4

An existing LSI raid is using 2 3.5 7200rpm drives. I would like to replace one of the drives with a 2.5 5400rpm drives.

Is this ok? If performance is lowered, it is ok for me.

Magellan
  • 4,451
  • 3
  • 30
  • 53
peter chan
  • 51
  • 1
  • 2

2 Answers2

6

If using the same connection protocol (read: SATA), it should work. However, 5.4k RPM disk are consumer (laptop?) ones, absolutely not adequate for any production RAID.

In other words: if the server does any real work, don't do that. If for lab/testing, you can try.

shodanshok
  • 47,711
  • 7
  • 111
  • 180
4

Usually, the array runs at the speed of the slowest disk, so putting a 5k 2.5" in a 7k 3.5 array probably more than halves the speed.

Also, some RAID controllers particularly don't like mixing disks of different speeds - performance may be impacted even further. You'll have to try though, no vendor explicitly supports this kind of setup.

As shodanshok has already pointed out, for production storage you'll want to match disks.

Zac67
  • 10,320
  • 2
  • 12
  • 32
  • RAID1 reads can ideally be distributed according to which disk currently has a shorter read queue (they're mirrors so you can read either member). So ideally (e.g. Linux software `md` RAID) you should get multi-threaded RAID1 read iops about equal to the sum of the individual disk iops (I think). If your controller insists on alternating between devices, then yes the slow device slows down the fast device. (I have no idea if HW RAID controllers tend to handle this queue balancing). Of course writes have to make it to both disks so you bottleneck there. – Peter Cordes Nov 30 '19 at 21:51
  • 1
    @PeterCordes Yes, *ideally* you're correct. But sadly I have yet to see a RAID controller that does everything perfectly. Linux software RAID is most probably the best implementation you can find (apart from being in software). – Zac67 Nov 30 '19 at 21:58