On Thu, 2 Feb 2017 18:48:41 +0200 Ran Shalit <ranshalit@xxxxxxxxx> wrote: > On Thu, Feb 2, 2017 at 4:53 PM, Julia Cartwright <julia@xxxxxx> wrote: > > On Thu, Feb 02, 2017 at 07:53:55AM +0200, Ran Shalit wrote: > >> Hello, > > > > Hello Ran- > > > >> I have followed instruction for using cycletest in: > >> https://rt.wiki.kernel.org/index.php/Cyclictest > >> > >> It described command lines for different scenarios. > >> usually one with load and the next without load. > >> Yet, I see no difference between the load 100% and no-load commands. > >> > >> How can you simulate load with cycletest ? > > > > cyclictest itself doesn't do any loading, it's common to run it with > > some other processes in the background which generate a simulated load. > > > > Keep in mind that the using a utility like the ones I've listed below is > > only _simulating_ what a load _might_ be like in a real application. > > The best thing to do is to actually test your application. > > > > - hackbench - scheduling load (nice because it's part of rt-tests) > > - stress - load-generating swiss-army knife > > - xfstests - disk / filesystem load > > - 'make -j' in a kernel source tree > > - iperf - network load > > - glxgears - gfx load > > > > (those are the ones that come to mind, there are others I'm forgetting) > > > > Julia > > Hi Julia , Joakim , Kannan, > > Thank you very much for answering my question ! You might want to look at 'rteval', which runs hackbench and 'make -j' while running cyclictest. When the run duration is done it does some statistics calculations on the cyclictest histogram. https://git.kernel.org/cgit/linux/kernel/git/clrkwllms/rteval.git/ I'm looking at adding a 'stress' module to it as well. /me takes note of the other loads that Julia listed... Clark -- The United States Coast Guard Ruining Natural Selection since 1790
Attachment:
pgpqLEuVl_t4W.pgp
Description: OpenPGP digital signature
