Code Review / controller.git / commit
? search:
summary | shortlog | log | commit | commitdiff | review | tree
(parent: 94374bc) | patch
BUG 2734 : Use the 95th percentile number to determine backpressure 35/15535/1
authorMoiz Raja <moraja@cisco.com>
Fri, 20 Feb 2015 10:53:20 +0000 (02:53 -0800)
committerMoiz Raja <moraja@cisco.com>
Fri, 20 Feb 2015 10:53:20 +0000 (02:53 -0800)
commit6056a678a4409d4d43a0f44b8d0326d387bb81e2
tree2c6a67ebe3704b407a4d57589b9a4ce4ac8d5218tree | snapshot
parent94374bc65cc6c57f1d2e77e1cd15902bb69d4feacommit | diff
BUG 2734 : Use the 95th percentile number to determine backpressure

When running tests with OpenFlow I see the following transaction completion
times consistently for 127 and 255 switches with statistics on.

50% - 5-6 ms
75% - 7-8 ms
95% - 20-60 ms
98% - 100-160 ms

By using the 95th percentile number to determine transaction creation backpressure
we can safely increase our throughput without causing timeouts due to outliers.

Change-Id: I0d9739b3045afb389583f2dd18ac431215a28f84
Signed-off-by: Moiz Raja <moraja@cisco.com>
opendaylight/md-sal/sal-distributed-datastore/src/main/java/org/opendaylight/controller/cluster/datastore/ThreePhaseCommitCohortProxy.java diff | blob | history
opendaylight/md-sal/sal-distributed-datastore/src/test/java/org/opendaylight/controller/cluster/datastore/ThreePhaseCommitCohortProxyTest.java diff | blob | history
OpenDaylight controller