You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Describe the bug
Recent I was trying to update the grpcio version from v0.10.4 to the newest master commit for tikv. And when I did some benchmark using sysbench I found there are performance regression within different workloads(oltp_read_only, oltp_read_write and oltp_write_only), the qps dropped around 5% ~ 10%.
The v0.10.4 build is based on commit: tikv/tikv@966e134
The master build is based on: https://github.com/glorv/tikv/commits/grpc-master/
From the metrics I saw master grpc threads consumed more cpu time than v0.10.4:
From the cpu profiling, I saw the cq_next call consumed more cpu time while other part are almost the same:
(v0.10.4)
(master)
With v0.10.4, grpc cost about 14.78% of all cpu time:
CallTag::resolve(4.12%), CompletionQueue::next(1.45%), UnfinishedWork::finish(9.22%)
With master commit, gRPC cost about 21.35% of all cpu time:
CallTag::resolve(5.87%), CompletionQueue::next(4.77%), UnfinishedWork::finish(10.71%)
It seems that some changes in CompletionQueue may be the root cause of this regression. To Reproduce
Steps to reproduce the behavior:
Go to '...'
Click on '....'
Scroll down to '....'
See error
Expected behavior
A clear and concise description of what you expected to happen.
System information
CPU architecture:
Distribution and kernel version:
SELinux on?:
Any other system details we should know?:
Additional context
Add any other context about the problem here.
The text was updated successfully, but these errors were encountered:
Describe the bug
Recent I was trying to update the grpcio version from v0.10.4 to the newest master commit for tikv. And when I did some benchmark using sysbench I found there are performance regression within different workloads(oltp_read_only, oltp_read_write and oltp_write_only), the qps dropped around 5% ~ 10%.
The v0.10.4 build is based on commit: tikv/tikv@966e134
The master build is based on: https://github.com/glorv/tikv/commits/grpc-master/
From the metrics I saw master grpc threads consumed more cpu time than v0.10.4:
From the cpu profiling, I saw the
cq_next
call consumed more cpu time while other part are almost the same:(v0.10.4)
(master)
With v0.10.4, grpc cost about 14.78% of all cpu time:
CallTag::resolve(4.12%), CompletionQueue::next(1.45%), UnfinishedWork::finish(9.22%)
With master commit, gRPC cost about 21.35% of all cpu time:
CallTag::resolve(5.87%), CompletionQueue::next(4.77%), UnfinishedWork::finish(10.71%)
It seems that some changes in
CompletionQueue
may be the root cause of this regression.To Reproduce
Steps to reproduce the behavior:
Expected behavior
A clear and concise description of what you expected to happen.
System information
Additional context
Add any other context about the problem here.
The text was updated successfully, but these errors were encountered: