LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: [lvs-users] WLC Slow start for 2.6.x kernel

To: "'Krzysztof Wilczynski'" <krzysztof.wilczynski@xxxxxxxxx>, "'Simon Horman'" <horms@xxxxxxxxxxxx>
Subject: Re: [lvs-users] WLC Slow start for 2.6.x kernel
Cc: lvs-users@xxxxxxxxxxxxxxxxxxxxxx
From: "Aleksey Chudov" <aleksey.chudov@xxxxxxxxx>
Date: Sun, 23 Oct 2011 01:42:27 +0300
Hello!

Thank you for great work. We have few heavy loaded LVS system with a lot of 
connections.
I'll test patch as soon as possible.

Best regards,
Aleksey


-----Original Message-----
From: kswilczynski@xxxxxxxxxxxxxx [mailto:kswilczynski@xxxxxxxxxxxxxx] On 
Behalf Of Krzysztof Wilczynski
Sent: Saturday, October 22, 2011 4:08 PM
To: Aleksey Chudov; Simon Horman
Cc: lvs-users@xxxxxxxxxxxxxxxxxxxxxx
Subject: WLC Slow start for 2.6.x kernel

Hello,

[...]
> > It will be difficult because I'm not C/C++ programmer.
> > Could you please port this patch to 2.6 kernel?
>
> Sure, I will try and find some time to up-port the patch.
[...]

I took Simon's patch and made an attempt to forward-port onto 2.6 and latest 
3.0.x (working mainly with "ipvs-next" and "net-next" trees):

https://github.com/kwilczynski/miscellaneous/blob/master/ipvs-thundering-herd-2.6.patch
https://github.com/kwilczynski/miscellaneous/blob/master/ipvs-thundering-herd-3.0.patch

Making sure that it holds water in my test environment / set-up.
Unfortunately, I do not have the ability to test this under a real life load 
and/or in a network / system that was suffering from the problem that this 
patch is trying to address. I other words, my notebook solely is not enough to 
see real benefit of the slow-start introduced by the patch, since it will reach 
limit of its capacity long before I am able to create enough congestion to spot 
the difference in a statistically significant manner.

Therefore, would you help me Aleksey and test it for me? :-) I am asking for 
your assistance, since you requested the forward-port in the first place, 
therefore you are probably the person whom will benefit from it the most.

Hopefully, everything will work as it says on the tin, and if so, then we can 
start working on merging this with upstream kernel and exposing ability to 
toggle this functionality on-demand via /proc and/or /sys entries per algorithm.

Thanks for help in advance! :)

KW


_______________________________________________
Please read the documentation before posting - it's available at:
http://www.linuxvirtualserver.org/

LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
or go to http://lists.graemef.net/mailman/listinfo/lvs-users

<Prev in Thread] Current Thread [Next in Thread>