On 10/17/14 5:38 , Andre Sachs via smartos-discuss wrote:
> Hi
> 
> We have a bunch of Samsung EVO 840's that we are trying to use as log
> devices on some old nodes to give them a bit of life.
> 
> The nodes are using an LSI 2008 controller and run well.
> 
> When we add the EVO as an slog device writes max out at 30Mb/s. We then
> swapped the log for an Intel DC3700 and got 200MB's on the pool.
> 
> We added the EVO to a pool on its own and we get 350 Mb/s plus write speeds
> just checking that it performs and it does.
> 
> The only thing we noticed is that the EVO has 8K blocks (ashift = 13) but
> in both scenarios as the log device and in the pool itself zfs sets ashift
> = 9, so that does not appear to be the issue.
> 
> Any pointers on where to investigate would be much appreciated ?
> 
> Thanks
>   Andre
> 

The very first thing to look at is how are you actually performing the
tests in these cases? Using a device as a slog is very different from
using the device as part of a primary storage pool. For example, the
first thing you'd want to understand is how do synchronous writes behave
and in particular that will cause synchronize cache commands to fire,
which may not perform particularly well for that kind of device.

Robert


-------------------------------------------
smartos-discuss
Archives: https://www.listbox.com/member/archive/184463/=now
RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=25769125&id_secret=25769125-7688e9fb
Powered by Listbox: http://www.listbox.com

Reply via email to