On 9/4/13 10:49 PM, Andrey Borzenkov wrote:
В Wed, 04 Sep 2013 12:04:58 -0400 Jeff Mahoney <jeffm@suse.com> пишет:
- VM image performance - Performance is generally regarded as horrible. - This is because CoW on what is essentially a block device backing store means a ton of write amplification for each write that the VM issues.
Why VM is special cased here? What makes it so different from any normal write activity?
That it's significant overwrite activity with a strong desire for snapshot functionality. The typical database workload wants /just/ overwrite. The typical VM workload wants both.
Did anyone look at what Solaris does? They must have hit the same problem with LDOM on ZFS years ago.
Nope. We haven't examined ZFS behavior. That gets tricky WRT legality. -Jeff -- Jeff Mahoney SUSE Labs