Clariion Drive Swaps (Smaller to Larger)

The speed of the replacements really depends on the priorities used and what is going on with the box at those times (IO wise). Take a look at this white paper:

http://powerlink.emc.com/km/live1/en_US/Offering_Technical/White_Paper/H4153-the-influence-of-priorities-on-lun-mgmt-optns-wp.pdf

When you replace a 146GB drive with a 300GB drive in a raid group, the extra space will no be available until all drives in that group are 300GBs in size. Also, if a 300GB drive happens to fail when there are still 146GB drives in the group, a 146GB hot spare could invoke to it seeing you will have less that 146GBs of data on that drive.

Are you replacing 15k rpm drives with 10k drives? Slower does mean less IO. Does performance matter? If not then ignore this next part. If so, then you have to keep in mind you are putting more data on the drives, which can mean more IO, which could mean more problems. Just my 2¢

-Ryan

Hi,

That is perfect, exactly the info i was looking for.

Replacing 10K drives with 10K drives, just need more space and the drives are readily available in the organization as well.

Thank You
Devang

You may want to seriously consider the option of replacing the 146GB 10k drives with 300GB 15k drives. Ryan pointed out the downside of dropping to a lower performance drive, but I'd be worried as well that the larger capacity might cause lower performance on the same speed drive. It all depends on how heavily utilized the drives are today, and how you would be using the additional capacity after you upgrade them.

More data means more seeking, which means longer average service time for each IO. Plain and simple.

-Ryan

True, but if the drives are currently being used for something that isn't driving the performance characteristics (e.g. archive file servers), then there wouldn't be a need for any more performance if they are just increasing the capacity for that function.

If on the other hand they are using the additional capacity to support some new Exchange Servers (for example) then they should likely be more concerned.

I know that every change to the configuration should consider the possible performance implications, but the reasoning behind some changes make it more critical than others.