Re: [PATCH] mac80211: improve PID rate control mechanism by avoiding rate oscillation problem
From: Felix Fietkau <hidden>
Date: 2012-02-29 12:45:17
Also in:
lkml
On 2012-02-29 1:14 PM, YIN Wei wrote:
quoted hunk ↗ jump to hunk
From: Wei YIN (Wei.Yin@nicta.com.au) Improve PID rate control mechanism by solving the rate oscillation problem. Current PID mechanism is based on a PID controller which tries to minimise the difference between the frame loss ratio (FLR) and the target FLR. Therefore it is straight forward that it increases to a higher rate when the FLR is less than the target without considering whether the higher rate can be supported. If the higher rate cannot be supported, significant FLR will occur, which causes the mechanism to decrease the rate sharply. The proposed approach only updates the rate when the proposed rate by the PID controller can achieve better throughput than the old rate. This patch applies to kernel 3.3.0. Signed-off-by: Wei YIN (Wei.Yin@nicta.com.au) --- diff -uprN wireless-testing_orig/net/mac80211/Kconfig wireless-testing/net/mac80211/Kconfig--- wireless-testing_orig/net/mac80211/Kconfig 2012-02-1713:59:53.495254495 +1000+++ wireless-testing/net/mac80211/Kconfig 2012-02-21 11:35:40.495706869 +1000@@ -21,6 +21,7 @@ config MAC80211_HAS_RC config MAC80211_RC_PID bool "PID controller based rate control algorithm" if EXPERT select MAC80211_HAS_RC + default y ---help--- This option enables a TX rate control algorithm for mac80211 that uses a PID controller to select the TX
Several issues with this patch: line wrapping, weird encoding in the comments: — Also, please drop this change that alters the default value of MAC80211_RC_PID. - Felix