Recommedn RAID Group size

For highest availability, a 14-drive RAID 6 (12+2) with a single hot spare should also be considered.

IMO in RAID 5 (13+1) would be the best bet for you considering you want capacitiy utilization with performance.

The problem with the 13+1 is that the stripe size is wacky.  With a 832 KB stripe the write cache isn't going to be able to work terribly efficiently. 

The other problem is, that rebuild times on very large RAID groups like that is very long.  Hot sparing is not infallible and drive failures tend to cluster.  You'd have a very long window of vulnerablity on your backup with a 13+1 group rebuilding.

Information on stripe size and its affect on back-end write performance, hot sparing, and rebuild times can be found in EMC Unified Storage System Fundamentals for Performance and Availability.  The Fundamentals document is available on Powerlink.

HTH

I just wanted to add to jps00's great points, while a double-drive failure is certainly possible, the main reason EMC makes the recommendation for RAID-6 at the capacities being propsed in a single RAID group: 16x 600GB (or say 6x 2TB/3TB SATA/NL-SAS drives), is the much higher probability (relative to a double drive fault) of an Unexpected Error Rate (UER). 

During the rebuild not only will it take a while as jps00 mentions, as an operation that has to read in every block, this becomes a concern especially at large capacities.  As drives get even larger, can we imagine a time when we need to look into triple parity?  Again, the main reason would not be to protect from +1 drives failing from what the RAID type is tolerant of, but to protect from UER.

Oops sorry about that jps00.  Thanks for the correction. 

Just to add to it as I failed to mention it, UER is a condition where a single block (versus double drive failure) is unreadable during the rebuild.

The 'Hard drive reliability classifications' section of EMC Unified Storage System Fundamentals for Performance and Availability contains a discussion of Unrecoverable Error Rate (UER) .  Coincidently, the example given uses 600 GB Enterprise drives like the OP describes using. 

The Fundamentals document is available on Powerlink.

Thanks friends for your responses.

I am keen to know the performance difference between the various RAID Group sizes.

Generally speaking, the larger the RG size, the better the performance will be. But it more depends on the I/O profile.

Hello amediratta,

Peromance and RAID Group will entierly depend upon what kind of I/O is coming in.

For example FOR small blocks of data like SQL/Database etc the best RAID TYPE is RAID TYPE-10

However this RAID would require more numbe of disks.

So whenever you choose a RAID Type, the first question that needs to be addressed is .

Availibiliy or performance.

Hi amediratta

I am assuming the shelf of disks (15) contain a HS.

I would recommend you this:-

create a 4+1 and a 8+1. with the remaining disk, use that as HS.

4+1 and 8+1 are the best practice RAID setup for RAID 5.

4+1 should able to give 2149GB and 8+1 give 4298.

in both 4+1 and 8+1, select max. capacity and select bind 2 luns - make sure lun balanced on SPs.

with 13+1, yes you will get maximum capacity however its not a nice RAID configuration because stripe size is "dodgy".

another issue with 13+1 is that rebuild time will take alot longer as well. increase the potential risk if another disk in that RAID setup decide to take a hike while a rebuild is happening.