Attachment uploads are currently disabled. This is a temporary situation and will resume as normal in the coming days.
Solved!
Go to SolutionOne Large Raid vs Multiple Raids
Hello Everyone,
We have a MD3600F, it has a 12 600SAS drives, and I was hoping you can help me decide how to handle the Raid and Lun creation.
this san will store everything we have:
- VMware VM's
- SQL databases (about 5)
- Exchange database
- Users folders and profiles
- Archives
I was planning to create a large raid 10 array (all drives) and dump everything in it, but after some reading, it might be better to separate them and create several raid arrays, at least one for the DB's.
so my questions are:
- Is it better (performance wise) to break the Large raid to two/three smaller raids?
- If yes, will it perform better on a raid 1 (two drives, I can't spare more than two drives for the DB's) than on the large raid 10?
Thank you for your help and time.
I don't believe the MD-series uses separate I/O queues for multiple LUNs. This only applies to Equallogic if I'm not mistaken.
So if everything will be presented to a single host or single hypervisor cluster, I don't think it'll make a difference other than the side-effects.
With the side-effects I mean that using multiple smaller LUNs means that 1 application cannot take diskspace away from another (possibly impacting the other application's functionality), but also means that one application cannot borrow space from another.
Dev Mgr
6 Operator
•
9308 Posts
•
3 Points
4552
1
Posted May 13th, 2013 11:00
A SAN virtual disk/lun/volume cannot be shared between servers unless they are clustered (and/or use a cluster-aware filesystem (like VMFS or GFS)).
With raid 10 more drives pretty much means more performance independent of your read/write ratio (unlike parity based raid where writes get slower the more drives you add, so high write ratio IO could actually get slower with more drives).
However, splitting into 2 6-disk raid 10's does mean that if one of the 2 is being used heavily, the other isn't affected by this (so you can split your high IO data between the 2 diskgroups).