On Tue, Jun 16, 2009 at 11:42 PM, Tharindu Rukshan Bamunuarachchi<tharindu@xxxxxxxxxxx> wrote: > hi All, > > recently we were developing high performance / low latency transaction > processing system with Linux. (i.e. SuSE 11 but 2.6.29 vanilla kernel) > > we tried to send high volume of traffic, typically 20K messages per > second. each message is about 400 bytes to 800 bytes. > > we saw ksoftirq thread is taking about 19 % CPU at 15K rate and at 20K > rate it was about 100%. so we could not pump messages more than 20K. > > do you guys have any idea about ksoftirq behaviour, why is it taking 100% CPU. > > we have tried both 1G NIC and 10G NIC. we saw high % for both NICs ? Besides what other people has said about NAPI and such, there's a chance it has something to do with the various enhancement made into CFS scheduler. Other thing that might influence your result is the preemption model you pick? do you use voluntary preemption, full preemption or no preemption at all? As you could easily interpret, ksoftirqd is a process which need to be scheduled, like every other processes do. So, the faster it could preempt other process when it has jobs to do, IMO theoritically the faster data rate you get. CMIIW people regards, Mulyadi. -- To unsubscribe from this list: send an email with "unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx Please read the FAQ at http://kernelnewbies.org/FAQ
