On Thu, Aug 20, 2009 at 9:09 PM, Avi Kivity<avi@xxxxxxxxxx> wrote: > On 08/19/2009 03:53 PM, Ryota Ozaki wrote: >> >> Hi, >> >> This patch let kvm exposes per-vm statistics about >> such as the total number of vm exits, via debugfs. >> >> Existing kvm already collects the per-vm statistics, >> but it has no interface to expose them to users. >> This patch creates directories named a pid of the >> corresponding vm in debugfs of kvm, containing the >> same files that original debugfs exposes. >> >> The per-vm statistics are useful to know activities >> of vms (and to identify anomalous vms for example) >> with more detailed information than cpu and memory >> usage, and network traffics. And also the patch >> introduces no performance overhead, thus, it should >> be familiar with online operations, e.g., dynamic >> adaptation of assigning vm resources using the >> statistics. >> >> Note that this patch requires a trivial modification >> to kvm_stat script. Once this patch is accepted I will >> send a patch for it. >> >> Thanks, >> ozaki-r >> >> > > My plan is to completely remove the current statistics in favour of > tracepoints. You can already display tracepoint statistics with 'perf stat' > (see tools/perf); tracepoints have the advantage that they can be completely > disabled at runtime and thus have no performance impact. Oh, I don't really know it. I'll try it later. BTW, is the function feasible to my purpose, I mean, is it low overhead enough even if it is enabled? ozaki-r > > -- > error compiling committee.c: too many arguments to function > > -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
