HPE Storage Users Group

Guys,

I've always been told that service times for disks should be under 15ms otherwise you notice at the hosts.

So in system reporter, when you check this, does thie refer to write / read or total or should they all be under 15 ms?

Also any idea if it makes a difference if you're using NL or FC and if so what the ms delay should be?

Thanks

koopa79 wrote:Guys,

I've always been told that service times for disks should be under 15ms otherwise you notice at the hosts.

So in system reporter, when you check this, does thie refer to write / read or total or should they all be under 15 ms?

Also any idea if it makes a difference if you're using NL or FC and if so what the ms delay should be?

Thanks

I believe per HP definitions that "disk latency should be over 25ms before it is considered a latency issue". This refers to the total latency of the disk which you should be able to see from a statpd command

Thanks for the reply.

So on system reporter you get a read, write and total. I assume that's the same total as you see in statpd?

Do you know if 25ms is any different on the NL to FC or is it the same figure?

Thanks

koopa79 wrote:Thanks for the reply.

So on system reporter you get a read, write and total. I assume that's the same total as you see in statpd?

Do you know if 25ms is any different on the NL to FC or is it the same figure?

Thanks

That is the same info you would see in system reporter correct. I believe that technically the 25ms is the same for NL and FC however in practicality i would assume that if you are hitting 25ms on FC drives you are on the verge of having a performance problem if not already having one.

Hope that helps!

I would say use which ever one is greater, read or write ms.

However, wouldn't it be better to monitor the actually response time of the VVs to the host(s)? I think generally speaking MS SQL server starts to put entries in the logs about slow disk when its response time exceeds 15ms.