SSD RAID0 Scaling

[Boogerlad]

Hello everyone. I'm planning on ordering a large amount of ocz vertex plus for raid0, but I'm not too sure about the scaling factors. I'm sure many people on xs have 2 or more ssds in their storage array, so this is the perfect place to ask. So I know that, given the right controller:

SSDs scale linearly in sequential writes,
sequential reads,
random read & write at high queue depths,

BUT access time stays the same or gets worse because of overhead from the controller. What I don't know is random read and write at low queue depths. (0-4) Could anyone provide some benchmarks of single and raid0 random read and write at low queue depths (0-4) with iometer or as-ssd?

[One_Hertz]

Raid can not improve things for small block QD1 random reads. It improves results very slightly for QD2-4 random reads.

For the low QD random writes, it depends on the controller/setup. There can be a lot of improvement.

[Boogerlad]

Let's say p67/x79 6x/10x ocz vertex plus. As for stripe sizes, according to anandtech

SSDs are made up of millions of NAND flash cells. They can be written to in groups called pages (generally 4KB in size) but can only be erased in larger groups called blocks (generally 128 pages or 512KB). These stipulations are partially the source of many SSD performance issues.

would that mean that 512kb stripe be the best?

[Anvil]

What's it for and why the Vertex Plus?

I'm not implying there's anything wrong with the VP, could be that it's a perfect drive for your purposes.

Intel (and most other on-board controllers) limit strip size to 128KB, 64-128KB looks to be a great all-round strip size.

[Boogerlad]

I chose the vertex plus because I can get it for less than $1/gb, and in non compressible transfers, it matches the ocz solid 3(the cheapest sf2281)

This is for an os drive, and lossless video recording, 1080p.
I know I can achieve the sequential speeds I want, but would raiding benefit 4k, which in turn benefit my overall os experience?

[Computurd]

Raid can not improve things for small block QD1 random reads. It improves results very slightly for QD2-4 random reads.

partially true here. for QD1 it will not scale. for 9260 here is 8 vertex with fastpath. there is scaling there. cant believe im still posting vertex results LOL

[Boogerlad]

would you happen to have results for 4k random writes?