基于 MPI 的多服务器时间校准(C++
案例介绍
1. record current timestamp TA[n-1](use RDTSC time counter), call GstRouterSend(GstRouterSend is actually MPI_Send with some wrapper) in GstRouterOMPI.cpp to send timestamp data to server tb 10.10.70.201
2. after sending successfully, call GstRouterRecv in GstRouterOMPI.cpp to receive the timestamp data sent by server tb(use RDTSC time counter)
3. after receiving successfully, resolve and record the timestamp Ttb[n-1](RDTSC counter) of server tb to record current timestamp(RDTSC counter) TA[n]
4. when the above steps are done, loop back to step 1, do the operaton N times according to the loop number N
when the aboves are done, server A could get all the data in {TA[n]} and {Ttb[n]}, thus get the estimation value of C. the equation for estimating C is as below:
d= (sum of (TA[n]-TA[n-1]) from n=1 to n=N)/2N;
C= (sum of (TA[n] - Ttb[n-1]) from n=1 to n=N)/N - d;
5. dump nodename.dat files to record the following information
基于 MPI 的多服务器时间校准(C++
案例介绍
there is a server A 10.10.70.200 and server tb 10.10.70.201:
1. record current timestamp TA[n-1](use RDTSC time counter), call GstRouterSend(GstRouterSend is actually MPI_Send with some wrapper) in GstRouterOMPI.cpp to send timestamp data to server tb 10.10.70.201
2. after sending successfully, call GstRouterRecv in GstRouterOMPI.cpp to receive the timestamp data sent by server tb(use RDTSC time counter)
3. after receiving successfully, resolve and record the timestamp Ttb[n-1](RDTSC counter) of server tb to record current timestamp(RDTSC counter) TA[n]
4. when the above steps are done, loop back to step 1, do the operaton N times according to the loop number N
when the aboves are done, server A could get all the data in {TA[n]} and {Ttb[n]}, thus get the estimation value of C. the equation for estimating C is as below:
d= (sum of (TA[n]-TA[n-1]) from n=1 to n=N)/2N;
C= (sum of (TA[n] - Ttb[n-1]) from n=1 to n=N)/N - d;
5. dump nodename.dat files to record the following information
案例图片
1小时精推人才
需求方专属客服，免费梳理匹配
