[Sheepdog] [PATCH v5 4/8] sheep: teach sheep to use object cache

huxinwei huxinwei at huawei.com
Tue Mar 27 06:01:31 CEST 2012


> -----Original Message-----
> From: Liu Yuan [mailto:namei.unix at gmail.com]
> Sent: Monday, March 26, 2012 6:08 PM
> To: huxinwei
> Cc: MORITA Kazutaka; sheepdog at lists.wpkg.org
> Subject: Re: [Sheepdog] [PATCH v5 4/8] sheep: teach sheep to use object cache
> 
> On 03/26/2012 05:26 PM, huxinwei wrote:
> 
> > My point is about _different_ VMs reading the same object.
> > It doesn't matter how you optimize guest here ...
> 
> 
> We can't read the same object from mutiple VMs right now. (Farm
> originally supports naming the object by hash content for data
> deduplication)

We can, if these VMs are actually cloned from the same snapshot ;)

BTW: I'm not aware that you are planning data dedup already for farm.
That'll be really awesome ;)
However, 4M is far too big for effective deduplication, IMHO.
It seems we need a patch to change the size of object, e.g. 128K as ZFS.

> If we do, yes, in this scenario, we would benefit some workloads (mostly
> read, seldom write) from a unified and content-named cache.
> But for a general cache, if we have a unified cache (be it in memory or
> on disk) that serve a lot of VM requests concurrently, I think we would
> meet the lock contention problem, hardly to be scalable.

Agreed.

> So we might later have a unified cache complementarity(a transparent
> write through cache) in memory for this kind of read acceleration on top
> of object cache, which just aims to be generally reducing the network
> traffic by absorbing IO requests locally.
> 
> Thanks,
> Yuan



More information about the sheepdog mailing list