Hi, Jan I know performance is not a concern for this patch that people can always turn it off. I just run some p2p test with the stats on for a reference point. Basically it introduces around 10 cycles overhead per packet. In real pvp use cases, the performance influence should be small.
flow-subtable cnt Throughput influence 10k-1t -3.51% 10k-3t -3.33% 10k-5t -2.84% 10k-10t -2.30% 10k-20t -3.32% 100k-1t -3.02% 100k-3t -2.83% 100k-5t -2.11% 100k-10t -1.33% 100k-20t -3.81% 1M-1t -3.25% 1M-3t -2.75% 1M-5t -1.98% 1M-10t -1.78% 1M-20t -0.60% Tested-by: Yipeng Wang <yipeng1.w...@intel.com> Thanks Yipeng >-----Original Message----- >From: ovs-dev-boun...@openvswitch.org [mailto:ovs-dev- >boun...@openvswitch.org] On Behalf Of Jan Scheurich >Sent: Monday, December 18, 2017 6:14 AM >To: d...@openvswitch.org >Subject: [ovs-dev] [PATCH v4 0/3] dpif-netdev: Detailed PMD performance >metrics and supervision > >The run-time performance of PMDs is often difficult to understand and >trouble-shoot. The existing PMD statistics counters only provide a coarse >grained average picture. At packet rates of several Mpps sporadic drops of >packet bursts happen at sub-millisecond time scales and are impossible to >capture and analyze with existing tools. > >This patch set refactors the existing PMD statistics into a dedicated submodule >and collects a large number of important PMD performance metrics per PMD >iteration, maintaining histograms and circular histories for iteration metrics >and millisecond averages. To capture sporadic drop events, the patch set can >be configured to monitor iterations for suspicious metrics and to log the >neighborhood of such iterations for off-line analysis. > >The extra cost for the performance metric collection and the supervision has >been measured to be in the order of 1% compared to the base commit in a >PVP setup with L3 pipeline over VXLAN tunnels. For that reason the metrics >collection is disabled by default and can be enabled at run-time through >configuration. > >The first patch in the series fully includes the changes proposed in Darrel's >earlier "[patch_v5 0/3] dpif-netdev: Fix and refactor pmd stats" >(https://mail.openvswitch.org/pipermail/ovs-dev/2017-August/337686.html). > > >v3->v4: > Rebased to master (commit 4d0a31b) > Reverting changes to struct dp_netdev_pmd_thread > Make metrics collection configurable > Several bugfixes > >v2->v3: > Rebased to OVS master (commit 3728b3b) > Non-trivial adaptation to struct dp_netdev_pmd_thread > refactored in commit a807c157 (Bhanu) > No other changes compared to v2 > >v1 -> v2: > Rebased to OVS master (commit 7468ec788) > No other changes compared to v1 > > >Jan Scheurich (3): > dpif-netdev: Refactor PMD performance into dpif-netdev-perf > dpif-netdev: Detailed performance stats for PMDs > dpif-netdev: Detection and logging of suspicious PMD iterations > > lib/automake.mk | 2 + > lib/dp-packet.h | 2 + > lib/dpif-netdev-perf.c | 515 >+++++++++++++++++++++++++++++++++++++++++++++++++ > lib/dpif-netdev-perf.h | 322 +++++++++++++++++++++++++++++++ > lib/dpif-netdev.c | 471 ++++++++++++++++++++++++++------------------ > lib/netdev-dpdk.c | 23 ++- > lib/netdev-dpdk.h | 14 ++ > ofproto/ofproto-dpif.c | 3 +- > tests/pmd.at | 22 ++- > 9 files changed, 1168 insertions(+), 206 deletions(-) > create mode 100644 lib/dpif-netdev-perf.c > create mode 100644 lib/dpif-netdev-perf.h > >-- >1.9.1 > >_______________________________________________ >dev mailing list >d...@openvswitch.org >https://mail.openvswitch.org/mailman/listinfo/ovs-dev _______________________________________________ dev mailing list d...@openvswitch.org https://mail.openvswitch.org/mailman/listinfo/ovs-dev
