number of sectors per block (question)

[Phan7]

Hi folks,

couldnt' find any posts relating to this so perhaps this will be an interesting one...

when creating a new disk in starwind, specifying number of sectors per block...

According to this : http://www.starwindsoftware.com/starwin ... cdpnew.htm
"The Number of sector per block value can help you to improve performance of the target. If the applications that operate with data on the target mostly operate with 4 KB data blocks, it is recommended to set Number of sector per block value to 16. If the application operates with 64 KB data blocks, it is recommended to set Number of sector per block value to 128."

So, does this refer to the allocation unit size on the disk? I'm thinking not, I'm actually thinking SQL Server and 8KB pages =) as it does say "application".

The big question: Is it a good idea is to set this to 32 to better support 8KB page retrieval for SQL Server? And perhaps when formatting this disk, use 8KB allocation unit size, rather than 4KB?


Or is it better to set this to support the allocation unit size that the disk will be formatted with?

[Anatoly (staff)]

First of all I want to inform you that >60% of performance issues are related to disk alligns, so that is why our software have trial period - that gives our users time to pass the tests to tune their systems to get best performance and reliability levels.
In 90% of cases show that 64k aligned systems show better performance, but still your block and stripe size choice should be test based.
Performance of the HA will depend on the performance of the RAID array used.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.
I hope it was helpful.

[Phan7]

Thanks for that, just what I was expecting to hear if I was wrong about the "application" being higher level.

I will do some tests with SQL and report back, in any case I think that the stripe/AU size (or a multiple of) will be best.

I'm not using RAID at the moment but I will be placing a tempDB on an mSATA SSD I have just bought, so to start I will be formatting with 8KB AU, setting 8KB in starwind and obviously SQL will use its 8KB page size.

I will try the same as above but @ 64KB, with SQL page extents in mind!

[anton (staff)]

Run ATTO disk benchmark for a quick test of what block is doing the best read / write. After passing this value increasing I/O size does not scale. So use this block (64KB or 128KB or 256KB) as allocation unit for NTFS. And keep RAID stripe as a multiply of allocation unit size. That's all...

Thanks for that, just what I was expecting to hear if I was wrong about the "application" being higher level.

I will do some tests with SQL and report back, in any case I think that the stripe/AU size (or a multiple of) will be best.

I'm not using RAID at the moment but I will be placing a tempDB on an mSATA SSD I have just bought, so to start I will be formatting with 8KB AU, setting 8KB in starwind and obviously SQL will use its 8KB page size.

I will try the same as above but @ 64KB, with SQL page extents in mind!