Sign in with
Sign up | Sign in

Conclusion

SSD RAID: Do You Want A Cheap Array Or One Larger Drive?
By

This article aims at enthusiasts who want to maximize storage performance and those who want to know how to best spend their money: by either purchasing a single SSD or by going with a couple of low-budget, low-capacity solid state drives. Since most retail motherboards come with RAID functionality built-in, it makes sense to put an inexpensive SSD-based RAID array up against one powerful SSD. However, we found that this makes only limited sense. Let’s look at some details.

Zalman’s entry into the SSD segment is successful. The N-series SSD is based on the SandForce SF-1220 controller, which currently is one of the best architectures available. Companies like Adata, Corsair, G.Skill, and OCZ all offer solid state drives based on this platform, and they all deliver comparable performance and efficiency. The throughput of Zalman’s SSD0128N1 is at the top end of the SATA 3Gb/s range. PCMark Vantage storage application performance is nice as well, and the I/O performance of all SandForce SF-1200 SSDs is as high as it can be. Once again, we can only recommend looking at prices and warranty, as the performance differences between several SandForce products are too small to be turned into purchase criteria. The bottom line is that the Zalman N-series SSD is a great product.

Now let’s take a look at the SSD RAID performance results, as these aren’t as clear. First of all, we need to repeat that Kingston’s SSDNow V 30 GB drives are being replaced by the V100-series, employing a JMF-618 controller by JMicron. This means that the 30 GB capacity will be gone soon. But it could also represent an opportunity if stores want to clear their inventory. However, the SSDNow V 30 GB isn’t a great performer when it comes to I/O performance or write throughput. This is because of the Toshiba controller, which just isn’t very strong, and can only convince in sequential read testing.

Yet, two or even four of these entry-level drives will boost performance levels quite a bit. We found that throughput increases way over 600 MB/s and that write performance is much more consistent on RAID setups than it is on an individual drive. While almost all results benefit from utilizing two or even four drives in RAID 0, the RAID setups still cannot make up for the weaknesses of the aged Toshiba controller. Throughput is stellar; everything else is not.

Users who primarily want sequential throughput can of course consider an SSD RAID array. But for everyone else, we recommend investing in a single more enthusiast-oriented SSD for similar money. We did not only come to this conclusion because of the ambivalent performance results, but also because of the potential trouble you can run into if a RAID 0 array should ever fail. Since your entire array is lost if only one drive fails, we recommend staying away from such configurations unless really necessary for performance reasons--and back your data up regularly. The results would change if you were to use faster products for such an array, but then cost would increase quite a bit.

We didn’t measure power consumption or efficiency, because these items are mostly negligible if you aim at maximizing storage performance. Efficient SSDs require only 2-3 W at peak load situations, and using up to four would result in 10-15 W maximum power consumption based on modern drives. Have a look at our last SSD roundup for power consumption details. Zalman’s N-series is as power efficient as other SandForce drives, and the Kingston SSDNow V 30 GB is similarly low because it runs fewer flash memory channels.

We'll also dive into the implications of using an SSD-based RAID array on TRIM support in an upcoming story, so stay tuned!

Ask a Category Expert

Create a new thread in the UK Article comments forum about this subject

Example: Notebook, Android, SSD hard drive

Display all 6 comments.
This thread is closed for comments
  • 1 Hide
    mi1ez , 3 December 2010 14:48
    First thoughts, before even reading the article, I would rather have TRIM.
  • 1 Hide
    mi1ez , 3 December 2010 15:23
    I would be more interested in RAID5 with 4 drives. Will look forward to the follow-up!
  • 2 Hide
    LePhuronn , 6 December 2010 20:29
    The big thing I get from this article is, regardless of single drive or array, you want a decent SSD to begin with - RAIDing up those (comparitvely) poor Kingston drives doesn't overcome their inherent flaws against other SSDs.

    What I'm looking at personally is 2 60GB OCZ Vertex 2E against a single 120GB Vertex 2E (or 3 60GB vs a single 180GB) - the former is only £20 more, but will I get a performance boost to justify it?
  • 1 Hide
    Gonemad , 7 December 2010 17:02
    I would compare instead a Ramdrive with 20GB and SSDs raided on the same budget. And then find out if there is a sweet spot combining both, instead of trying to compare outdated SSDs with a larger new one.
  • 0 Hide
    acer0169 , 9 December 2010 16:13
    Who has the money to have SSD in raid!? my god.
  • 1 Hide
    Anonymous , 22 December 2010 17:34
    Agree with LePhuronn's comment above, in that there's little point comparing shonky SSDs in R0 with a single much better one... You want apples with apples, not apples with spider monkeys or sloths or...

    The conclusion starts with "This article aims at enthusiasts who want to maximize storage performance and those who want to know how to best spend their money: by either purchasing a single SSD or by going with a couple of low-budget, low-capacity solid state drives.", however these are clearly not the only two choices for ~the same expenditure & it's foolish to suggest that they are.


    Okay, it's possible to go all out (atm am currently running 4x 50GB OCZ V2s in R0 on a 9260-8i & the speeds (not simply sequential ones) are pretty snazzy...), but the cost differential between 1 larger 3Gb/s SSD & 2 SSDs of the same model/brand that are half the size (which can be run in R0 with no bandwidth limitation on at least the latter ich controllers) is minimal...


    So just seems like a bit of a half-arsed experiment...