LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: Performance testing

To: Brian Jackson <brian-lists@xxxxxxxx>
Subject: Re: Performance testing
Cc: "lvs-users@xxxxxxxxxxxxxxxxxxxxxx" <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
From: Julian Anastasov <ja@xxxxxx>
Date: Sat, 7 Sep 2002 02:33:38 +0000 (GMT)
        Hello,

On Fri, 6 Sep 2002, Brian Jackson wrote:

> LVS with Preempt and lock-break patches from Robert Love
> Avg. packets/sec
> RS1           RS2
> 4677          4675
>
> I didn't know if anybody would be interested in these numbers.

        More interesting would be the numbers for NAPI+LVS (I
didn't tried it myself yet, no free time). NAPI (already in
2.4.20pre) solves exactly such problems where the higher input
load causes strange/lower results in the output due to big IRQ
rate. IMO, your goal should be to send 25KP/s (SYNs) and to receive 25K 
packets/sec. This is achievable with LVS on PCI 33MHz without NAPI.

        As for the differences when extra patches are used,
may be it only proves that these patches add extra overhead for
the LVS forwarding path, they simply have another purpose.

> Let me what you think or if you have any questions.
>
>  --Brian Jackson

Regards

--
Julian Anastasov <ja@xxxxxx>



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