Block Size on RAID Disk Array?

**ChrisK2972** · June 26th, 2003, 01:05 PM

I am setting up a RAID controller (Adaptec 1200A; Highpoint chip-w, 2- Maxtor 40G 7200 ATA133) and need some suggestions about block size. My options are 16, 32, 64; what would be the best size? I am thinking that I can get less waste with a 32 block size? I do very little video editing, so, I do not think that I need a large block size. What would you use a 16KB block size for?

Thanks for the help

ChrisK

**Siliconjunkie** · June 26th, 2003, 01:07 PM

I would think that 16 would waste less space since the total #/size of the blocks could be within 16k of the actual size of the data. I could be completely wrong though.

**ChrisK2972** · June 26th, 2003, 01:16 PM

I was reading my manual for a Promise Fast Trak 100 controller. It indicates that a small block size would be used for a file server where there were numerous small files. Large block size would be better for like Video editing, ect, where you are working with primarily large files. A medium block size is used for a typical home computer where there is usually a combination of both large and small files.

I am not sure how accurate this information is, but, it kinda make sense. Anyone know for sure what the best route to take is? I mean even if you do do allot of video editing, why not set the size to 16 so as to waste less space. What you say, Siliconjunkie, sounds logical to me. But, computers are not always logical

**waterdd** · August 2nd, 2003, 02:33 PM

Let me expand on the original question....

What are your thoughts on the NTFS Cluster size vs. the RAID block size? Should they match?

In my case, I mainly use my computer for internet browsing and DVD ripping. So I set my RAID volume for a block size of 64K (for video editing) and then partitioned it into different "virtual" drives, each with different NTFS cluster sizes. For example, the drive for Windows XP and programs has a cluster size set to 4K, but the drive for my DVD rips is set to 64K.

By doing so, I thought I would get the benefit of both worlds, but sometimes I believe the RAID block size it too big for my day-to-day stuff and perhaps it slows my computer down when it has to handle the small files. So maybe I should have set my RAID block size to 32K as a comprise.

Dave

**waterdd** · August 3rd, 2003, 09:17 AM

Originally posted by Steve R Jones
If you’re doing raid 0 for preformance, the large the block size the more “preformance at a price” you get.

Can you expand a little more..."the more performance at a price"...what price?

**Kuasimodem** · August 3rd, 2003, 01:00 PM

I'm running WindowsXP Pro SP1, NTFS, RAID 0 with 2 80Gb Maxtor ATA133 drives (3 partitions + 2Gb Pagefile partition) on a KD7-RAID mobo with onboard HPT372 controller. I'm running a 4k cluster size on XP with 16k stripes on the RAID array. This way I get 4 clusters per stripe, and a 1k text file will only take up 1 4k cluster.

My Sandra benchmarks give me a rating of approximately 41000.

Go with 16k stripes, leave your NTFS cluster size at the default 4k, and you will be very happy with the performance/disk size ballance

**waterdd** · August 3rd, 2003, 02:05 PM

So let me see if I understand you correctly...
If I set my RAID block size to 64K, then my little 1k text file will be written to the hard drives as 64K? ...and this is done regardless of what my NTFS cluster size.

For example ... in one of my partitions, I have the cluster size set to 4K, with my stripe block size set to 64K (so that's 16 NTFS clusters per RAID block, right). So if my computer decides to save this little 1K text file, will it write the text file into one NTFS cluster, then "pad" the rest of the RAID block with 15 empty clusters?

Dave

**Kuasimodem** · August 3rd, 2003, 05:42 PM

Close... The other 15 clusters will be filled with other files. Windows doesn't see the striping of the RAID array, the stripes are only a reference for the RAID controller. Windows sees one huge hard drive (in my case 160Gb).