The attached patch attempts to deprioritize bulk ack flows in cake.
Once we start accumulating enough acks to filter out, and we start
filtering them out, the "sparse flow optimization" in cake will
start prioritizing the shorter queues. This patch attempts to stop
that, which should give more time for more acks to be eliminated,
and deprioritize bulk ack flows slightly in favor of other traffic
that needs less latency.
Interesting stats from testing would be to measure a change in
ack_drops between the two versions, and changes in the rate of
increase of tcp. The 50x1 settings would be the most dramatic test...
--
Dave Täht
CEO, TekLibre, LLC
http://www.teklibre.com
Tel: 1-669-226-2619
From dd5bb48a4744cbd5e3081379e1e0954a31fbca02 Mon Sep 17 00:00:00 2001
From: Dave Taht <dave.t...@gmail.com>
Date: Tue, 5 Dec 2017 13:52:30 -0800
Subject: [PATCH] set filtered ack flows to bulk mode
Once we start accumulating enough acks to filter out, and we start
filtering them out, the "sparse flow optimization" in cake will
start prioritizing the shorter queues. This patch attempts to stop
that, which should give more time for more acks to be eliminated,
and deprioritize the ack flows slightly in favor of other traffic
that needs less latency.
---
sch_cake.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/sch_cake.c b/sch_cake.c
index 4010388..72ab02b 100644
--- a/sch_cake.c
+++ b/sch_cake.c
@@ -1492,6 +1492,7 @@ static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch,
ack = cake_ack_filter(q, flow);
if (ack) {
+ flow->set = CAKE_SET_BULK;
b->ack_drops++;
sch->qstats.drops++;
b->bytes += ack->len;
@@ -1532,6 +1533,7 @@ static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch,
ack = cake_ack_filter(q, flow);
if (ack) {
+ flow->set = CAKE_SET_BULK;
b->ack_drops++;
sch->qstats.drops++;
b->bytes += qdisc_pkt_len(ack);
--
2.7.4
_______________________________________________
Cake mailing list
Cake@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/cake
