new random value. The receiver
MUST report the QS Nonce back to the sender.
Floyd, et al. Experimental [Page 16]
RFC 4782 Quick-Start for TCP and IP January 2007
If the Rate Request was not decremented in the network, then the QS
Nonce should have its original value. Similarly, if the Rate Request
was decremented by N steps in the network, and the receiver reports
back a Rate Request of K, then the last 2K bits of the QS Nonce
should have their original value.
With the QS Nonce, the receiver has a 1/4 chance of cheating about
each step change in the rate request. Thus, if the rate request is
reduced by two steps in the network, the receiver has a 1/16 chance
of successfully reporting that the original request was approved, as
this requires reporting the original value for the QS nonce.
Similarly, if the rate request is reduced many steps in the network,
and the receiver receives a QS Option with a rate request of K, the
receiver has a 1/16 chance of guessing the original values for the
fields in the QS nonce for "Rate K+2 -> Rate K+1" and "Rate K+1 ->
Rate K". Thus, the receiver has a 1/16 chance of successfully lying
and saying that the received rate request was K+2 instead of K.
We note that the protection offered by the QS Nonce is the same
whether one router makes all the decrements in the rate request, or
whether they are made at different routers along the path.
The requirements for randomization for the sender and routers in
setting `random' values in the QS Nonce are not stringent -- almost
any form of pseudo-random numbers will do. The requirement is that
the original value for the QS Nonce is not easily predictable by the
receiver, and in particular, the nonce MUST NOT be easily determined
from inspection of the rest of the packet or from previous packets.
In particular, the nonce MUST NOT be based only on a combination of
specific packet header fields. Thus, if two bits of the QS Nonce are
changed by a router along the path, the receiver should not be able
to guess those two bits from the other 28 bits in the QS Nonce.
An additional requirement is that the receiver cannot be able to
tell, from the QS Nonce itself, which numbers in the QS Nonce were
generated by the sender, and which were generated by routers along
the path. This makes it harder for the receiver to infer the value
of the original rate request, making it one step harder for the
receiver to cheat.
Section 9.4 also considers issues of receiver cheating in more
detail.
Floyd, et al. Experimental [Page 17]
RFC 4782 Quick-Start for TCP and IP January 2007
4. The Quick-Start Mechanisms in TCP
This section describes how the Quick-Start mechanism would be used in
TCP. We first sketch the procedure and then tightly define it in the
subsequent subsections.
If a TCP sender (say, host A) would like to use Quick-Start, the TCP
sender puts the requested sending rate in bits per second,
appropriately formatted, in the Quick-Start Option in the IP header
of the TCP packet, called the Quick-Start Request packet. (We will
be somewhat loose in our use of "packet" vs. "segment" in this
section.) When used for initial start-up, the Quick-Start Request
packet can be either the SYN or SYN/ACK packet, as illustrated in
Figure 1. The requested rate includes an estimate for the transport
and IP header overhead. The TCP receiver (say, host B) returns the
Quick-Start Response option in the TCP header in the responding
SYN/ACK packet or ACK packet, called the Quick-Start Response packet,
informing host A of the results of their request.
If the acknowledging packet does not contain a Quick-Start Response,
or contains a Quick-Start Response with the wrong value for the TTL
Diff or the QS Nonce, then host A MUST assume that its Quick-Start
request failed. In this case, host A sends a Report of Approved Rate
with a Rate Report of zero, and uses TCP's default congestion control
procedure. For initial start-up, host A uses the default initial
congestion window ([RFC2581], [RFC3390]).
If the returning packet contains a valid Quick-Start Response, then
host A uses the information in the response, along with its
measurement of the round-trip time, to determine the Quick-Start
congestion window (QS-cwnd). Quick-Start data packets are defined as
data packets sent as the result of a successful Quick-Start request,
up to the time when the first Quick-Start packet is acknowledged.
The sender also sends a Report of Approved Rate. In order to use
Quick-Start, the TCP host MUST use rate-based pacing [VH97] to
transmit Quick-Start packets at the rate indicated in the Quick-Start
Response, at the level of granularity possible by the sending host.
We note that the limitations of interrupt timing on computers can
limit the ability of the TCP host in rate-pacing the outgoing
packets.
The two TCP end-hosts can independently decide whether to request
Quick-Start. For example, host A could send a Quick-Start Request in
the SYN packet, and host B could also send a Quick-Start Request in
the SYN/ACK packet.
Floyd, et al. Experimental [Page 18]
RFC 4782 Quick-Start for TCP and IP January 2007
4.1. Sending the Quick-Start Request
When sending a Quick-Start Request, the TCP sender SHOULD send the
request on a packet that requires an acknowledgement, such as a SYN,
SYN/ACK, or data packet. In this case, if the packet is acknowledged
but no Quick-Start Response is included, then the sender knows that
the Quick-Start Request has been denied, and can send a Report of
Approved Rate.
In addition to the use of Quick-Start when a connection is
established, there are several additional points in a connection when
a transport protocol may want to issue a Rate Request. We first
reiterate the notion that Quick-Start is a coarse-grained mechanism.
That is, Quick-Start's Rate Requests are not meant to be used for
fine-grained control of the transport's sending rate. Rather, the
transport MAY issue a Rate Request when no information about the
appropriate sending rate is available, and the default congestion
control mechanisms might be significantly underestimating the
appropriate sending rate.
The following are potential points where Quick-Start may be useful:
(1) At or soon after connection initiation, when the transport has no
idea of the capacity of the network path, as discussed above. (A
transport that uses TCP Control Block sharing [RFC2140], the
Congestion Manager [RFC3124], or other mechanisms for sharing
congestion information may not need Quick-Start to determine an
appropriate rate.)
(2) After an idle period when the transport no longer has a validated
estimate of the available bandwidth for this flow. (An example
could be a persistent-HTTP connection when a new HTTP request is
received after an idle period.)
(3) After a host has received explicit indications that one of the
endpoints has moved its point of network attachment. This can
happen due to some underlying mobility mechanism like Mobile IP
([RFC3344], [RFC3775]). Some transports, such as Steam Control
Transmission Protocol (SCTP) [RFC2960], may associate with
multiple IP addresses and can switch addresses (and therefore
network paths) in mid-connection. If the transport has concrete
knowledge of a changing network path, then the current sending
rate may not be appropriate, and the transport sender may use
Quick-Start to probe the network to see if it can send at a
higher rate. (Alternatively, traditional slow-start should be
used in this case when Quick-Start is not available.)
Floyd, et al. Experimental [Page 19]
RFC 4782 Quick-Start for TCP and IP January 2007
(4) After an application-limited period, when the sender has been
using only a small amount of its appropriate share of the network
capacity and has no valid estimate for its fair share. In this
case, Quick-Start may be an appropriate mechanism to determine if
the sender can send at a higher rate. For instance, consider an
application that steadily exchanges low- rate control messages
and suddenly needs to transmit a large amount of data.
Of the above, this document recommends that a TCP sender MAY attempt
to use Quick-Start in cases (1) and (2). It is NOT RECOMMENDED that
a TCP sender use Quick-Start for case (3) at the current time. Case
(3) requires external notifications not presently defined for TCP or
other transport protocols. Finally, a TCP SHOULD NOT use Quick-
Start for case (4) at the current time. Case (4) requires further
thought and investigation with regard to how the transport protocol
could determine it was in a situation that would warrant transmitting
a Quick-Start Request.
As a general guideline, a TCP sender SHOULD NOT request a sending
rate larger than it is able to use over the next round-trip time of
the connection (or over 100 ms, if the round-trip time is not known),
except as required to round up the desired sending rate to the next-
highest allowable request.
In any circumstances, the sender MUST NOT make a QS request if it has
made a QS request within the most recent round-trip time.
Section 4.7 discusses some of the issues of using Quick-Start at
connection initiation, and Section 4.8 discusses issues that arise
when Quick-Start is used to request a larger sending rate after an
idle period.
4.2. The Quick-Start Response Option in the TCP header
In order to approve the use of Quick-Start, the TCP receiver responds
to the receipt of a Quick-Start Request with a Quick-Start Response,
using the Quick-Start Response Option in the TCP header. TCP's
Quick-Start Response option is defined as follows:
Floyd, et al. Experimental [Page 20]
RFC 4782 Quick-Start for TCP and IP January 2007
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Kind | Length=8 | Resv. | Rate | TTL Diff |
| | | |Request| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| QS Nonce | R |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 5: The Quick-Start Response Option in the TCP Header.
The first byte of the Quick-Start Response option contains the option
kind, identifying the TCP option.
The second byte of the Quick-Start Response option contains the
option length in bytes. The length field MUST be set to 8 bytes.
The third byte of the Quick-Start Response option contains a four-
bit Reserved field, and the four-bit allowed Rate Request, formatted
as in the Quick-Start Rate Request option.
The fourth byte of the TCP option contains the TTL Diff. The TTL
Diff contains the difference between the IP TTL and QS TTL fields in
the received Quick-Start Request packet, as calculated in equations
(1) or (2) (depending on whether IPv4 or IPv6 is used).
Bytes 5-8 of the TCP option contain the 30-bit QS Nonce and a two-
bit Reserved field.
We note that, while there are limitations on the potential size of
the Quick-Start Response Option, a Quick-Start Response Option of
eight bytes should not be a problem. The TCP Options field can
contain up to 40 bytes. Other TCP options that might be used in a
SYN or SYN/ACK packet include Maximum Segment Size (four bytes), Time
Stamp (ten bytes), Window Scale (three bytes), and Selective
Acknowledgments Permitted (two bytes).
4.3. TCP: Sending the Quick-Start Response
An end host (say, host B) that receives an IP packet containing a
Quick-Start Request passes the Quick-Start Request, along with the
value in the IP TTL field, to the receiving TCP layer.
If the TCP host is willing to permit the Quick-Start Request, then a
Quick-Start Response option is included in the TCP header of the
corresponding acknowledgement packet. The Rate Request in the
Quick-Start Response option is set to the received value of the Rate
Request in the Quick-Start Option, or to a lower value if the TCP
Floyd, et al. Experimental [Page 21]
RFC 4782 Quick-Start for TCP and IP January 2007
receiver is only willing to allow a lower Rate Request. The TTL Diff
in the Quick-Start Response is set to the difference between the IP
TTL value and the QS TTL value as given in equation (1) or (2)
(depending on whether IPv4 or IPv6 is used). The QS Nonce in the
Response is set to the received value of the QS Nonce in the Quick-
Start Option.
If an end host receives an IP packet with a Quick-Start Request with
a rate request of zero, then that host SHOULD NOT send a Quick-Start
Response.
The Quick-Start Response MUST NOT be resent if it is lost in the
network. Packet loss could be an indication of congestion on the
return path, in which case it is better not to approve the Quick-
Start Request.
4.4. TCP: Receiving and Using the Quick-Start Response Packet
A TCP host (say, TCP host A) that sent a Quick-Start Request and
receives a Quick-Start Response in an acknowledgement first checks
that the Quick-Start Response is valid. The Quick-Start Response is
valid if it contains the correct value for the TTL Diff, and an equal
or lesser value for the Rate Request than that transmitted in the
Quick-Start Request. In addition, if the received Rate Request is K,
then the rightmost 2K bits of the QS Nonce must match those bits in
the QS Nonce sent in the Quick-Start Request. If these checks are
not successful, then the Quick-Start Request failed, and the TCP host
MUST use the default TCP congestion window that it would have used
without Quick-Start. If the rightmost 2K bits of the QS Nonce do not
match those bits in the QS Nonce sent in the Quick-Start Request, for
a received Rate Request of K, then the TCP host MUST NOT send
additional Quick-Start Requests during the life of the connection.
Whether or not the Quick-Start Request was successful, the host
receiving the Quick-Start Response MUST send a Report of Approved
Rate. Similarly, if the packet containing the Quick-Start Request is
acknowledged, but the acknowledgement does not include a Quick-Start
Response, then the sender MUST send a Report of Approved Rate.
If the checks of the TTL Diff and the Rate Request are successful,
and the TCP host is going to use the Quick-Start Request, it MUST
start using it within one round-trip time of receiving the Quick-
Start Response, or within three seconds, whichever is smaller. To
use the Quick-Start Request, the host sets its Quick-Start congestion
window (in terms of MSS-sized segments), QS-cwnd, as follows:
QS-cwnd = (R * T) / (MSS + H) (3)
Floyd, et al. Experimental [Page 22]
RFC 4782 Quick-Start for TCP and IP January 2007
where R is the Rate Request in bytes per second, T is the measured
round-trip time in seconds, and H is the estimated TCP/IP header size
in bytes (e.g., 40 bytes).
Derivation: the sender is allowed to transmit at R bytes per second
including packet headers, but only R*MSS/(MSS+H) bytes per second, or
equivalently R*T*MSS/(MSS+H) bytes per round-trip time, of
application data.
The TCP host SHOULD set its congestion window cwnd to QS-cwnd only if
QS-cwnd is greater than cwnd; otherwise, QS-cwnd is ignored. If
QS-cwnd is used, the TCP host sets a flag that it is in Quick-Start
mode, and while in Quick-Start mode, the TCP sender MUST use rate-
based pacing to pace out Quick-Start packets at the approved rate.
If, during Quick-Start mode, the TCP sender receives ACKs for packets
sent before this Quick-Start mode was entered, these ACKs are
processed as usual, following the default congestion control
mechanisms. Quick-Start mode ends when the TCP host receives an ACK
for one of the Quick-Start packets.
If the congestion window has not been fully used when the first ack
arrives ending the Quick-Start mode, then the congestion window is
decreased to the amount that has actually been used so far. This is
necessary because when the Quick-Start Response is received, the TCP
sender's round-trip-time estimate might be longer than for succeeding
round-trip times, e.g., because of delays at routers processing the
IP Quick-Start Option, or because of delays at the receiver in
responding to the Quick-Start Request packet. In this case, an
overly large round-trip-time estimate could have caused the TCP
sender to translate the approved Quick-Start sending rate in bytes
per second into a congestion window that is larger than needed, with
the TCP sender receiving an ACK for the first Quick- Start packet
before the entire congestion window has been used. Thus, when the
TCP sender receives the first ACK for a Quick-Start packet, the
sender MUST reduce the congestion window to the amount that has
actually been used.
As an example, a TCP sender with an approved Quick-Start Request of R
KBps, B-byte packets including headers, and an RTT estimate of T
seconds, would translate the Rate Request of R KBps to a congestion
window of R*T/B packets. The TCP sender would send the Quick-Start
packets rate-paced at R KBps. However, if the actual current round-
trip time was T/2 seconds instead of T seconds, then the sender would
begin to receive acknowledgements for Quick-Start packets after T/2
seconds. Following the paragraph above, the TCP sender would then
reduce its congestion window from R*T/B to approximately R*T/(B*2)
packets, the actual number of packets that were needed to fill the
pipe at a sending rate of R KBps. (Note: this is just an
Floyd, et al. Experimental [Page 23]
RFC 4782 Quick-Start for TCP and IP January 2007
illustration; the congestion window is actually set to the amount of
data sent before the ACK arrives and not based on equations above.)
After Quick-Start mode is exited and the congestion window adjusted
if necessary, the TCP sender returns to using the default congestion-
control mechanisms, processing further incoming ACK packets as
specified by those congestion control mechanisms. For example, if
the TCP sender was in slow-start prior to the Quick-Start Request,
and no packets were lost or marked since that time, then the sender
continues in slow-start after exiting Quick-Start mode, as allowed by
ssthresh.
To add robustness, the TCP sender MUST use Limited Slow-Start
[RFC3742] along with Quick-Start. With Limited Slow-Start, the TCP
sender limits the number of packets by which the congestion window is
increased for one window of data during slow-start.
When Quick-Start is used at the beginning of a connection, before any
packet marks or losses have been reported, the TCP host MAY use the
reported Rate Request to set the slow-start threshold to a desired
value, e.g., to some small multiple of the congestion window. A
possible future research topic is how the sender might modify the
slow-start threshold at the beginning of a connection to avoid
overshooting the path capacity. (The initial value of ssthresh is
allowed to be arbitrarily high, and some TCP implementations use the
size of the advertised window for ssthresh [RFC2581].)
4.5. TCP: Controlling Acknowledgement Traffic on the Reverse Path
When a Quick-Start Request is approved for a TCP sender, the
resulting Quick-Start data traffic can result in a sudden increase in
traffic for pure ACK packets on the reverse path. For example, for
the largest Quick-Start Request of 1.3 Gbps, given a TCP sender with
1500-byte packets and a TCP receiver with delayed acknowledgements
acking every other packet, this could result in 17.3 Mbps of
acknowledgement traffic on the reverse path.
One possibility, in cases with large Quick-Start Requests, would be
for TCP receivers to send Quick-Start Requests to request bandwidth
for the acknowledgement traffic on the reverse path. However, in our
view, a better approach would be for TCP receivers to simply control
the rate of sending acknowledgement traffic. The optimal future
solution would involve the explicit use of congestion control for TCP
acknowledgement traffic, as is done now for the acknowledgement
traffic in DCCP's CCID2 [RFC4341].
Floyd, et al. Experimental [Page 24]
RFC 4782 Quick-Start for TCP and IP January 2007
In the absence of congestion control for acknowledgement traffic, the
TCP receiver could limit its sending rate for ACK packets sent in
response to Quick-Start data packets. The following information is
needed by the TCP receiver:
* The RTT: TCP naturally measures the RTT of the path and therefore
should have a sample of the RTT. If the TCP receiver does not have
a measurement of the round-trip time, it can use the time between
the receipt of the Quick-Start Request and the Report of Approved
Rate.
* The Approved Rate Request (R): When the TCP receiver receives the
Quick-Start Response packet, the receiver knows the value of the
approved Rate Request.
* The MSS: TCP advertises the MSS during the initial three-way
handshake; therefore, the receiver should have an understanding of
the packet size the sender will be using. If the receiver does not
have such an understanding or wishes to confirm the negotiated MSS,
the size of the first data packet can be used.
With this set of information, the TCP receiver can restrict its
sending rate for pure acknowledgment traffic to at most 100 pure ACK
packets per RTT by sending at most one ACK for every K data packets,
for the ACK Ratio K set to R*RTT/(100*8*MSS). The receiver would
acknowledge the first Quick-Start data packet, and every succeeding K
data packets. Thus, for a somewhat extreme case of R=1.3 Gbps,
RTT=0.2 seconds, and MSS=1500 bytes, K would be set to 216, and the
receiver would acknowledge every 216 data packets. From [RFC2581],
the ACK Ratio K should have a minimum value of two. When the ACK
Ratio is greater than two, and the TCP sender receives
acknowledgements each acknowledging more than two data packets, the
TCP sender may want to use rate-based pacing to control the
burstiness of its outgoing data traffic.
In the absence of explicit congestion control mechanisms, the TCP end
nodes cannot determine the packet drop rate for pure acknowledgement
traffic. This is true with or without Quick-Start. However, the TCP
receiver could limit its increase in the sending rate for pure ACK
packets by at most doubling the sending rate for pure ACK packets
from one round-trip time to the next. The TCP receiver would do this
by halving the ACK Ratio each round-trip time.
Note that the above is one particular mechanism that could be used to
control the ACK stream. Future work that investigates this scheme
and others in detail is encouraged.
Floyd, et al. Experimental [Page 25]
RFC 4782 Quick-Start for TCP and IP January 2007
4.6. TCP: Responding to a Loss of a Quick-Start Packet
For TCP, we have defined a "Quick-Start packet" as one of the packets
sent in the window immediately following a successful Quick-Start
Request. After detecting the loss or ECN-marking of a Quick-Start
packet, TCP MUST revert to the default congestion control procedures
that would have been used if the Quick-Start Request had not been
approved. For example, if Quick-Start is used for setting the
initial window, and a packet from the initial window is lost or
marked, then the TCP sender MUST then slow-start with the default
initial window that would have been used if Quick-Start had not been
used. In addition to reverting to the default congestion control
mechanisms, the sender MUST take into account that the Quick-Start
congestion window was too large. Thus, the sender SHOULD decrease
ssthresh to, at most, half the number of Quick-Start packets that
were successfully transmitted. Appendix B.5 discusses possible
alternatives in responding to the loss of a Quick-Start packet.
If a Quick-Start packet is lost or ECN-marked, then the sender SHOULD
NOT make future Quick-Start Requests for this connection.
We note that ECN [RFC3168] MAY be used with Quick-Start. As is
always the case with ECN, the sender's congestion control response to
an ECN-marked Quick-Start packet is the same as the response to a
dropped Quick-Start packet, thus reverting to slow start in the case
of Quick-Start packets marked as experiencing congestion.
4.7. TCP: A Quick-Start Request for a Larger Initial Window
Some of the issues of using Quick-Start are related to the specific
scenario in which Quick-Start is used. This section discusses the
following issues that arise when Quick-Start is used by TCP to
request a larger initial window: (1) interactions with Path MTU
Discovery (PMTUD); and (2) Quick-Start Request packets that are
discarded by middleboxes.
4.7.1. Interactions with Path MTU Discovery
One issue when Quick-Start is used to request a large initial window
concerns the interactions between the large initial window and Path
MTU Discovery. Some of the issues are discussed in RFC 3390:
"When larger initial windows are implemented along with Path MTU
Discovery [RFC1191], alternatives are to set the `Don't Fragment'
(DF) bit in all segments in the initial window, or to set the `Don't
Fragment' (DF) bit in one of the segments. It is an open question as
to which of these two alternatives is best."
Floyd, et al. Experimental [Page 26]
RFC 4782 Quick-Start for TCP and IP January 2007
If the sender knows the Path MTU when the initial window is sent
(e.g., from a PMTUD cache or from some other IETF-approved method),
then the sender SHOULD use that MTU for segments in the initial
window. Unfortunately, the sender doesn't necessarily know the Path
MTU when it sends packets in the initial window. In this case, the
sender should be conservative in the packet size used. Sending a
large number of overly large packets with the DF bit set is not
desirable, but sending a large number of packets that are fragmented
in the network can be equally undesirable.
If the sender doesn't know the Path MTU when the initial window is
sent, the sender SHOULD send one large packet in the initial window
with the DF bit set, and send the remaining packets in the initial
window with a smaller MTU of 576 bytes (or 1280 bytes with IPv6).
A second possibility would be for the sender to delay sending the
Quick-Start Request for one round-trip time by sending the Quick-
Start Request with the first window of data, while also doing Path
MTU Discovery.
The sender may be using an iterative approach such as Packetization
Layer Path MTU Discovery (PLPMTUD) [MH06] for Path MTU Discovery,
where the sender tests successively larger MTUs. If a probe is
successfully delivered, then the MTU can be raised to reflect the
value used in that probe. We would note that PLPMTUD does not allow
the sender to determine the Path MTU before sending the initial
window of data.
4.7.2. Quick-Start Request Packets that are Discarded by Routers or
Middleboxes
It is always possible for a TCP SYN packet carrying a Quick-Start
request to be dropped in the network due to congestion, or to be
blocked due to interactions with routers or middleboxes, where a
middlebox is defined as any intermediary box performing functions
apart from normal, standard functions of an IP router on the data
path between a source host and destination host [RFC3234].
Measurement studies of interactions between transport protocols and
middleboxes [MAF04] show that for 70% of the Web servers
investigated, no connection is established if the TCP SYN packet
contains an unknown IP option (and for 43% of the Web servers, no
connection is established if the TCP SYN packet contains an IP
TimeStamp Option). In both cases, this is presumably due to routers
or middleboxes along that path.
If the TCP sender doesn't receive a response to the SYN or SYN/ACK
packet containing the Quick-Start Request, then the TCP sender SHOULD
resend the SYN or SYN/ACK packet without the Quick-Start Request.
Floyd, et al. Experimental [Page 27]
RFC 4782 Quick-Start for TCP and IP January 2007
Similarly, if the TCP sender receives a TCP reset in response to the
SYN or SYN/ACK packet containing the Quick-Start Request, then the
TCP sender SHOULD resend the SYN or SYN/ACK packet without the
Quick-Start Request [RFC3360].
RFCs 1122 and 2988 specify that the sender should set the initial RTO
(retransmission timeout) to three seconds, though many TCP
implementations set the initial RTO to one second. For a TCP SYN
packet sent with a Quick-Start request, the TCP sender SHOULD use an
initial RTO of three seconds.
We note that if the TCP SYN packet is using the IP Quick-Start Option
for a Quick-Start Request, and it is also using bits in the TCP
header to negotiate ECN-capability with the TCP host at the other
end, then the drop of a TCP SYN packet could be due to congestion, a
router or middlebox dropping the packet because of the IP Option, or
a router or middlebox dropping the packet because of the information
in the TCP header negotiating ECN. In this case, the sender could
resend the dropped packet without either the Quick-Start or the ECN
requests. Alternately, the sender could resend the dropped packet
with only the ECN request in the TCP header, resending the TCP SYN
packet without either the Quick-Start or the ECN requests if the
second TCP SYN packet is dropped. The second choice seems
reasonable, given that a TCP SYN packet today is more likely to be
blocked due to policies that discard packets with IP Options than due
to policies that discard packets with ECN requests in the TCP header
[MAF04].
4.8. TCP: A Quick-Start Request in the Middle of a Connection
This section discusses the following issues that arise when Quick-
Start is used by TCP to request a larger window in the middle of a
connection, such as after an idle period: (1) determining the rate to
request; (2) when to make a request; and (3) the response if Quick-
Start packets are dropped.
(1) Determining the rate to request:
For a connection that has not yet had a congestion event (that
is, a marked or dropped packet), the TCP sender is not restricted
in the rate that it requests. As an example, a server might wait
and send a Quick-Start Request after a short interaction with the
client.
To use a Quick-Start Request in a connection that has already
experienced a congestion event, and that has not had a recent
mobility event, the TCP sender can determine the largest
congestion window that the TCP connection achieved since the last
packet drop and translate this to a sending rate to get the
Floyd, et al. Experimental [Page 28]
RFC 4782 Quick-Start for TCP and IP January 2007
maximum allowed request rate. If the request is granted, then
the sender essentially restarts with its old congestion window
from before it was reduced, for example, during an idle period.
A Quick-Start Request sent in the middle of a TCP connection
SHOULD be sent on a data packet.
(2) When to make a request:
A TCP connection MAY make a Quick-Start Request before the
connection has experienced a congestion event, or after an idle
period of at least one RTO.
(3) Response if Quick-Start packets are dropped:
If Quick-Start packets are dropped in the middle of connection,
then the sender MUST revert to half the Quick-Start window, or to
the congestion window that the sender would have used if the
Quick-Start request had not been approved, whichever is smaller.
4.9. An Example Quick-Start Scenario with TCP
The following is an example scenario of when both hosts request
Quick-Start for setting their initial windows. This is similar to
Figures 1 and 2 in Section 2.1, except that it illustrates a TCP
connection with both TCP hosts sending Quick-Start Requests.
* The TCP SYN packet from Host A contains a Quick-Start Request in
the IP header.
* Routers along the forward path modify the Quick-Start Request as
appropriate.
* Host B receives the Quick-Start Request in the SYN packet, and
calculates the TTL Diff. If Host B approves the Quick-Start
Request, then Host B sends a Quick-Start Response in the TCP header
of the SYN/ACK packet. Host B also sends a Quick-Start Request in
the IP header of the SYN/ACK packet.
* Routers along the reverse path modify the Quick-Start Request as
appropriate.
* Host A receives the Quick-Start Response in the SYN/ACK packet, and
checks the TTL Diff, Rate Request, and QS Nonce for validity. If
they are valid, then Host A sets its initial congestion window
appropriately, and sets up rate-based pacing to be used with the
initial window. If the Quick-Start Response is not valid, then
Host A uses TCP's default initial window. In either case, Host A
sends a Report of Approved Rate.
Floyd, et al. Experimental [Page 29]
RFC 4782 Quick-Start for TCP and IP January 2007
Host A also calculates the TTL Diff for the Quick-Start Request in
the incoming SYN/ACK packet, and sends a Quick-Start Response in
the TCP header of the ACK packet.
* Host B receives the Quick-Start Response in an ACK packet, and
checks the TTL Diff, Rate Request, and QS Nonce for validity. If
the Quick-Start Response is valid, then Host B sets its initial
congestion window appropriately, and sets up rate-based pacing to
be used with its initial window. If the Quick-Start Response is
not valid, then Host B uses TCP's default initial window. In
either case, Host B sends a Report of Approved Rate.
5. Quick-Start and IPsec AH
This section shows that Quick-Start is compatible with IPsec
Authentication Header (AH). AH uses an Integrity Check Value (ICV)
in the IPsec Authentication Header to verify both message
authentication and integrity [RFC4302]. Changes to the Quick-Start
Option in the IP header do not affect this AH ICV. The tunnel
considerations in Section 6 below apply to all IPsec tunnels,
regardless of what IPsec headers or processing are used in
conjunction with the tunnel.
Because the contents of the Quick-Start Option can change along the
path, it is important that these changes not affect the IPsec
Authentication Header Integrity Check Value (AH ICV). For IPv4, RFC
4302 requires that unrecognized IPv4 options be zeroed for AH ICV
computation purposes, so Quick-Start IP Option data changing en route
does not cause problems with existing IPsec AH implementations for
IPv4. If the Quick-Start Option is recognized, it MUST be treated as
a mutable IPv4 option, and hence be completely zeroed for AH ICV
calculation purposes. IPv6 option numbers explicitly indicate
whether the option is mutable; the third-highest order bit in the
IANA-allocated option type has the value 1 to indicate that the
Quick-Start Option data can change en route. RFC 4302 requires that
the option data of any such option be zeroed for AH ICV computation
purposes. Therefore, changes to the Quick-Start Option in the IP
header do not affect the calculation of the AH ICV.
Floyd, et al. Experimental [Page 30]
RFC 4782 Quick-Start for TCP and IP January 2007
6. Quick-Start in IP Tunnels and MPLS
This section considers interactions between Quick-Start and IP
tunnels, including IPsec ([RFC4301]), IP in IP [RFC2003], GRE
[RFC2784], and others. This section also considers interactions
between Quick-Start and MPLS [RFC3031].
In the discussion, we use TTL Diff, defined earlier as the difference
between the IP TTL and the Quick-Start TTL, mod 256. Recall that the
sender considers the Quick-Start Request approved only if the value
of TTL Diff for the packet entering the network is the same as the
value of TTL Diff for the packet exiting the network.
Simple tunnels: IP tunnel modes are generally based on adding a new
"outer" IP header that encapsulates the original or "inner" IP header
and its associated packet. In many cases, the new "outer" IP header
may be added and removed at intermediate points along a path,
enabling the network to establish a tunnel without requiring endpoint
participation. We denote tunnels that specify that the outer header
be discarded at tunnel egress as "simple tunnels", and we denote
tunnels where the egress saves and uses information from the outer
header before discarding it as "non-simple tunnels". An example of a
"non-simple tunnel" would be a tunnel configured to support ECN,
where the egress router might copy the ECN codepoint in the outer
header to the inner header before discarding the outer header
[RFC3168].
__ Tunnels Compatible with Quick-Start
/
Simple Tunnels __/
\
\__ Tunnels Not Compatible with Quick-Start
(False Positives!)
__ Tunnels Supporting Quick-Start
/
/
Non-Simple Tunnels __/_____ Tunnels Compatible with Quick-Start,
\ but Not Supporting Quick-Start
\
\__ Tunnels Not Compatible with Quick-Start?
Figure 6: Categories of Tunnels.
Floyd, et al. Experimental [Page 31]
RFC 4782 Quick-Start for TCP and IP January 2007
Tunnels that are compatible with Quick-Start: We say that an IP
tunnel `is not compatible with Quick-Start' if the use of a Quick-
Start Request over such a tunnel allows false positives, where the
TCP sender incorrectly believes that the Quick-Start Request was
approved by all routers along the path. If the use of Quick-Start
over the tunnel does not cause false positives, we say that the IP
tunnel `is compatible with Quick-Start'.
If the IP TTL of the inner header is decremented during forwarding
before tunnel encapsulation takes place, then the simple tunnel is
compatible with Quick-Start, with Quick-Start Requests being
rejected. Section 6.1 describes in more detail the ways that a
simple tunnel can be compatible with Quick-Start.
There are some simple tunnels that are not compatible with Quick-
Start, allowing `false positives' where the TCP sender incorrectly
believes that the Quick-Start Request was approved by all routers
along the path. This is discussed in Section 6.2 below.
One of our tasks in the future will be to investigate the occurrence
of tunnels that are not compatible with Quick-Start, and to track the
extent to which such tunnels are modified over time. The evaluation
of the problem of false positives from tunnels that are not
compatible with Quick-Start will affect the progression of Quick-
Start from Experimental to Proposed Standard, and will affect the
degree of deployment of Quick-Start while in Experimental mode.
Tunnels that support Quick-Start: We say that an IP tunnel `supports
Quick-Start' if it allows routers along the tunnel path to process
the Quick-Start Request and give feedback, resulting in the
appropriate possible acceptance of the Quick-Start Request. Some
tunnels that are compatible with Quick-Start support Quick-Start,
while others do not. We note that a simple tunnel is not able to
support Quick-Start.
From a security point of view, the use of Quick-Start in the outer
header of an IP tunnel might raise security concerns because an
adversary could tamper with the Quick-Start information that
propagates beyond the tunnel endpoint, or because the Quick-Start
Option exposes information to network scanners. Our approach is to
make supporting Quick-Start an option for IP tunnels. That is, in
environments or tunneling protocols where the risks of using Quick-
Start are judged to outweigh its benefits, the tunnel can simply
delete the Quick-Start Option or zero the Quick-Start rate request
and QS TTL fields before encapsulation. The result is that there are
two viable options for IP tunnels to be compatible with Quick-Start.
The first option is the simple tunnel described above and in Section
6.1, where the tunnel is compatible with Quick-Start but does not
Floyd, et al. Experimental [Page 32]
RFC 4782 Quick-Start for TCP and IP January 2007
support Quick-Start, where all Quick-Start Requests along the path
will be rejected. The second approach is a Quick-Start-capable mode,
described in Section 6.3, where the tunnel actively supports Quick-
Start.
6.1. Simple Tunnels that Are Compatible with Quick-Start
This section describes the ways that a simple tunnel can be
compatible with Quick-Start but not support Quick-Start, resulting in
the rejection of all Quick-Start Requests that traverse the tunnel.
If the tunnel ingress for the simple tunnel is at a router, the IP
TTL of the inner header is generally decremented during forwarding
before tunnel encapsulation takes place. In this case, TTL Diff will
be changed, correctly causing the Quick-Start Request to be rejected.
For a simple tunnel, it is preferable if the Quick-Start Request is
not copied to the outer header, saving the routers within the tunnel
from unnecessarily processing the Quick-Start Request. However, the
Quick-Start Request will be rejected correctly in this case whether
or not the Quick-Start Request is copied to the outer header.
6.1.1. Simple Tunnels that Are Aware of Quick-Start
If a tunnel ingress is aware of Quick-Start, but does not want to
support Quick-Start, then the tunnel ingress MUST either zero the
Quick-Start rate request, QS TTL, and QS Nonce fields, or remove the
Quick-Start Option from the inner header before encapsulation.
Section 6.3 describes the procedures for a tunnel that does want to
support Quick-Start.
Deleting the Quick-Start Option or zeroing the Quick-Start rate
request *after decapsulation* also serves to prevent the propagation
of Quick-Start information, and is compatible with Quick-Start. If
the outer header does not contain a Quick-Start Request, a Quick-
Start-aware tunnel egress MUST reject the inner Quick-Start Request
by zeroing the Rate Request field in the inner header, or by deleting
the Quick-Start Option.
If the tunnel ingress is at a sending host or router where the IP TTL
is not decremented prior to encapsulation, and neither tunnel
endpoint is aware of Quick-Start, then this allows false positives,
described in the section below.
Floyd, et al. Experimental [Page 33]
RFC 4782 Quick-Start for TCP and IP January 2007
6.2. Simple Tunnels that Are Not Compatible with Quick-Start
Sometimes a tunnel implementation that does not support Quick-Start
is independent of the TCP sender or a router implementation that
supports Quick-Start. In these cases, it is possible that a Quick-
Start Request gets erroneously approved without the routers in the
tunnel having individually approved the request, causing a false
positive.
If a tunnel ingress is a separate component from the TCP sender or IP
forwarding, it is possible that a packet with a Quick-Start option is
encapsulated without the IP TTL being decremented first, or with both
IP TTL and QS TTL being decremented before the tunnel encapsulation
takes place. If the tunnel ingress does not know about Quick-Start,
a valid Quick-Start Request with unchanged TTL Diff traverses in the
inner header, while the outer header most likely does not carry a
Quick-Start Request. If the tunnel egress also does not support
Quick-Start, it remains possible that the Quick-Start Request would
be falsely approved, because the packet is decapsulated using the
Quick-Start Request from the inner header, and the value of TTL Diff
echoed to the sender remains unchanged. For example, such a scenario
can occur with a Bump-In-The-Stack (BITS), an IPsec encryption
implementation where the data encryption occurs between the network
drivers and the TCP/IP protocol stack [RFC4301].
As one example, if a remote access VPN client uses a BITS structure,
then Quick-Start obstacles between the client and the VPN gateway
won't be seen. This is a particular problem because the path between
the client and the VPN gateway is likely to contain the most
congested part of the path. Because most VPN clients are reported to
use BITS [H05], we will explore this in more detail.
A Bump-In-The-Wire (BITW) is an IPsec encryption implementation where
the encryption occurs on an outboard processor, offloading the
encryption processing overhead from the host or router [RFC4301].
The BITW device is usually IP addressable, which means that the IP
TTL is decremented before the packet is passed to the BITW. If the
QS TTL is not decremented, then the value of TTL Diff is changed, and
the Quick-Start Request will be denied. However, if the BITW
supports a host and does not have its own IP address, then the IP TTL
is not decremented before the packet is passed from the host to the
BITW, and a false positive could occur.
Other tunnels that need to be looked at are IP tunnels over non-
network protocols, such as IP over TCP and IP over UDP [RFC3948], and
tunnels using the Layer Two Tunneling Protocol [RFC2661].
Floyd, et al. Experimental [Page 34]
RFC 4782 Quick-Start for TCP and IP January 2007
Section 9.2 discusses the related issue of non-IP queues, such as
layer-two Ethernet or ATM (Asynchronous Transfer Mode) networks, as
another instance of possible bottlenecks that do not participate in
the Quick-Start feedback.
6.3. Tunnels That Support Quick-Start
This section discusses tunnels configured to support Quick-Start.
If the tunnel ingress node chooses to locally approve the Quick-
Start Request, then the ingress node MUST decrement the Quick-Start
TTL at the same time it decrements the IP TTL, and MUST copy IP TTL
and the Quick-Start Option from the inner IP header to the outer
header. During encapsulation, the tunnel ingress MUST zero the
Quick-Start rate request field in the inner header to ensure that the
Quick-Start Request will be rejected if the tunnel egress does not
support Quick-Start.
If the tunnel ingress node does not choose to locally approve the
Quick-Start Request, then it MUST either delete the Quick-Start
option from the inner header before encapsulation, or zero the QS TTL
and the Rate Request fields before encapsulation.
Upon decapsulation, if the outer header contains a Quick-Start
option, the tunnel egress MUST copy the IP TTL and the Quick-Start
option from the outer IP header to the inner header.
IPsec uses the IKE (Internet Key Exchange) Protocol for security
associations. We do not consider the interactions between Quick-
Start and IPsec with IKEv1 [RFC2409] in this document. Now that the
RFC for IKEv2 [RFC4306] is published, we plan to specify a
modification of IPsec to allow the support of Quick-Start to be
negotiated; this modification will specify the negotiation between
tunnel endpoints to allow or forbid support for Quick-Start within
the tunnel. This was done for ECN for IPsec tunnels, with IKEv1
[RFC3168, Section 9.2]. This negotiation of Quick-Start capability
in an IPsec tunnel will be specified in a separate IPsec document.
This document will also include a discussion of the potential effects
of an adversary's modifications of the Quick-Start field (as in
Sections 18 and 19 of RFC 3168), and of the security considerations
of exposing the Quick-Start rate request to network scanners.
6.4. Quick-Start and MPLS
The behavior of Quick-Start with MPLS is similar to the behavior of
Quick-Start with IP Tunnels. For those MPLS paths where the IP TTL
is decremented as part of traversing the MPLS path, these paths are
compatible with Quick-Start, but do not support Quick-Start; Quick-
Floyd, et al. Experimental [Page 35]
RFC 4782 Quick-Start for TCP and IP January 2007
Start Requests that are traversing these paths will be correctly
understood by the transport sender as having been denied. Any MPLS
paths where the IP TTL is not decremented as part of traversing the
MPLS path would be not compatible with Quick-Start; such paths would
result in false positives, where the TCP sender incorrectly believes
that the Quick-Start Request was approved by all routers along the
path.
For cases where the ingress node to the MPLS path is aware of Quick-
Start, this node should either zero the Quick-Start rate request, QS
TTL, and QS Nonce fields, or remove the Quick-Start Option from the
IP header.
7. The Quick-Start Mechanism in Other Transport Protocols
The section earlier specified the use of Quick-Start in TCP. In this
section, we generalize this to give guidelines for the use of Quick-
Start with other transport protocols. We also discuss briefly how
Quick-Start could be specified for other transport protocols.
The general guidelines for Quick-Start in transport protocols are as
follows:
* Quick-Start is only specified for unicast transport protocols with
appropriate congestion control mechanisms. Note: Quick-Start is
not a replacement for standard congestion control techniques, but
meant to augment their operation.
* A transport-level mechanism is needed for the Quick-Start Response
from the receiver to the sender. This response contains the Rate
Request, TTL Diff, and QS Nonce.
* The sender checks the validity of the Quick-Start Response.
* The sender has an estimate of the round-trip time, and translates
the Quick-Start Response into an allowed window or allowed sending
rate. The sender sends a Report of the Approved Rate. The sender
starts sending Quick-Start packets, rate-paced out at the approved
sending rate.
* After the sender receives the first acknowledgement packet for a
Quick-Start packet, no more Quick-Start packets are sent. The
sender adjusts its current congestion window or sending rate to be
consistent with the actual amount of data that was transmitted in
that round-trip time.
Floyd, et al. Experimental [Page 36]
RFC 4782 Quick-Start for TCP and IP January 2007
* When the last Quick-Start packet is acknowledged, the sender
continues using the standard congestion control mechanisms of that
protocol.
* If one of the Quick-Start packets is lost, then the sender reverts
to the standard congestion control method of that protocol that
would have been used if the Quick-Start Request had not been
approved. In addition, the sender takes into account the
information that the Quick-Start congestion window was too large
(e.g., by decreasing ssthresh in TCP).
8. Using Quick-Start
8.1. Determining the Rate to Request
As discussed in [SAF06], the data sender does not necessarily have
information about the size of the data transfer at connection
initiation; for example, in request-response protocols such as HTTP,
the server doesn't know the size or name of the requested object
during connection initiation. [SAF06] explores some of the
performance implications of overly large Quick-Start Requests, and
discusses heuristics that end-nodes could use to size their requests
appropriately. For example, the sender might have information about
the bandwidth of the last-mile hop, the size of the local socket
buffer, or of the TCP receive window, and could use this information
in determining the rate to request. Web servers that mostly have
small objects to transfer might decide not to use Quick-Start at all,
since Quick-Start would be of little benefit to them.
Quick-Start will be more effective if Quick-Start Requests are not
larger than necessary; every Quick-Start Request that is approved but
not used (or not fully used) takes away from the bandwidth pool
available for granting successive Quick-Start Requests.
8.2. Deciding the Permitted Rate Request at a Router
In this section, we briefly outline how a router might decide whether
or not to approve a Quick-Start Request. The router should ask the
following questions:
* Has the router's output link been underutilized for some time
(e.g., several seconds)?
* Would the output link remain underutilized if the arrival rate were
to increase by the aggregate rate requests that the router has
approved over the last fraction of a second?
Floyd, et al. Experimental [Page 37]
RFC 4782 Quick-Start for TCP and IP January 2007
In order to answer the last question, the router must have some
knowledge of the available bandwidth on the output link and of the
Quick-Start bandwidth that could arrive due to recently approved
Quick-Start Requests. In this way, if an underutilized router
experiences a flood of Quick-Start Requests, the router can begin to
deny Quick-Start Requests while the output link is still
underutilized.
A simple way for the router to keep track of the potential bandwidth
from recently approved requests is to maintain two counters: one for
the total aggregate Rate Requests that have been approved in the
current time interval [T1, T2], and one for the total aggregate Rate
Requests approved over a previous time interval [T0, T1]. However,
this document doesn't specify router algorithms for approving Quick-
Start Requests, or make requirements for the appropriate time
intervals for remembering the aggregate approved Quick-Start
bandwidth. A possible router algorithm is given in Appendix E, and
more discussion of these issues is available in [SAF06].
* If the router's output link has been underutilized and the
aggregate of the Quick-Start Request Rate options granted is low
enough to prevent a near-term bandwidth shortage, then the router
could approve the Quick-Start Request.
Section 10.2 discusses some of the implementation issues in
processing Quick-Start Requests at routers. [SAF06] discusses the
range of possible Quick-Start algorithms at the router for deciding
whether to approve a Quick-Start Request. In order to explore the
limits of the possible functionality at routers, [SAF06] also
discusses Extreme Quick-Start mechanisms at routers, where the router
would keep per-flow state concerning approved Quick-Start requests.
9. Evaluation of Quick-Start
9.1. Benefits of Quick-Start
The main benefit of Quick-Start is the faster start-up for the
transport connection itself. For a small TCP transfer of one to five
packets, Quick-Start is probably of very little benefit; at best, it
might shorten the connection lifetime from three to two round-trip
times (including the round-trip time for connection establishment).
Similarly, for a very large transfer, where the slow-start phase
would have been only a small fraction of the connection lifetime,
Quick-Start would be of limited benefit. Quick-Start would not
significantly shorten the connection lifetime, but it might eliminate
or at least shorten the start-up phase. However, for moderate-sized
connections in a well-provisioned environment, Quick-Start could
possibly allow the entire transfer of M packets to be completed in
Floyd, et al. Experimental [Page 38]
RFC 4782 Quick-Start for TCP and IP January 2007
one round-trip time (after the initial round-trip time for the SYN
exchange), instead of the log_2(M)-2 round-trip times that it would
normally take for the data transfer, in an uncongested environments
(assuming an initial window of four packets).
9.2. Costs of Quick-Start
This section discusses the costs of Quick-Start for the connection
and for the routers along the path.
The cost of having a Quick-Start Request packet dropped:
Measurement studies cited earlier [MAF04] suggest that on a wide
range of paths in the Internet, TCP SYN packets containing unknown IP
options will be dropped. Thus, for the sender one risk in using
Quick-Start is that the packet carrying the Quick-Start Request could
be dropped in the network. It is particularly costly to the sender
when a TCP SYN packet is dropped, because in this case the sender
should wait for an RTO of three seconds before re-sending the SYN
packet, as specified in Section 4.7.2.
The cost of having a Quick-Start data packet dropped:
Another risk for the sender in using Quick-Start lies in the
possibility of suffering from congestion-related losses of the
Quick-Start data packets. This should be an unlikely situation
because routers are expected to approve Quick-Start Requests only
when they are significantly underutilized. However, a transient
increase in cross-traffic in one of the routers, a sudden decrease in
available bandwidth on one of the links, or congestion at a non-IP
queue could result in packet losses even when the Quick-Start Request
was approved by all of the routers along the path. If a Quick-Start
packet is dropped, then the sender reverts to the congestion control
mechanisms it would have used if the Quick-Start Request had not been
approved, so the performance cost to the connection of having a
Quick-Start packet dropped is small, compared to the performance
without Quick-Start. (On the other hand, the performance difference
between Quick-Start with a Quick-Start packet dropped and Quick-
Start with no Quick-Start packet dropped can be considerable.)
Added complexity at routers:
The main cost of Quick-Start at routers concerns the costs of added
complexity. The added complexity at the end-points is moderate, and
might easily be outweighed by the benefit of Quick-Start to the end
hosts. The added complexity at the routers is also somewhat
moderate; it involves estimating the unused bandwidth on the output
link over the last several seconds, processing the Quick-Start
request, and keeping a counter of the aggregate Quick-Start rate
approved over the last fraction of a second. However, this added
complexity at routers adds to the development cycle, and could
Floyd, et al. Experimental [Page 39]
RFC 4782 Quick-Start for TCP and IP January 2007
prevent the addition of other competing functionality to routers.
Thus, careful thought would have to be given to the addition of
Quick-Start to IP.
The slow path in routers:
Another drawback of Quick-Start is that packets containing the
Quick-Start Request message might not take the fast path in routers,
particularly in the beginning of Quick-Start's deployment in the
Internet. This would mean some extra delay for the end hosts, and
extra processing burden for the routers. However, as discussed in
Sections 4.1 and 4.7, not all packets would carry the Quick-Start
option. In addition, for the underutilized links where Quick-Start
Requests could actually be approved, or in typical environments where
most of the packets belong to large flows, the burden of the Quick-
Start Option on routers would be considerably reduced. Nevertheless,
it is still conceivable, in the worst case, that many packets would
carry Quick-Start Requests; this could slow down the processing of
Quick-Start packets in routers considerably. As discussed in Section
9.6, routers can easily protect against this by enforcing a limit on
the rate at which Quick-Start Requests will be considered. [RW03]
and [RW04] contain measurements of the impact of IP Option Processing
on packet round-trip times.
Multiple paths:
One limitation of Quick-Start is that it presumes that the data
packets of a connection will follow the same path as the Quick-Start
request packet. If this is not the case, then the connection could
be sending the Quick-Start packets, at the approved rate, along a
path that was already congested, or that became congested as a result
of this connection. Thus, Quick-Start could give poor performance
when there is a routing change immediately after the Quick-Start
Request is approved, and the Quick-Start data packets follow a
different path from that of the original Quick-Start Request. This
is, however, similar to what would happen for a connection with
sufficient data, if the connection's path was changed in the middle
of the connection, which had already established the allowed initial
rate.
As specified in Section 3.3, a router that uses multipath routing for
packets within a single connection must not approve a Quick-Start
Request. Quick-Start would not perform robustly in an environment
with multipath routing, where different packets in a connection
routinely follow different paths. In such an environment, the
Quick-Start Request and some fraction of the packets in the
connection might take an underutilized path, while the rest of the
packets take an alternate, congested path.
Floyd, et al. Experimental [Page 40]
RFC 4782 Quick-Start for TCP and IP January 2007
Non-IP queues:
A problem of any mechanism for feedback from routers at the IP level
is that there can be queues and bottlenecks in the end-to-end path
that are not in IP-level routers. As an example, these include
queues in layer-two Ethernet or ATM networks. One possibility would
be that an IP-level router adjacent to such a non-IP queue or
bottleneck would be configured to reject Quick-Start Requests if that
was appropriate. One would hope that, in general, IP networks are
configured so that non-IP queues between IP routers do not end up
being the congested bottlenecks.
9.3. Quick-Start with QoS-Enabled Traffic
The discussion in this document has largely been of Quick-Start with
default, best-effort traffic. However, Quick-Start could also be
used by traffic using some form of differentiated services, and
routers could take the traffic class into account when deciding
whether or not to grant the Quick-Start Request. We don't address
this context further in this paper, since it is orthogonal to the
specification of Quick-Start.
Routers are also free to take into account their own priority
classifications in processing Quick-Start Requests.
9.4. Protection against Misbehaving Nodes
In this section, we discuss the protection against senders,
receivers, or colluding routers or middleboxes lying about the
Quick-Start Request.
9.4.1. Misbehaving Senders
A transport sender could try to transmit data at a higher rate than
that approved in the Quick-Start Request. The network could use a
traffic policer to protect against misbehaving senders that exceed
the approved rate, for example, by dropping packets that exceed the
allowed transmission rate. The required Report of Approved Rate
allows traffic policers to check that the Report of Approved Rate
does not exceed the Rate Request actually approved at that point in
the network in the previous Quick-Start Request from that connection.
The required Approved Rate report also allows traffic policers to
check that the sender's sending rate does not exceed the rate in the
Report of Approved Rate.
If a router or receiver receives an Approved Rate report that is
larger than the Rate Request in the Quick-Start Request approved for
that sender for that connection in the previous round-trip time, then
the router or receiver could deny future Quick-Start Requests from
Floyd, et al. Experimental [Page 41]
RFC 4782 Quick-Start for TCP and IP January 2007
that sender, e.g., by deleting the Quick-Start Request from future
packets from that sender. We note that routers are not required to
use Approved Rate reports to check if senders are cheating; this is
at the discretion of the router.
If a router sees a Report of Approved Rate, and did not see an
earlier Quick-Start Request, then either the sender could be
cheating, or the connection's path could have changed since the
Quick-Start Request was sent. In either case, the router could
decide to deny future Quick-Start Requests for this connection. In
particular, it is reasonable for the router to deny a Quick-Start
request if either the sender is cheating, or if the connection path
suffers from path changes or multipathing.
If a router approved a Quick-Start Request, but does not see a
subsequent Approved Rate report, then there are several
possibilities: (1) the request was denied and/or dropped downstream,
and the sender did not send a Report of Approved Rate; (2) the
request was approved, but the sender did not send a Report of
Approved Rate; (3) the Approved Rate report was dropped in the
network; or (4) the Approved Rate report took a different path from
the Quick-Start Request. In any of these cases, the router would be
justified in denying future Quick-Start Requests for this connection.
In any of the cases mentioned in the three paragraphs above (i.e., an
Approved Rate report that is larger than the Rate Request in the
earlier Quick-Start Request, a Report of Approved Rate with no
preceding Rate Request, or a Rate Request with no Report of Approved
Rate), a traffic policer may assume that Quick-Start is not being
used appropriately, or is being used in an unsuitable environment
(e.g., with multiple paths), and take some corresponding action.
What are the incentives for a sender to cheat by over-sending after a
Quick-Start Request? Assuming that the sender's interests are
measured by a performance metric such as the completion time for its
connections, sometimes it might be in the sender's interests to
cheat, and sometimes it might not; in some cases, it could be
difficult for the sender to judge whether it would be in its
interests to cheat. The incentives for a sender to cheat by over-
sending after a Quick-Start Request are not that different from the
incentives for a sender to cheat by over-sending even in the absence
of Quick-Start, with one difference: the use of Quick-Start could
help a sender evade policing actions from policers in the network.
The Report of Approved Rate is designed to address this and to make
it harder for senders to use Quick-Start to `cover' their cheating.
Floyd, et al. Experimental [Page 42]
RFC 4782 Quick-Start for TCP and IP January 2007
9.4.2. Receivers Lying about Whether the Request was Approved
One form of misbehavior would be for the receiver to lie to the
sender about whether the Quick-Start Request was approved, by falsely
reporting the TTL Diff and QS Nonce. If a router that understands
the Quick-Start Request denies the request by deleting the request or
by zeroing the QS TTL and QS Nonce, then the receiver can "lie" about
whether the request was approved only by successfully guessing the
value of the TTL Diff and QS Nonce to report. The chance of the
receiver successfully guessing the correct value for the TTL Diff is
1/256, and the chance of the receiver successfully guessing the QS
nonce for a reported rate request of K is 1/(2K).
However, if the Quick-Start Request is denied only by a non-Quick-
Start-capable router, or by a router that is unable to zero the QS
TTL and QS Nonce fields, then the receiver could lie about whether
the Quick-Start Requests were approved by modifying the QS TTL in
successive requests received from the same host. In particular, if
the sender does not act on a Quick-Start Request, then the receiver
could decrement the QS TTL by one in the next request received from
that host before calculating the TTL Diff, and decrement the QS TTL
by two in the following received request, until the sender acts on
one of the Quick-Start Requests.
Unfortunately, if a router doesn't understand Quick-Start, then it is
not possible for that router to take an active step such as zeroing
the QS TTL and QS Nonce to deny a request. As a result, the QS TTL
is not a fail-safe mechanism for preventing lying by receivers in the
case of non-Quick-Start-capable routers.
What would be the incentives for a receiver to cheat in reporting on
a Quick-Start Request, in the absence of a mechanism such as the QS
Nonce? In some cases, cheating would be of clear benefit to the
receiver, resulting in a faster completion time for the transfer. In
other cases, where cheating would result in Quick-Start packets being
dropped in the network, cheating might or might not improve the
receiver's performance metric, depending on the details of that
particular scenario.
9.4.3. Receivers Lying about the Approved Rate
A second form of receiver misbehavior would be for the receiver to
lie to the sender about the Rate Request for an approved Quick-Start
Request, by increasing the value of the Rate Request field. However,
the receiver doesn't necessarily know the Rate Request in the
original Quick-Start Request sent by the sender, and a higher Rate
Request reported by the receiver will only be considered valid by the
sender if it is no higher than the Rate Request originally requested
Floyd, et al. Experimental [Page 43]
RFC 4782 Quick-Start for TCP and IP January 2007
by the sender. For example, if the sender sends a Quick-Start
Request with a Rate Request of X, and the receiver reports receiving
a Quick-Start Request with a Rate Request of Y > X, then the sender
knows that either some router along the path malfunctioned
(increasing the Rate Request inappropriately), or the receiver is
lying about the Rate Request in the received packet.
If the sender sends a Quick-Start Request with a Rate Request of Z,
the receiver receives the Quick-Start Request with an approved Rate
Request of X, and reports a Rate Request of Y, for X < Y <= Z, then
the receiver only succeeds in lying to the sender about the approved
rate if the receiver successfully reports the rightmost 2Y bits in
the QS nonce.
If senders often use a configured default value for the Rate Request,
then receivers would often be able to guess the original Rate
Request, and this would make it easier for the receiver to lie about
the value of the Rate Request field. Similarly, if the receiver
often communicates with a particular sender, and the sender always
uses the same Rate Request for that receiver, then the receiver might
over time be able to infer the original Rate Request used by the
sender.
There are several possible additional forms of protection against
receivers lying about the value of the Rate Request. One possible
additional protection would be for a router that decreases a Rate
Request in a Quick-Start Request to report the decrease directly to
the sender. However, this could lead to many reports back to the
sender for a single request, and could also be used in address-
spoofing attacks.
A second limited form of protection would be for senders to use some
degree of randomization in the requested Rate Request, so that it is
difficult for receivers to guess the original value for the Rate
Request. However, this is difficult because there is a fairly coarse
granularity in the set of rate requests available to the sender, and
randomizing the initial request only offers limited protection, in
any case.
9.4.4. Collusion between Misbehaving Routers
In addition to protecting against misbehaving receivers, it is
necessary to protect against misbehaving routers. Consider collusion
between an ingress router and an egress router belonging to the same
intranet. The ingress router could decrement the Rate Request at the
ingress, with the egress router increasing it again at the egress.
The routers between the ingress and egress that approved the
Floyd, et al. Experimental [Page 44]
RFC 4782 Quick-Start for TCP and IP January 2007
decremented rate request might not have been willing to approve the
larger, original request.
Another form of collusion would be for the ingress router to inform
the egress router out-of-band of the TTL Diff and QS Nonce for the
request packet at the ingress. This would enable the egress router
to modify the QS TTL and QS N |
