What's Wrong with Intel Matrix RAID

Article Index
What's Wrong with Intel Matrix RAID
Page 2
Page 3
All Pages
No Caching Capability

Enterprise RAID arrays always have cache memory. Usually the modern controllers have 128MB, 256MB or even 512MB. Controllers in SANs can have multiple gigabytes. Why? Because memory is way, way faster than the fastest disk.

If fast hard disks are like the fastest man in the world covering 100 metres in 10 seconds, then RAM acting as cache is a bullet covering that same 100M stretch in a tenth or a quarter of a second.

Cache memory has multiple purposes, and the two most critical are to: a. Hold frequently accessed parts of the disk in faster memory b. Store up data going to the disks so that instead of lots of tiny changes the disk sees one stream of larger changes

The Matrix RAID controller can't do any of that.

Now at this point some of you are undoubtedly cringing and saying things like "But it's a free RAID controller", or "If you don't like it don't use it".

I actually like having the Intel RAID controller. I don't build PCs any more without setting up a mirror set. If a disk dies, the person keeps on using the machine without losing data.

And Intel have the right idea about drivers and management. The same driver used for the ICH7R also controls the ICH10R. And if you swap the board, you should be able to recover the array.

But it could still be better. Memory is so, so cheap. Give the chip an interface to a normal DDR or DDR2 DIMM. I don't care if it doesn't have a battery and only caches reads. But let the user stick a SODIMM on the board and reap the benefits. Heck, even if it's only the server boards at first.

Oh, as an aside to you, the reader; Intel suggests using a RAID 0 set for the Windows page file. Please don’t do this. Don’t put the Windows page file on a RAID 0 disk. Any RAID 0 disk, even if you dedicate the disk to paging. It's a surefire path to STOP error 7B, KERNEL?DATA?INPAGE?ERROR, which usually indicates a dying disk or disk controller. I suspect BSD and Linux might have similar responses (kernel panic) if swap was on a RAID 0 disk.



Comments (0)
Comment
Your Contact Details:
Comment:
[b] [i] [u] [url] [quote] [code] [img]   
Banner