On Thu, Sep 26, 2013 at 10:40 PM, Robert <[email protected]> wrote: > Hello Rainer, David, > > The spirent is set to send out 300k fps. Attached are the pstats, mpstats, > iostats., and its writing locally, this is what I get back: >
Thanks again, this is useful as a baseline. As discussed with David, we see big improvement (queue does not fill up during test time frame), but we still can't run at this sustained rate. Now let me elaborate a bit on my next steps. I am trying to find the bottleneck. I am NOT trying to improve throughput. That's something we can do once we have found the bottleneck and thought about how we can discuss it. IT IS VITALLY IMPORTANT THAT YOU USE THE EXACT CONFIG THAT I PROVIDED. Otherwise, I can't draw conclusions and this turns out to a waste of time. My next step is to see how NFS affects the overall picture. So please use the EXACT SAME config that you used for the local test, but replace all local directories with those on the nfs drive again (those that you used so far). You can leave the impstats log on the local disk. Then re-run the test with THE SAME SETTINGS and gather the same information (CPU utilization and impstats is all I need). Report that information. Once we have it, we can conclude how much nfs affects the overall picture. The steps thereafter (NOT TO BE CARRIED OUT NOW or in parallel) will focus on why even with local writing we have a queue buildup. So we will go back from nfs to local disk (no matter what the nfs results are - nfs introduces too much potential randomness to be useful for that type of testing). There are two suspects: "contains" filters and missing "stop" statements. We will probably need three more test runs to get us a useful matrix. We may need to re-run some of the test scenarios if they do not fit expectation (so we can make sure results are consistent and not affected by something random). Once we have done this, we should be able to better understand the bottlenecks and *then* can begin to think about a route forward. Thanks, Rainer _______________________________________________ rsyslog mailing list http://lists.adiscon.net/mailman/listinfo/rsyslog http://www.rsyslog.com/professional-services/ What's up with rsyslog? Follow https://twitter.com/rgerhards NOTE WELL: This is a PUBLIC mailing list, posts are ARCHIVED by a myriad of sites beyond our control. PLEASE UNSUBSCRIBE and DO NOT POST if you DON'T LIKE THAT.
