1

I just ordered a new PowerEdge R650 which will be the primary server for our company. It will run HyperV with 10 - 12 virtuals on it, mostly Windows but some Ubuntu. All the standards: domain controller, file server, web server, Exchange (hybrid so little traffic/storage) etc. The most disk intensive will be a SQL Server, 800Gb of databases expected to grow to 1.1Tb over the next three years before another server replacement.

Currently for the data drive I have 5x 1.92Tb SAS 12Gb enterprise mixed use SSDs in a RAID 5 for a total of 7.2Tb. We are using 3.8Tb of that. The new server will have 6x of the same type of drive on a PERC 755 raid controller. The OS drive will be RAID1 with two 480 SSDs.

Each virtual is backed up fully each night and they are also replicated off site weekly. So I am concerned with fault tolerance but not overly concerned as if we had a failure we'd boot up the replica's and restore anything that changed since the last full backup. What would be the most efficient/tolerant yet fastest config for these? I'm thinking one large RAID5 array isn't even though that's what we are currently running without any issues. Two RAID5 arrays would be 3.84Tb each and I could split up the VMs accordingly. A single RAID10 gives me 5.76 and still plenty of future space but I'm reading/writing to a single array. I could even create three separate RAID1 arrays but that seems like overkill to me but might make more sense then the RAID10.

Any suggestions? In each above scenario (RAID5, 2x RAID5, 1x RAID10, or 3x RAID1) I will have a minimum of 2Tb of expansion space so final disk size is not a concern, just performance/efficiency/tolerance.

ADY
  • 73
  • 1
  • 11

2 Answers2

3

You need to define, what has the highest priority for you:

  1. Performance. RAID10, especially on write operation.
  2. Redundancy. RAID6.
  3. Capacity efficiency. Single RAID5.

In my opinion 2 x RAID5 makes no much sense, because it gives performance and capacity of RAID6, but less redundancy.

Actually, any RAID will provide you enough redundancy, taking into account, that the PERC and the server itself is SPOF. For real tolerance improvement look at Failover Clustering.

batistuta09
  • 8,723
  • 9
  • 21
  • In my head two separate RAID 5 arrays would be beneficial when you have different VM's assigned to different arrays. Leaning twords RAID10. As for SPOF we have thought about this but for the money one server with on site 4 hour response is good enough. – ADY Jan 31 '22 at 19:07
2

If you are not really concerned about wasting space, you should definitely go for RAID10.

You'll get a lot better performance than RAID5 or RAID6, 50% fault tolerance and no heavy rebuilds upon a disk replacement.

Massimo
  • 68,714
  • 56
  • 196
  • 319
  • Thanks. Think this is what we are going to do. We don't "need" the extra space of RAID5 so the speed of RAID10 sound like the way to go. – ADY Jan 31 '22 at 19:08