From: Rick Jones <rick.jones2@hp.com> To: Jason Wang <jasowang@redhat.com> Cc: mst@redhat.com, mashirle@us.ibm.com, krkumar2@in.ibm.com, habanero@linux.vnet.ibm.com, rusty@rustcorp.com.au, netdev@vger.kernel.org, linux-kernel@vger.kernel.org, virtualization@lists.linux-foundation.org, edumazet@google.com, tahm@linux.vnet.ibm.com, jwhan@filewood.snu.ac.kr, davem@davemloft.net, akong@redhat.com, kvm@vger.kernel.org, sri@us.ibm.com Subject: Re: [net-next RFC V5 0/5] Multiqueue virtio-net Date: Fri, 06 Jul 2012 09:23:25 -0700 [thread overview] Message-ID: <4FF710FD.2090100@hp.com> (raw) In-Reply-To: <4FF696C9.5070907@redhat.com> On 07/06/2012 12:42 AM, Jason Wang wrote: > I'm not expert of tcp, but looks like the changes are reasonable: > - we can do full-sized TSO check in tcp_tso_should_defer() only for > westwood, according to tcp westwood > - run tcp_tso_should_defer for tso_segs = 1 when tso is enabled. I'm sure Eric and David will weigh-in on the TCP change. My initial inclination would have been to say "well, if multiqueue is draining faster, that means ACKs come-back faster, which means the "race" between more data being queued by netperf and ACKs will go more to the ACKs which means the segments being sent will be smaller - as TCP_NODELAY is not set, the Nagle algorithm is in force, which means once there is data outstanding on the connection, no more will be sent until either the outstanding data is ACKed, or there is an accumulation of > MSS worth of data to send. >> Also, how are you combining the concurrent netperf results? Are you >> taking sums of what netperf reports, or are you gathering statistics >> outside of netperf? >> > > The throughput were just sumed from netperf result like what netperf > manual suggests. The cpu utilization were measured by mpstat. Which mechanism to address skew error? The netperf manual describes more than one: http://www.netperf.org/svn/netperf2/trunk/doc/netperf.html#Using-Netperf-to-Measure-Aggregate-Performance Personally, my preference these days is to use the "demo mode" method of aggregate results as it can be rather faster than (ab)using the confidence intervals mechanism, which I suspect may not really scale all that well to large numbers of concurrent netperfs. I also tend to use the --enable-burst configure option to allow me to minimize the number of concurrent netperfs in the first place. Set TCP_NODELAY (the test-specific -D option) and then have several transactions outstanding at one time (test-specific -b option with a number of additional in-flight transactions). This is expressed in the runemomniaggdemo.sh script: http://www.netperf.org/svn/netperf2/trunk/doc/examples/runemomniaggdemo.sh which uses the find_max_burst.sh script: http://www.netperf.org/svn/netperf2/trunk/doc/examples/find_max_burst.sh to pick the burst size to use in the concurrent netperfs, the results of which can be post-processed with: http://www.netperf.org/svn/netperf2/trunk/doc/examples/post_proc.py The nice feature of using the "demo mode" mechanism is when it is coupled with systems with reasonably synchronized clocks (eg NTP) it can be used for many-to-many testing in addition to one-to-many testing (which cannot be dealt with by the confidence interval method of dealing with skew error) >> A single instance TCP_RR test would help confirm/refute any >> non-trivial change in (effective) path length between the two cases. >> > > Yes, I would test this thanks. Excellent. happy benchmarking, rick jones
WARNING: multiple messages have this Message-ID (diff)
From: Rick Jones <rick.jones2@hp.com> To: Jason Wang <jasowang@redhat.com> Cc: krkumar2@in.ibm.com, habanero@linux.vnet.ibm.com, mashirle@us.ibm.com, kvm@vger.kernel.org, mst@redhat.com, netdev@vger.kernel.org, linux-kernel@vger.kernel.org, virtualization@lists.linux-foundation.org, edumazet@google.com, tahm@linux.vnet.ibm.com, jwhan@filewood.snu.ac.kr, davem@davemloft.net, sri@us.ibm.com Subject: Re: [net-next RFC V5 0/5] Multiqueue virtio-net Date: Fri, 06 Jul 2012 09:23:25 -0700 [thread overview] Message-ID: <4FF710FD.2090100@hp.com> (raw) In-Reply-To: <4FF696C9.5070907@redhat.com> On 07/06/2012 12:42 AM, Jason Wang wrote: > I'm not expert of tcp, but looks like the changes are reasonable: > - we can do full-sized TSO check in tcp_tso_should_defer() only for > westwood, according to tcp westwood > - run tcp_tso_should_defer for tso_segs = 1 when tso is enabled. I'm sure Eric and David will weigh-in on the TCP change. My initial inclination would have been to say "well, if multiqueue is draining faster, that means ACKs come-back faster, which means the "race" between more data being queued by netperf and ACKs will go more to the ACKs which means the segments being sent will be smaller - as TCP_NODELAY is not set, the Nagle algorithm is in force, which means once there is data outstanding on the connection, no more will be sent until either the outstanding data is ACKed, or there is an accumulation of > MSS worth of data to send. >> Also, how are you combining the concurrent netperf results? Are you >> taking sums of what netperf reports, or are you gathering statistics >> outside of netperf? >> > > The throughput were just sumed from netperf result like what netperf > manual suggests. The cpu utilization were measured by mpstat. Which mechanism to address skew error? The netperf manual describes more than one: http://www.netperf.org/svn/netperf2/trunk/doc/netperf.html#Using-Netperf-to-Measure-Aggregate-Performance Personally, my preference these days is to use the "demo mode" method of aggregate results as it can be rather faster than (ab)using the confidence intervals mechanism, which I suspect may not really scale all that well to large numbers of concurrent netperfs. I also tend to use the --enable-burst configure option to allow me to minimize the number of concurrent netperfs in the first place. Set TCP_NODELAY (the test-specific -D option) and then have several transactions outstanding at one time (test-specific -b option with a number of additional in-flight transactions). This is expressed in the runemomniaggdemo.sh script: http://www.netperf.org/svn/netperf2/trunk/doc/examples/runemomniaggdemo.sh which uses the find_max_burst.sh script: http://www.netperf.org/svn/netperf2/trunk/doc/examples/find_max_burst.sh to pick the burst size to use in the concurrent netperfs, the results of which can be post-processed with: http://www.netperf.org/svn/netperf2/trunk/doc/examples/post_proc.py The nice feature of using the "demo mode" mechanism is when it is coupled with systems with reasonably synchronized clocks (eg NTP) it can be used for many-to-many testing in addition to one-to-many testing (which cannot be dealt with by the confidence interval method of dealing with skew error) >> A single instance TCP_RR test would help confirm/refute any >> non-trivial change in (effective) path length between the two cases. >> > > Yes, I would test this thanks. Excellent. happy benchmarking, rick jones
next prev parent reply other threads:[~2012-07-06 16:23 UTC|newest] Thread overview: 91+ messages / expand[flat|nested] mbox.gz Atom feed top 2012-07-05 10:29 [net-next RFC V5 0/5] Multiqueue virtio-net Jason Wang 2012-07-05 10:29 ` Jason Wang 2012-07-05 10:29 ` [net-next RFC V5 1/5] virtio_net: Introduce VIRTIO_NET_F_MULTIQUEUE Jason Wang 2012-07-05 10:29 ` Jason Wang 2012-07-05 10:29 ` [net-next RFC V5 2/5] virtio_ring: move queue_index to vring_virtqueue Jason Wang 2012-07-05 10:29 ` Jason Wang 2012-07-05 11:40 ` Sasha Levin 2012-07-05 11:40 ` Sasha Levin 2012-07-06 3:17 ` Jason Wang 2012-07-06 3:17 ` Jason Wang 2012-07-26 8:20 ` Paolo Bonzini 2012-07-26 8:20 ` Paolo Bonzini 2012-07-30 3:30 ` Jason Wang 2012-07-30 3:30 ` Jason Wang 2012-07-05 10:29 ` [net-next RFC V5 3/5] virtio: intorduce an API to set affinity for a virtqueue Jason Wang 2012-07-05 10:29 ` Jason Wang 2012-07-27 14:38 ` Paolo Bonzini 2012-07-27 14:38 ` Paolo Bonzini 2012-07-29 20:40 ` Michael S. Tsirkin 2012-07-29 20:40 ` Michael S. Tsirkin 2012-07-30 6:27 ` Paolo Bonzini 2012-08-09 15:14 ` Paolo Bonzini 2012-08-09 15:14 ` Paolo Bonzini 2012-08-09 15:13 ` Paolo Bonzini 2012-08-09 15:13 ` Paolo Bonzini 2012-08-09 15:35 ` Avi Kivity 2012-08-09 15:35 ` Avi Kivity 2012-07-05 10:29 ` [net-next RFC V5 4/5] virtio_net: multiqueue support Jason Wang 2012-07-05 10:29 ` Jason Wang 2012-07-05 20:02 ` Amos Kong 2012-07-05 20:02 ` Amos Kong 2012-07-06 7:45 ` Jason Wang 2012-07-06 7:45 ` Jason Wang 2012-07-20 13:40 ` Michael S. Tsirkin 2012-07-20 13:40 ` Michael S. Tsirkin 2012-07-21 12:02 ` Sasha Levin 2012-07-21 12:02 ` Sasha Levin 2012-07-23 5:54 ` Jason Wang 2012-07-23 5:54 ` Jason Wang 2012-07-23 9:28 ` Sasha Levin 2012-07-23 9:28 ` Sasha Levin 2012-07-30 3:29 ` Jason Wang 2012-07-30 3:29 ` Jason Wang 2012-07-29 9:44 ` Michael S. Tsirkin 2012-07-29 9:44 ` Michael S. Tsirkin 2012-07-30 3:26 ` Jason Wang 2012-07-30 3:26 ` Jason Wang 2012-07-30 13:00 ` Sasha Levin 2012-07-30 13:00 ` Sasha Levin 2012-07-23 5:48 ` Jason Wang 2012-07-23 5:48 ` Jason Wang 2012-07-29 9:50 ` Michael S. Tsirkin 2012-07-29 9:50 ` Michael S. Tsirkin 2012-07-30 5:15 ` Jason Wang 2012-07-30 5:15 ` Jason Wang 2012-07-05 10:29 ` [net-next RFC V5 5/5] virtio_net: support negotiating the number of queues through ctrl vq Jason Wang 2012-07-05 10:29 ` Jason Wang 2012-07-05 12:51 ` Sasha Levin 2012-07-05 12:51 ` Sasha Levin 2012-07-05 20:07 ` Amos Kong 2012-07-05 20:07 ` Amos Kong 2012-07-06 7:46 ` Jason Wang 2012-07-06 7:46 ` Jason Wang 2012-07-06 3:20 ` Jason Wang 2012-07-06 3:20 ` Jason Wang 2012-07-06 6:38 ` Stephen Hemminger 2012-07-06 6:38 ` Stephen Hemminger 2012-07-06 9:26 ` Jason Wang 2012-07-06 9:26 ` Jason Wang 2012-07-06 8:10 ` Sasha Levin 2012-07-06 8:10 ` Sasha Levin 2012-07-09 20:13 ` Ben Hutchings 2012-07-09 20:13 ` Ben Hutchings 2012-07-20 12:33 ` Michael S. Tsirkin 2012-07-20 12:33 ` Michael S. Tsirkin 2012-07-23 5:32 ` Jason Wang 2012-07-23 5:32 ` Jason Wang 2012-07-05 17:45 ` [net-next RFC V5 0/5] Multiqueue virtio-net Rick Jones 2012-07-05 17:45 ` Rick Jones 2012-07-06 7:42 ` Jason Wang 2012-07-06 7:42 ` Jason Wang 2012-07-06 16:23 ` Rick Jones [this message] 2012-07-06 16:23 ` Rick Jones 2012-07-09 3:23 ` Jason Wang 2012-07-09 3:23 ` Jason Wang 2012-07-09 16:46 ` Rick Jones 2012-07-09 16:46 ` Rick Jones 2012-07-08 8:19 ` Ronen Hod 2012-07-08 8:19 ` Ronen Hod 2012-07-09 5:35 ` Jason Wang 2012-07-09 5:35 ` Jason Wang
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=4FF710FD.2090100@hp.com \ --to=rick.jones2@hp.com \ --cc=akong@redhat.com \ --cc=davem@davemloft.net \ --cc=edumazet@google.com \ --cc=habanero@linux.vnet.ibm.com \ --cc=jasowang@redhat.com \ --cc=jwhan@filewood.snu.ac.kr \ --cc=krkumar2@in.ibm.com \ --cc=kvm@vger.kernel.org \ --cc=linux-kernel@vger.kernel.org \ --cc=mashirle@us.ibm.com \ --cc=mst@redhat.com \ --cc=netdev@vger.kernel.org \ --cc=rusty@rustcorp.com.au \ --cc=sri@us.ibm.com \ --cc=tahm@linux.vnet.ibm.com \ --cc=virtualization@lists.linux-foundation.org \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.