The output from prof:eval will probably make your queries look slower than they really are, because profiling adds its own overhead to the evaluation. Profiling shows where queries spend time, but those timings will include profiler overhead.
If you're just trying to get the server elapsed time, use xdmp:elapsed-time. That has a lot less overhead than prof:eval does. You could also return or log the output from xdmp:query-meters, which is more expensive that xdmp:elapsed-time but probably cheaper than profiling. Also I'm not sure how easy it'll be to interpret multiple profiler reports that ran in parallel. You might be start by looking at the output from a single request first. -- Mike > On 13 Jan 2015, at 03:18 , priya dharshni <[email protected]> wrote: > > Hi, > > > > We are working on analyzing the performance of our MarkLogic XQueries. We > are using threads to concurrently hit the marklogic xqueries (in marklogic > we are using prof.eval and returns back the profiling report). > > > Example : > > We are using 5 threads and each thread returns one profile report. > > > We are calculating the total average time from the profile reports. Now > please let me know is it a proper way to estimate the performance numbers? > > > Regards, > > Priya > > _______________________________________________ > General mailing list > [email protected] > http://developer.marklogic.com/mailman/listinfo/general _______________________________________________ General mailing list [email protected] http://developer.marklogic.com/mailman/listinfo/general
