It looks like MacOS had CPU PM options removed from pmset in the last few years, so I’m not sure it’s possible to disable in recent versions. On Ubuntu I was able to get cpufrequtils to work a few months ago: https://askubuntu.com/questions/523640/how-i-can-disable-cpu-frequency-scaling-and-set-the-system-to-performance
Have you tried setting process priority with nice? https://superuser.com/questions/42817/is-there-any-way-to-set-the-priority-of-a-process-in-mac-os-x Matt On Friday, January 19, 2018 at 1:14:25 AM UTC-6, sotte...@gmail.com wrote: > > a. request client and server running on the same computer. > b. Unfortunately, I have not find the method how to disable OS power > management. I re-running this case on linux server, result shows that > there is almost no time out. I suspect the P's block is affected by the > external operating environment, and the M was not scheduled in time. > > 在 2018年1月19日星期五 UTC+8上午1:25:43，matthe...@gmail.com写道： >> >> Is your test request source running on the same computer? Have you >> disabled OS power management features? >> >> Matt >> >> On Thursday, January 18, 2018 at 10:23:43 AM UTC-6, sotte...@gmail.com >> wrote: >>> >>> I developed an RPC Service with high requirements for real-time >>> performance. There was an unexpected delay at a small probability, and I do >>> not know how this happened. I suspect this is go schedule related, of >>> course, this may also be caused by a code bug. >>> >>> Anyone can solve this problem, it will be a great help to me！ >>> >>> Following is my test process. trace.out in the attachment. >>> >>> >>> GO-Version：OS X go1.9.2 go1.10beta2 compile package >>> ::go1.10beta2.darwin-amd64.tar.gz >>> Net : localhost To localhost >>> QPS ： 11K >>> >>> RT statistics >>> >>> *TestTime* >>> >>> *Test Count* >>> >>> *0-1ms* >>> >>> *1-2ms* >>> >>> *2-5ms* >>> >>> *5-10ms* >>> >>> *10-20ms* >>> >>> *20+ms* >>> >>> *2018-01-16* >>> >>> 58409 >>> >>> 99.97432% >>> >>> 0.022256844% >>> >>> 0.00342413% >>> >>> 0 >>> >>> 0 >>> >>> 0 >>> >>> 51707 >>> >>> 99.99226% >>> >>> 77358964% >>> >>> 0 >>> >>> 0 >>> >>> 0 >>> >>> 0 >>> >>> 54650 >>> >>> 99.93413 % >>> >>> >>> 0.027447393% >>> >>> 0.021957913% >>> >>> 0.0036596523% >>> >>> 0.0054894784% >>> >>> 0.0073193046% >>> >>> >>> By analyzing trace.out, in addition to the GC, I found two places that >>> lead to very long blocks >>> >>> - Net IO operator >>> - Channel operator >>> >>> >>> <https://lh3.googleusercontent.com/-f3CwZlmkYo8/WmBYqfozeeI/AAAAAAAAAAo/fTdh0qtEmNYSHJqLwU-RJvLFhJiOcdmRACLcBGAs/s1600/9135C3F2-C6DA-4199-A27D-0660497B553B.png> >>> >>> >>> <https://lh3.googleusercontent.com/-qoqFDYLVtZg/WmBYm8unb3I/AAAAAAAAAAk/-V_-ck6IqVgtuMOT6uKvC3mm_oilw1bgQCLcBGAs/s1600/9135C3F2-C6DA-4199-A27D-0660497B553B.png> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> ReadLoop Code : >>> >>> for { head := c.dataBuf[0:WafBaseHeaderLen] _, err >>> := io.ReadFull(c.TcpConn, head) if err != nil { >>> log.Println("WafConn Err :", err.Error()) return nil, >>> err } dataLen, err := c.GetLength(head) >>> >>> //omit some exception handler _, err = >>> io.ReadFull(c.TcpConn, c.dataBuf[WafBaseHeaderLen: >>> WafBaseHeaderLen+dataLen]) if err != nil { >>> log.Println("WafConn Err :", err.Error()) return nil, >>> err } msg = &conncodec.WafMessage{ } >>> err = msg.Decode(c.dataBuf[:WafBaseHeaderLen + dataLen]) >>> return msg, err } >>> >>> >>> >>> channel init >>> >>> waf_resp := make(chan WafProxyResp, 1) >>> >>> waiting channel read event code: >>> >>> select { case result := <-waf_resp: *//do >>> something* case <-timer.C: //10 ms timeout >>> this.request_cache.DisposeRequest(request_id) >>> monitor.QPSStatAddWafRecvTimeout(1) return >>> GenWafResp(WAF_SUCCESS, "receive time out") } >>> >>> >>> Send Channel Code : >>> >>> select { case resp_chan <- resp: >>> //只使用一次，写入完成后关闭； close(resp_chan) >>> default: log.Print("Default fail !!!!! request id : >>> ", request_id) } >>> >>> >>> -- You received this message because you are subscribed to the Google Groups "golang-nuts" group. To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
