RAID Performance w/ four Patriot Warp 2 SSDs - Up to 660MB/s

[Guest]

Please feel free to comment about our story entitled "RAID Performance w/ four Patriot Warp 2 SSDs - Up to 660MB/s"

[dmaster97]

Re: RAID Performance w/ four Patriot Warp 2 SSDs - Up to 660MB/s

there are articles around this.

mostly covers about disk partition not aligned.
see Disk performance may be slower than expected when you use multiple disks in Windows Server 2003, in Windows XP, and in Windows 2000

interesting reading at ocztech forum
Partition alignment importance under Windows XP (32-bit and 64-bit)..why it cures stuttering and increases drive working life. - OCZ Forum

microsoft windows address disk cluster best when using partition with first sector align to cluster.

one easiest way i found is format the harddrives using fdisk under linux-live cd.
linux fdisk can specify first n last sector when creating partition.

but if using fake raid, or onchip raid. one must find drivers first so linux can recognise the raid array.

even on single disk, i experience improved performance after creating aligned partition.

[Tommy_Boy]

Re: RAID Performance w/ four Patriot Warp 2 SSDs - Up to 660MB/s

Something that may be limiting the ability of the controller to allow a continually high throughput may be the 256mb of cache on the controller.

It would be interesting to see if changing the cache module to a 2gb DDR2 module made an difference on the ability of the controller to maintain the 350+MB/sec of the first 22% of the graph.

I'm not too well versed in how dedicated RAID controllers work, but if I was being asked to maintain a super high amount of data throughput to a RAID0 array of 4 SSD hard drives, I would want as much cache to play with as possible. And for the price of a 2Gb DDR2 ECC module it may be interesting to follow up.

Cheers,
T

[Efassi]

Re: RAID Performance w/ four Patriot Warp 2 SSDs - Up to 660MB/s

The increased performance for the first 22% could be due to the SSD's keeping an area cleared for immediate writing.

When writing to flash it must be cleared first with a higher energy zap. This is a relatively slow process. Some SSD's can work in the background when they're idle to clear unused sectors in advance. Then when data needs to be written they're ready to write giving much better performance. The 22% mark could be where the pre-cleared area runs out.

If this is what's happening then I would treat the higher speed as the real write rate because there are few instances in everyday use where I would re-write more than 22% of the drive at once.