| Internet-Draft | QUIC Accurate ECN Acknowledgements | May 2024 |
| Seemann & Goel | Expires 24 November 2024 | [Page] |
QUIC defines a variant of the ACK frame that carries cumulative count for each of the three ECN codepoints (ECT(1), ECT(0) and CE). From this information, the recipient of the ACK frame cannot deduce which ECN marking the individual packets were received with.¶
This document defines an alternative ACK frame that encodes enough information to indicate which ECN mark each individual packet was received with. This information is essential for accurately performing adjustments to congestion window and sending rate of the sender.¶
This note is to be removed before publishing as an RFC.¶
The latest revision of this draft can be found at https://marten-seemann.github.io/draft-seemann-quic-accurate-ack-ecn/draft-seemann-quic-accurate-ack-ecn.html. Status information for this document may be found at https://datatracker.ietf.org/doc/draft-seemann-quic-accurate-ack-ecn/.¶
Discussion of this document takes place on the QUIC Working Group mailing list (mailto:quic@ietf.org), which is archived at https://mailarchive.ietf.org/arch/browse/quic/. Subscribe at https://www.ietf.org/mailman/listinfo/quic/.¶
Source for this draft and an issue tracker can be found at https://github.com/marten-seemann/draft-seemann-quic-accurate-ack-ecn.¶
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.¶
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.¶
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."¶
This Internet-Draft will expire on 24 November 2024.¶
Copyright (c) 2024 IETF Trust and the persons identified as the document authors. All rights reserved.¶
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶
Some congestion control algorithms would benefit from not only knowing that some packets were marked with Congestion Experienced (CE) bit, but exactly which ones. In the general case, this is not possible with the standard [RFC9000] ACK frame, since it only contains cumulative ECN counts. This information is helpful to congestion control algorithms in following ways:¶
To perform additive increase and multiplicative decrease accurately, it is important to know the exact sequence of CE marked and non-CE marked packets, as the sequence determines how the congestion was experienced at the bottleneck.¶
Some congestion control algorithms (for example L4S congestion controllers, see [RFC9330]) would benefit from knowing exactly which packets were not CE-marked. These algorithms apply an additive increase for non-CE marked packets even if some other packets were CE marked in the same round trip.¶
This document defines an alternative ACK frame, the ACCURATE_ACK_ECN frame, which encodes the corresponding ECN codepoint alongside the ACK range. This encoding comes at a cost: In the presence of ECN markings, this will lead to ACCURATE_ACK_ECN frames containing more ACK ranges compared to a regular ACK frame. However, this is not expected to significantly inflate the size of ACCURATE_ACK_ECN frames. For example, in the steady state, L4S [RFC9331] applies the CE marking to two packets per roundtrip. In the absence of packet loss, two of the ACCURATE_ACK_ECN frames sent during that RTT would contain two ACK ranges instead of one.¶
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
The ACCURATE_ACK_ECN frame looks similar to an [RFC9000] ACK frame. It uses a different encoding for ACK ranges (see Section 3.1 and Section 3.2).¶
ACCURATE_ACK_ECN Frame {
Type (i) = 0x2051a5fa,
Largest Acknowledged (i),
ACK Delay (i),
ACK Range Count (i),
First ACK Range (..),
ACK Range (..) ...,
}
¶
Except for the two ACK Range fields, all the fields are the same as defined in Section 19.3 of [RFC9000].¶
All packets within an ACK range MUST have been received with the same ECN code point. If a range of packets with contiguous packet numbers, but different ECN markings is received, this MUST be reported using multiple ACK ranges.¶
Similar to regular ACK frames, ACCURATE_ACK_ECN frames are not ack-eliciting (see Section 13.2 of [RFC9000]), nor are they congestion-controlled.¶
First ACK Range {
ACK Range Length (i),
ECN Marking (8),
}
¶
A variable-length integer indicating the number of contiguous packets preceding the Largest Acknowledged that are being acknowledged with the same ECN code point. That is, the smallest packet acknowledged in the range with the same ECN code point is determined by subtracting the First ACK Range Length value from the Largest Acknowledged field.¶
The ECN code point all packets in this range were received with: Non-ECT is encoded as 0, ECT(1) as 1, ECT(0) as 2 and CE as 3. Values larger than or equal to 4 are invalid, and the receiver MUST close the connection with a FRAME_ENCODING_ERROR if it receives an ACK range with an invalid ECN marking value.¶
Each ACK Range consists of alternating Gap, ACK Range Length and ECN Marking values in descending packet number order. ACK Ranges can be repeated. The number of ranges is determined by the ACK Range Count field; one of each value is present for each value in the ACK Range Count field.¶
ACK Ranges are structured as shown in Figure 1.¶
ACK Range {
Gap (i),
ACK Range Length (i),
ECN Marking (8),
}
The fields that form each ACK Range are:¶
A variable-length integer indicating the number of contiguous unacknowledged packets preceding the packet number given by the smallest in the preceding ACK Range. Note that this definition differs by one from the Gap definition of the standard QUIC ACK frame in Section 19.3.1 of [RFC9000]. This is necessary to allow encoding of contiguous ranges of packet numbers that were received with different ECN markings.¶
A variable-length integer indicating the number of contiguous acknowledged packets preceding the largest packet number, as determined by the preceding Gap.¶
The ECN code point all packets in this range were received with, as defined in Section 3.1.¶
As described in Section 19.3.1 of [RFC9000], given a largest packet number for an ACK range, the smallest value is determined by:¶
smallest = largest - ack_range¶
To calculate the largest value for a subsequent ACK Range, the formula differs from the standard QUIC ACK frame which can be calculated using:¶
largest = previous_smallest - gap - 1¶
If any computed packet number is negative, an endpoint MUST generate a connection error of type FRAME_ENCODING_ERROR.¶
Consider a scenario where 10 packets (from packet number 1 to 10) were sent with ECT(1) but receiver received a total of 9 packets where packet number 8 was lost and packet number 6 and 9 were CE marked. The ACCURATE_ACK_ECN frame would look like below.¶
ACCURATE_ACK_ECN Frame {
Type (i) = 0x2051a5fa,
Largest Acknowledged (10),
ACK Delay (i),
ACK Range Count (4),
First ACK Range {
ACK Range Length (0),
ECN Marking (1),
},
ACK Range {
Gap (0),
ACK Range Length (0),
ECN Marking (3),
}
ACK Range {
Gap (1),
ACK Range Length (0),
ECN Marking (1),
}
ACK Range {
Gap (0),
ACK Range Length (0),
ECN Marking (3),
}
ACK Range {
Gap (0),
ACK Range Length (4),
ECN Marking (1),
}
}
¶
Endpoints advertise their support of the extension by sending the accurate_ack_ecn (0x2051a5fa8648af) transport parameter (Section 7.4 of [RFC9000]) with an empty value. Implementations that understand this transport parameter MUST treat the receipt of a non-empty value as a connection error of type TRANSPORT_PARAMETER_ERROR.¶
After negotiating this extension, endpoints MUST report received packets using the ACCURATE_ACK_ECN frame. This only applies to the application data packet number space. Initial and Handshake packets are acknowledged using the regular ACK frame.¶
It is invalid to send regular ACK frames in the application data packet number space after negotiating this extension. Endpoints MUST close the connection using a PROTOCOL_VIOLATION error when they receive an ACK frame in the application data packet number space after this extension was negotiated.¶
When using 0-RTT, both endpoints MUST remember the value of this transport parameter. This allows acknowledging 0-RTT packets using ACCURATE_ACK_ECN frames. If 0-RTT data is accepted by the server, the server MUST NOT disable this extension on the resumed connection.¶
The sender of an ACCURATE_ACK_ECN frame might be able to burden its peer by encoding a large number of ACK ranges. With the ACK frame defined in [RFC9000] it is not possible to split a contiguous sequence of packet numbers into multiple ranges, which becomes possible when using the ACCURATE_ACK_ECN frame. The number of ACK ranges is implicitely by the requirement that each frame fits into a QUIC packet. Receivers SHOULD make sure that they can process an ACCURATE_ACK_ECN frame containing a few hundred ACK ranges efficiently.¶
TODO consider IANA¶
TODO acknowledge.¶