On Tue, Jun 01, 2021 at 10:08:31PM +0800, Hyman wrote: > > > 在 2021/6/1 8:57, Peter Xu 写道: > > These two commands are missing when adding the QMP sister commands. Add > > them, > > so developers can play with them easier. > > > > Cc: Dr. David Alan Gilbert <dgilb...@redhat.com> > > Cc: Juan Quintela <quint...@redhat.com> > > Cc: Leonardo Bras Soares Passos <lsoar...@redhat.com> > > Cc: Chuan Zheng <zhengch...@huawei.com> > > Cc: huang...@chinatelecom.cn > > Signed-off-by: Peter Xu <pet...@redhat.com> > > --- > > PS: I really doubt whether this is working as expected... I ran one 200MB/s > > workload inside, what I measured is 20MB/s with current algorithm... > > Sampling > > 512 pages out of 1G mem is not wise enough I guess, especially that assumes > > dirty workload is spread across the memories while it's normally not the > > case.. > I doubt whether the sampling can cope with the situation that the guest > dirty memory too fast so that the sampling within a given time can not > finish, this may happens when vm is in large scale.
Not my case, though.. I'm with a 1G super small VM, starting a malloc() workload with 200MB, dirty rate 200MB/s. As I said in the other thread, I think it's the algorithm that may not really work well with such workload, while it could be one of the major workloads.. -- Peter Xu
