From e8bff3302c3cfe085fafdb0df7c349b8448cb689 Mon Sep 17 00:00:00 2001 From: ianswett Date: Fri, 1 Nov 2019 13:59:51 -0400 Subject: [PATCH] Limit CWND increase in slow start When not pacing packets, limit the CWND increase to 2 packets per ACK frame received and cite RFC3465. Fixes #3094 --- draft-ietf-quic-recovery.md | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/draft-ietf-quic-recovery.md b/draft-ietf-quic-recovery.md index ffbcea3095..738df563b0 100644 --- a/draft-ietf-quic-recovery.md +++ b/draft-ietf-quic-recovery.md @@ -754,7 +754,10 @@ Implementations MUST either use pacing or limit such bursts to the initial congestion window, which is recommended to be the minimum of 10 * max_datagram_size and max(2* max_datagram_size, 14720)), where max_datagram_size is the current maximum size of a datagram for the connection, -not including UDP or IP overhead. +not including UDP or IP overhead. Implementations not using a pacer SHOULD +limit the congestion window increase during slow start to +2 * max_datagram_size per ACK frame received in order to limit bursts, as +specified by {{?RFC3465}}. As an example of a well-known and publicly available implementation of a flow pacer, implementers are referred to the Fair Queue packet scheduler (fq qdisc)