"Benchmarks"

"Benchmarks"

[Brad Campbell]

I use the term loosely.

This is with the raw drive being a Maxtor MaxlineII 250GB 7200RPM SATA 
unit. It's about a 2004 vintage with over 30,000 hours on it. Slow and 
steady.

The Cache device is an OCZ Vertex Plus 120GB Unit. Not all that fast or 
clever. The Cache has writeback enabled and is formatted with -w512 so 
that direct IO worked as previously discussed.

The benchmark is conducted inside an XP VM running under KVM. The device 
is a qcow2 backing file formatted NTFS. I ran the test multiple times 
without the cache until the numbers stabilised (the backing file was 
being expanded as required)

http://www.fnarfbargle.com/private/111022_bcache/Without_Cache.png
http://www.fnarfbargle.com/private/111022_bcache/With_Cache.png

As expected a significant improvement in random small I/O. The VM is far 
more responsive with the cache.

Re: "Benchmarks"

[Kent Overstreet]

Cool!

The thing I'd be really curious about is what the benchmarks would
look like if you were using only the SSD, no bcache.

How happy are you with it?

On Fri, Oct 21, 2011 at 8:29 PM, Brad Campbell
<lists2009-+nnirC7rrGZibQn6LdNjmg@public.gmane.org> wrote:
> I use the term loosely.
>
> This is with the raw drive being a Maxtor MaxlineII 250GB 7200RPM SATA unit.
> It's about a 2004 vintage with over 30,000 hours on it. Slow and steady.
>
> The Cache device is an OCZ Vertex Plus 120GB Unit. Not all that fast or
> clever. The Cache has writeback enabled and is formatted with -w512 so that
> direct IO worked as previously discussed.
>
> The benchmark is conducted inside an XP VM running under KVM. The device is
> a qcow2 backing file formatted NTFS. I ran the test multiple times without
> the cache until the numbers stabilised (the backing file was being expanded
> as required)
>
> http://www.fnarfbargle.com/private/111022_bcache/Without_Cache.png
> http://www.fnarfbargle.com/private/111022_bcache/With_Cache.png
>
> As expected a significant improvement in random small I/O. The VM is far
> more responsive with the cache.
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-bcache" in
> the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>

Re: "Benchmarks"

[Brad Campbell]

On 25/10/11 01:10, Kent Overstreet wrote:
> Cool!
>
> The thing I'd be really curious about is what the benchmarks would
> look like if you were using only the SSD, no bcache.

I'll see what I can do when I next fire up the staging server.

> How happy are you with it?

Well, it does what it says on the tin for quick and dirty benchmarks, 
but I really can't even contemplate it on a production machine until it 
is compatible with md, and preferably has dirty write mirroring.

Now I have a permanent staging server, I'll keep playing with it as the 
code evolves though.