[
https://issues.apache.org/jira/browse/TS-1034?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Leif Hedstrom updated TS-1034:
------------------------------
Fix Version/s: 3.1.4
> reduce futex locking period
> ---------------------------
>
> Key: TS-1034
> URL: https://issues.apache.org/jira/browse/TS-1034
> Project: Traffic Server
> Issue Type: Improvement
> Components: Core, HTTP
> Affects Versions: 3.1.1
> Reporter: Zhao Yongming
> Assignee: Zhao Yongming
> Fix For: 3.1.4
>
>
> we need to reduce futex locking period, here is a simple testing in my
> 24cores HP380 system, with 24 ab, all cached in memory:
> {code}
> #!/bin/sh
> for i in {1..24}
> do
> ab -n 100000000 -c 16 -X 127.0.0.$i:8080
> http://img02.taobaocdn.com/tps/i2/T1o0ypXk4wXXXXXXXX-1000-40.png?$i &
> done
> {code}
> result:
> {code}
> Every 2.0s: echo show:proxy-stats | traffic_shell
> Mon Nov 28 16:06:42 2011
> Successfully Initialized MgmtAPI in /var/run/trafficserver
> Document Hit Rate -------- 100.000000 % *
> Bandwidth Saving --------- 100.000000 % *
> Cache Percent Free ------- 99.999619 %
> Open Server Connections -- 0
> Open Client Connections -- 9
> Open Cache Connections --- 2
> Client Throughput -------- 6824.747070 MBit/Sec
> Transaction Per Second --- 53914.925781
> * Value represents 10 second average.
> strace -c -p 11712
> Process 11712 attached - interrupt to quit
> ^CProcess 11712 detached
> % time seconds usecs/call calls errors syscall
> ------ ----------- ----------- --------- --------- ----------------
> 26.85 0.890335 15 58920 writev
> 24.45 0.810866 7 118147 epoll_ctl
> 22.27 0.738451 13 58920 close
> 11.50 0.381362 6 59227 getsockname
> 9.86 0.326843 3 119192 59228 read
> 3.53 0.117058 16 7100 1931 futex
> 1.53 0.050884 58 884 epoll_wait
> 0.00 0.000037 0 404 rt_sigprocmask
> 0.00 0.000000 0 3 write
> 0.00 0.000000 0 2 brk
> 0.00 0.000000 0 10 msync
> ------ ----------- ----------- --------- --------- ----------------
> 100.00 3.315836 422809 61159 total
> {code}
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
