Setting up for General storage

staze · Been around the blocks Joined: 15 Oct 2007 Posts: 25

All,

I'm setting up a new ActiveRAID 16TB to replace our aging Xserve RAIDs, and I have a few questions.

1. Since this is general file storage, is it worth hacking the RAID configuration profiles Active provides by default to adjust the stripe/block size? We're resharing it over AFP/SMB largely, and doing some other general storage. No direct editing of video.
2. It's a single unit, set with two RAID5 Luns, for data, and a single RAID1 for Metadata (yeah, wasting two 1TB drives for MD seems like a waste). I'm confused what the Stripe Breadth should be. Block size, I figure, should be 4kb. So... ? Everything is on, or will be on Xsan 2.2.2. Everything is currently AT LEAST 2.1.

Thanks!

staze · Been around the blocks Joined: 15 Oct 2007 Posts: 25

If I'm understanding correcty:

stripe breadth (blocks) = (transfer size (bytes)/number of LUNs in the storage pool)/ block allocation size (bytes)

So:
Transfer size is 1MB.
LUNs = 2
Block allocation size: 4KB.

1,048,576/2 => 524 288 / 4092. So stripe breadth should be 128 blocks? Does that sound right?

abstractrude · Xsan Master Joined: 13 Mar 2008 Posts: 881

why are you not just using hfs+?
general file storage?

rstasel · Xsan Master Joined: 03 Aug 2007 Posts: 120

why not just help with the question or not post anything if you can't?

Seems like a straight forward question, I thought. It's my other account (which I'd love to merge).

General HFS wouldn't give me the ability to use the volume across multiple servers. Wouldn't give me block level storage. Not a SAN. If I wanted a NAS, I wouldn't have bothered purchasing an FC storage array...

So, any real help here? All I am wondering is clarification on the stripe breadth calculation.

rstasel · Xsan Master Joined: 03 Aug 2007 Posts: 120

And maybe I'm reading things wrong, but when you go to create the volume, you can click settings on each pool, and you get "Stripe Breadth", and then in parenthesis (in my case) it says 1MB when I have the breadth set to 256. But this would seem to counter the idea that you divide by the number of LUNs.

Or do you only divide by the number of LUNs if you have more than 2?!

I've just set the breadth to 128 and am comparing the performance numbers... but...

abstractrude · Xsan Master Joined: 13 Mar 2008 Posts: 881

Stripe Breadth has always been a bit of grey area for me.(i know what it is, but real world performance makes this grey in my experience) only because I have been fortunate enough to have test environments to actually test performance on different configurations.

I personally set my metadata stripe breadth to 16. I have my reasons for this, mostly real world performance testing. as volumes get larger or may become larger i up that. 64 is good when we are talking huge volumes with lots of files.

I also like a 16K block size. And this will be even crazier, I set a storage pool breadth to 16,32 or 64. sometimes 128 but never 256. all based on performance testing for that workflow. with 4 luns in that pool. I like building volumes with 4 LUN pools and setting balance for my allocation strategy. As long as that pool of 4 can meet the bandwidth requirements of the workflow.

This has been my best experience with these settings. i hope i dont get flamed:/