RAID 0 ON SSD'S FOR OS?

Why would you want RAID 0 on the OS? A slightly faster boot time....then what?
 
Why would you want RAID 0 on the OS? A slightly faster boot time....then what?

Faster boot time. instant response when opening or working with large files on the same hard drive. games load a lot faster. RAID 0 on the OS drive for those of us that don't like waiting that extra few seconds. I may be impatient, but my RAID 0 OS drive doesn't slow me down anymore. :D
 
with the improvements of ssd's being so large over such a small period of time. getting two ssd for raid 0 is a waste of money. i bought 2x 60gb ocz agility for around $200 a pop only to watch the sandforce drives take the market, then the micron drives.

if i had a chance to redo it, i would just buy the last gen drive and choose to upgrade frequently then sink the investment of a raid array. the newer controllers offer incredible performance increases that makes raid not worth the trouble.
 
Personally I wouldn't use raid 0 of SSDs for my own OS drive due to the loss of TRIM, and I have found the performance of a single SSD to meet my high standards. I would be more likely to just by a larger SSD which typically offers higher performance (like a 256GB 6gbps Crucial C300).
 
but with newer drives on their way of saturating the 6 Gbps spec...

what? they arent even close to that. SSD's that support SATA3 only show a maybe 50-100MB/sec advantage over the same drive being plugged into a SATA2 port. will be a long time before a single device saturates a SATA3 link.
 
BTW: If you have good garbage collection (GC) on a drive, especially on a desktop where there are periods of inactivity for algorithms like the SandForce's GC, then TRIM in RAID is not a big issue.

A lot of people read about TRIM and totally forget that you can basically get the same effect over time with GC. The difference is between instant and a few hours to recondition a drive, but unless you are benchmarking constantly, you don't write/ delete enough data to *NEED* TRIM v. GC.

One caveat is that by GC, the newer generations are better so you want something at least as good as the Indilinx Barefoot GC and not some of the poor implementations on early drives.

Case and point, I have been using OCZ Vertex RAID 0 120GB SSDs for about 18 months with no TRIM. I think the GC firmware came out after month 6 so basically after one year of use, and only GC I'm at about 91% of original performance with only 50GB free. I also have larger Sandforce arrays which have better GC in terms of reclaiming performance, but they take longer to recover than the aggressive Indilinx GC.
 
I say the faster the better. Just ordered a ibm 1015m raid card to run 2 c400s 64g when they come out. Im running them raid 0 should get around 700meg reads don't see how that can be a bad thing. Plus have room for my OS and games.
 
A lot of people read about TRIM and totally forget that you can basically get the same effect over time with GC.

Not really. GC without TRIM is inferior to GC with TRIM unless the SSD is almost full (or unless you have an SSD with filesystem-aware GC, you are using a supported filesystem, and you are NOT using RAID).

After a drive has been thoroughly used (i.e., every LBA has been written to, even if the SSD is not currently full), then without TRIM the SSD only has the spare area (aka overprovisioned area) for the GC to work with. That can certainly provide some benefit, but it will not be as good as having the spare area PLUS (say) 20% of the SSD capacity to work with, as you would with an 80% full drive with TRIM.

If you take a fresh SSD (or secure-erased SSD) and leave some unused (unpartitioned, etc.) area, say 20%, and if the GC algorithm on the SSD is capable of using that extra space, then GC without TRIM could perform similarly to GC with TRIM on an 80% full SSD.
 
Not really. GC without TRIM is inferior to GC with TRIM unless the SSD is almost full (or unless you have an SSD with filesystem-aware GC, you are using a supported filesystem, and you are NOT using RAID).

After a drive has been thoroughly used (i.e., every LBA has been written to, even if the SSD is not currently full), then without TRIM the SSD only has the spare area (aka overprovisioned area) for the GC to work with. That can certainly provide some benefit, but it will not be as good as having the spare area PLUS (say) 20% of the SSD capacity to work with, as you would with an 80% full drive with TRIM.

If you take a fresh SSD (or secure-erased SSD) and leave some unused (unpartitioned, etc.) area, say 20%, and if the GC algorithm on the SSD is capable of using that extra space, then GC without TRIM could perform similarly to GC with TRIM on an 80% full SSD.

So that is true, GC + TRIM is better, but I was more focused on the idea that RAID without TRIM is terrible. It is not with GC, and most users, so long as they do have empty space, will be OK with good GC.

At the point where drives ultimately fill, TRIM is of course better when a large delete then write scenario is undertaken, especially with slower GC like SandForce's. Then again, I would be willing to bet most users are not writing and deleting more than 10GB/ day on average in non-benchmark scenarios so GC can keep up. Of course, if you had like 230GB on two 120GB drives, deleted 220GB of data (leaving 10GB) and then wrote 100GB, GC would not help much.
 
So that is true, GC + TRIM is better, but I was more focused on the idea that RAID without TRIM is terrible. It is not with GC, and most users, so long as they do have empty space, will be OK with good GC.

I agree that RAID without TRIM is not going to have terrible degradation if there is a decent GC implementation.

Of course, GC does better if it has more spare area to work with, at least the good GC implementations do. So, for example, if one were to RAID four Intel X25-M 120GB drives without TRIM, you could improve the fully-used performance by reserving some extra space on each SSD, say 20GB, so only use 100GB on each. The problem is that I do not know that many of the hardware RAID cards will allow you to set aside 20GB on each drive. It is easy to do with linux mdadm, but I don't know how to do it, for example, with Areca cards (I know you can truncate it a little to allow for slightly different HDD sizes, but I don't recall seeing a configurable "leave unused" box)
 
what? they arent even close to that. SSD's that support SATA3 only show a maybe 50-100MB/sec advantage over the same drive being plugged into a SATA2 port. will be a long time before a single device saturates a SATA3 link.


Did you click on the link http://www.bit-tech.net/news/hardware/2011/01/06/ocz-vertex-3-announcement/1 of the OCZ Vertex 3 coming out with a 550 MB/sec on the very first post? SATA3 theoretically has the bandwidth of 600MB/sec and 550 MB/sec is sure damn close in my book!
 
Back
Top