News - TDK takes magnetic-disk densities to the extreme

[scaryjim]

The RAID10's SSD's will be able to handle 10 or more times as many simultaneous connections in a database for instance.

Ah, but price-comparitive you're not talking about RAID10 SSDs. You're talking about one SSD. Just one. Singular. vs an 8 disk RAID array of enterprise-class mechanical drives. It simply doesn't make financial sense - regardless of the IOPS

I have no doubt that the time will come when SSDs will be worth the premium in the enterprise space - I just don't think they currently justify 8x the cost for the same capacity. At 3x the cost they might be worth it. At twice the cost I'm sure they would be. But not at 8x.

[howdee]

SSD's really, really make a huge difference in certain types of workload.

I don't think anyone was disputing this. But for large storage needs SSDs simply didn't match demands yet (I don't think they ever will, not NAND at least), and the reliability side of things is highly subjective with many people (wrongly so) assuming "solid state" somehow magically means their data is safe. There's all kinds of things that can happen and render a SSD non-functional, just as with HDDs. I believe that, for now, we need both of these technologies and they complete each other.

[shaithis]

Your missing a major point.

Many enterprises buy HDD and use only 20-30% of their capacity because of short-stroking them for performance.

Enterprises also have decent backup strategies, making long term degradation moot.

[howdee]

Your missing a major point.

Many enterprises buy HDD and use only 20-30% of their capacity because of short-stroking them for performance.

Enterprises also have decent backup strategies, making long term degradation moot.

I got lost and aren't quite sure who's replying to who any more LOL. I think we all agree, just pointing out different aspects of different storage tech. HDDs have a lot shorter seek times on first 1/3 of their capacity and enterprises (actually, I do the same and I wouldn't call myself an enterprise ) make good use of that fact to increase arrays' performance. The remaining space can be either left unallocated and with it the speed of the primary array intact (some advanced hardware controllers actually even have the capability of allocating the remaining space as hot spares in case of a hardware failure), or alternatively allocated to other volumes for additional storage that would be slower performing but less prone to degradation in case of hardware failures (simple arrays with any combination of striping and mirroring that are fast to rebuild and don't introduce a lot of system overhead). All this is actually achievable and available to us mere mortals as well either in OS (linux, win home server,...), by using software based storage controllers (Matrix can do IIRC up to two different volumes on a single array, perhaps now more?), or even by using simple SAFE33/SAFE50 enabled controllers that are present on many mobos as secondary SATA controllers (they allocate 33 or 50% of the array's space to raid 1 or 10, and the rest is either in separate JBOD volumes or joint together in one large spanned volume). I guess my point is we use the word "enterprise" too often with storage tech and unavoidable forget these (fail-safe, always on-line) methods have been around for long enough that they eventually found their way to consumer grade products as well. Heck, even SSDs actually use internal raid 1 built into their controllers to increase their performance.