Anyone storing VMs on an external RAID "disk"?

[gbullman]

The recent thread about storing VMs on an external disk and the merits of the various disk interfaces got me thinking about how much improvement in the storing disk performance will be noticeable when using the VM.

I'm trying to figure out what the limiting factor is in various configurations. I'm assuming that in each case the physical disks themselves will be 7200 RPM SATA drives.

Does it matter if they are 2.5 inch or 3.5 inch as far as throughput is concerned?

Does a FW400 interface limit the observed throughput of a 7200 RPM SATA drive? Same question for FW800 interface?

Will FW800 be the limiting factor in a 2 disk RAID 0 configuration?

I'm debating the merits of an eSATA 2 disk RAID 0 device to store my VMs on. Will I see the same performance with a FW800 interface with only 2 disks? With the type of disk access that is typical for VMs running mostly productivity applications will I notice any difference at all. Some of what I have read suggests that only HD Video editing will really benefit from a higher end eSATA RAID 0 storage set up.

I know these questions are not Fusion specific, but my main use of the external storage will be for hosting VMs. I'm hoping someone out there has tried out different configurations and can at least provide some anecdotal feedback.

[Mikero]

You might find this helpful:

http://www.hanselman.com/blog/VirtualMachinesAndExternalHardDriveThroughput.aspx

It's a bit old (2006), but still relevant.

I would say if you're doing activities which rely on lots of HD I/O, than a RAID 0 config would certainly be nice.

If you have a multi-GB Outlook data file, it might help with that too

Going over Firewire vs. eSata, I don't think you'd notice a difference. I would use which ever bus is less used. So, if you have a large chain of Firewire devices, use eSATA. Otherwise, Firewire800 is a good choice.

Hope that helps 😃

[WoodyZ]

JSYK and speaking in general RAID 0 while it offers increased performance it does nothing for fault tolerance. In other words with a 2 disk RAID if you want long term storage fault tolerance you need to use RAID 1 but if all you're looking for if a performance boost and you have full backups stored elsewhere if not on a RAID 1 or RAID 5 (4 Disks) then in multiple locations then using RAID 0 can have benefits but how much with Virtual Machine's will really depend on what you're doing with them. I would think unless you're doing something that is disk intensive you're not going to benefit enough to run that way.

[blackpuma]

I have a handful of VMs in that configuration…

Just want to echo the non-safe nature of RAID-0. I don't know why it has the label RAID because there is nothing Redundant about RAID-0. The danger of disk failure increases with each disk added to the array. We're stuck with the nomenclature, though.

Keep multiple daily backups, and understand that you may end up losing a day's worth of work when one of the hard disks give out.

I can't comment on eSATA, but FW800 has been solid for me. The limiting factor that I see with a two-disk setup is disk rotation speed. When doing large copies, the sustained data transfer rate is directly tied to the portion of the disk being copied.

I've never seen the FW bus saturate (except in micro bursts that hit the hard drive cache).

[gbullman]

Thanks to both of you.

I do realize that RAID 0 is a performance configuration and will maintain my current practice of copying the powered down VM to another drive as my backup strategy (so far so good, has saved me twice so far with minimal data loss).

I am questioning whether I will see any real world performance boost with my current use of my VMs. I am in the market for more storage anyway, and in the grand scheme of things the RAID 0 unit I am looking at is not that much more expensive than a comparably sized single disk unit (about 25 to 30% more) that I probably would have bought.

My intent is to attach the RAID 0 to an ExpressCard/34 eSATA interface on my MacBook Pro and leave my existing external drives attached to the Firewire 400 & 800 buses. With the multiple interfaces of the RAID unit I will be able to do some testing of throughput when copying large files (namely the VMs) and I'll see if I get any perceived performance improvement when using the VMs. I still have to believe that disk I/O is the slowest piece of my system and want to get a sense where I might notice speeding that up.

I'll try to remember to post back to this in a couple of weeks with what I am observing.

[blackpuma]

IMHO you're on the right track. Bus data transfer rates outstripped the moving mechanical parts of hard disk I/O long ago, so my gut says you'll see no statistically significant difference between the eSATA and FW800 buses.

Best wishes.

[gbullman]

I wanted to report back on how things are going. It has been a little tricky getting the RAID device to mount on my MacBook Pro. I now have a set of steps that work reliably each time and have a case open with the vendor to see if I can improve upon that. I'm using the eSATA interface because my Time Machine disk is only capable of Firewire 400 and I'm assuming that if I have a FW400 device connected it brings the whole FW bus down to 400 MB/sec. Copying my VMs for backup purposes is about 1 1/2 times faster, it would probably be better but the source or the target remains an FW400 external drive (2.5" 7200 RPM). The sustained throughput of the RAID device with eSATA interface is on the order of 3 to 4 times faster than my previous drive containing my VMs. I used to get about 17 MB/sec when copying between FW400 drives, and about 25 MB/sec copying between my internal drive and one of the FW400s. Now I get about 20-25 MB/sec copying between the RAID and one of the FW400s and a very nice 50-75 MB/sec copying from my internal to the RAID. During boot up of the VMs I saw spikes of 31 MB/sec which would not have been possible with the FW400 drive.

So far the VMs seem to boot up a little faster, they feel a bit more responsive overall. The biggest change I've noted so far is my weekly virus scan was much less noticeable while it was in progress and it seemed to complete in less time than before. Overall I'm happy with the bump up in speed. Seems to confirm what I've suspected all along, disks are the slowest piece of the puzzle, whatever one can do to speed them up will be noticeable.

[blackpuma]

Thanks for coming back with the report. It's interesting.

It's also consistent with my personal experience over the years.Lots of RAM and fast busses make for happy VMing.

[gbullman]

Well, I found the activity where the RAID 0 device + eSATA really shines. A maintenance cycle of Disk Cleanup, DeFrag, and Shrink Disk just took about 1/4 the time it took with FW400, 7200 RPM 2.5" drive. With Activity Monitor I was seeing I/O rates well in excess of 100 MB/sec (caught some as high as 130). Now it is relatively painless to do the period maintenance to keep the VMs as small as possible in the first place (which results in faster backups).