4

We have 2 servers which have 6 SATA 7200 disks inside (each). We want to configure both servers so that they will have Window Server 2008 R2 Enterprise installed and on top of it Exchange 2010 x2 (for 300 people) as one virtual machine and AD x2 as another virtual machine and maybe some additional VM's (but not that much.

Now we are asking ourselves if it's better to create 3 RAID groups RAID 1 or (RAID 1 and RAID 10 scenario). We're looking for performance and protection. Of course SATA drives were choose because of costs and there's no real option to choose something else.

Idea 1:

  • Raid A -> 2 x 1TB (RAID 1) -> C partition for system + D partition for Exchange logs
  • Raid B -> 2 x 1TB (RAID 1) -> E partition - all virtual machines
  • RAID C -> 2 x 1TB (RAID 1) -> F partition - exchange database

Idea 2:

  • Raid A -> 2 x 1TB (RAID 1) -> C system partition + exchange logs
  • Raid B -> 4 x 1TB (RAID 10) -> all virtual machines, exchange db

Idea 3:

  • Raid A -> 6 x 1TB (RAID10) -> system, all vm's, exchange db, exchange logs

Idea 4:

  • ?

Which of the option is the best? Will having RAID 10 on 4 drives mitigate the problem of putting multiple stuff on one big drive so access speed may be slower if all things at once hit the same target?

2 Answers 2

4

Just use one bit 6-way RAID10 array per server, it's kind of pointless doing anything else as you're using such slow drives and you'll lose some IO performance due to virtualisation anyway (you didn't mention hypervisor or disk access method by the way).

9
  • We will use Hyper-V (as per topic) on top of Windows Server 2008 R2. We will use pass-thru disks most likely.
    – MadBoy
    Jan 9, 2012 at 13:52
  • Also would Exchange benefit from having faster drives?
    – MadBoy
    Jan 9, 2012 at 13:53
  • Missed that in the topic sorry :(
    – Chopper3
    Jan 9, 2012 at 13:53
  • 2
    +1 Just one big RAID10, you'll need to squeeze every bit of performance out of those SATA disks that you can. It's probably too late but I would not run SATA disks as hypervisor storage.
    – Chris S
    Jan 9, 2012 at 13:55
  • 1
    I would NOT o with passthrough discson that pathetic level - and I would say that the disc layout is pathetic for a server of that size to start with. I would go with SuperMicro servers. 24 (!) disc slots in 1 HE. Put in a decent raid card (Adapted, ! series with SSD cache) and start with 12 disc per server. FAST discs - 10k rpm minimum. Raid 10. Add 2 discs as SSD cache. THEN you get perormance. WIth your latyout above youa re limited by IO right from the start.
    – TomTom
    Jan 9, 2012 at 13:59
0

I would definitely go for a one single RAID 10. Multiple RAIDs, separate for database/system/whatever are 90s thinking. In the age of LVMs and virtual disks, resizeable partitions, etc., this is obsolete and greatly impacts performance. And there is no real benefit. (I am answering this question despite the fact, that a lot of folks have already done the same just to emphasize the right answer. Utilize all the bandwidth and go for one big array).

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .