• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Question about hard drive read speeds

rudder

Lifer
Having an argument with a co-worker (well not really an argument, I am just trying to get out of some work)

Lets say I have a raid 5 array consisting of a number 10000 RPM Wide Ultra3 SCSI hard drives. The volume is filled with files that typically are around 40-50k in size. Currently it has a cluster size of 4k.

A co-worker feels that the cluster size should be brought up to 60k thinking that the drive will read the files quicker in one fell swoop.

My thinking is that maybe that would be true if the drive just had one platter and one head. Not only will the large cluster size waste tons of space, I think it would actually be slower because of the fact that there are multiple platters and heads on the drive. Or if we were talking about very large files maybe the large cluster size would defintely benefit performance. It is true that the 50K file will be broken up into 4k chunks, but with multiple platters and heads would it not read just as quickly?

 
Does the RAID array have any particular importance in this? Do you mean stripe size when you say cluster size?
 
When I say cluster size I am referring to block size. The total array size would be 140 gigs. Mainly I am looking at a performace perspective of reading these 40-50k files.
 
Not sure what you mean by block size, but assuming you actually do mean cluster size, in a RAID array, the stripe size will have a greater affect on performance than the cluster size. A larger cluster size will perform better because it reduces the size of the file table. This won't be a perceptible difference unless you do something retarded and try to format a 40GB drive with 512 byte clusters or something like that which would kill performance. The stripe size of the array would have a larger affect on performance because it along with the size of the file would determine how many drives it would be writting to. With a 60KB file, if your stripe size is 64KB it is only written to 1 drive, if you have a 32KB stripe the file gets striped across 2 drives and if you use something smaller, it will be striped across all 3 drives giving better STR performance.

A co-worker feels that the cluster size should be brought up to 60k thinking that the drive will read the files quicker in one fell swoop.

It doesn't matter what cluster size you use, the STR performance of the drive will remain the same.
 
yes, I did not even consider stripe size. It is a hardware raid on a smart array 5300 controller. Not even sure where the setting is.
 
Back
Top