A (hopefully) quick question about disk block sizes

[ralphw]

I have some old notes basically saying that for perfomance reasons, it's best to use 64k block sizes as much as possible. Is that still the recommended setting? I know there was some talk about a new release doing something with block sizes, so I just wanted to confirm.

So, (as an example) for a simple deployment of using Starwind SAN on Windows 2008 R2 to serve Windows OS guests from a VMware host:

1) We've got a dedicated array that we'll put our Starwind targets on. It is configured as a RAID10 with 64kb stripe size.
2) In Windows on the Starwind server, the volume where the targets are stored is formatted NTFS with a 64k allocation unit size.
3) In VMware, I don't think I have the ability to use 64k as a block size for VMFS, so I've been letting the max size of the virtual disk dictate what I use for the block size (1mb = 256gb, 2mb =512gb, etc).
4) In the Guest Windows OS, format volumes NTFS with 64k allocation unit size...though I don't know of a way to do this to the system drive (C) since it doesn't ask during Windows setup and it takes the default 4k for that volume.

Is what I'm doing still correct?

[Anatoly (staff)]

Hello ralhpw,

I would recommend you to use GPT formatting on the Windows side on the volumes where the StarWind related files are located.

And a little bit about block sizes:
Recommended RAID for implementing an HA are RAID 1, 0 or 10, RAID 5 or 6 are not recommended due to low write performance.
The performance of a RAID array directly depends on the Stripe Size used. There are no exact recommendations of which stripe size to use. It is a test-based choice. As best practice we recommend at first step to set recommended by vendor and run tests. Then set a bigger value and run tests again. In third step set a smaller value and test again. These 3 results should guide you to the optimal stripe size value to set. In some configuration smaller stripe size value like 4k or 8k give better performance and in some other cases 64k, 128k or even 256k values will give better performance.
Performance of the HA will depend on the performance of the RAID array used. It’s up to the customer to determine the optimal stripe size.

[ralphw]

Thanks for your response Anatoly.

Is using a GPT more to get over the disk size limitations of MBR? Or is there some performance benefit to this also?

Sounds like I'm good on the RAID 10, if I'm reading that correctly. RAID5 or 6 are the bad ones. How about RAID50?

And as far as stripe size, it sounds like there's no set rule and I just need to test. But once I find a good stripe size, I should carry that down to the block/allocation unit size in the OS, right?

[Anatoly (staff)]

You are welcome:)

Is using a GPT more to get over the disk size limitations of MBR? Or is there some performance benefit to this also?

Both actually - it eliminates size issue and excludes one of the possible reason of bad perfromance

Sounds like I'm good on the RAID 10, if I'm reading that correctly. RAID5 or 6 are the bad ones. How about RAID50?

Actually as I mentioned - the choice is test-based. If you are going well with RAID10, then I see no reason to change something.

And as far as stripe size, it sounds like there's no set rule and I just need to test. But once I find a good stripe size, I should carry that down to the block/allocation unit size in the OS, right?

I would recommend you to use same sttripe sizes on the disk and in the test simultaneously - you gotta change on both if you want to change something. Also please note that 95% performance issues are related to disk allignes.

[ralphw]

Thanks again, Anatoly!

[Anatoly (staff)]

You are welcome!