Am 22.05.2012 18:17, schrieb Paolo Bonzini: > This is an alternative implementation of writethrough caching. By always > opening protocols in writethrough mode and doing flushes manually after > every write, it achieves two results: 1) it makes flipping the cache mode > extremely easy; 2) it lets formats control flushes during metadata updates > even in writethrough mode, which makes the updates more efficient; 3) > it makes cache=writethrough automatically flush metadata without needing > extra work in the formats. > > In practice, the performance result is a wash. I measured "make -j3 > vmlinux" on a 2-core guest/4-core host, with 2GB memory in the guest > and 8GB in the host. > > Performance was measured starting qemu-kvm with an empty qcow2 image, > a virtio disk and cache=writethrough (F16 installation + exploded kernel > tarball in the backing file), and the results are as follows: > > without patch: > real 9m22.416s user 12m8.955s sys 3m46.331s > real 9m21.602s user 12m20.124s sys 3m51.126s > real 9m22.258s user 12m14.900s sys 3m48.595s > > with patch: > real 9m17.617s user 12m16.837s sys 3m48.637s > real 9m18.926s user 12m20.083s sys 3m50.458s > real 9m15.751s user 12m41.944s sys 3m56.447s > > Unlike the RFC, this is tested exactly with these seven patches. The RFC > was tested with follow-up cleanups that hid the problem in patch 3. > > v1->v2: only patch 3 changed, was completely backwards in v1
Are you going to send a v3 for patches 3 and possibly 5/6? Kevin