Home > General > SSD Storage Arrays

SSD Storage Arrays

January 18th, 2009

This seems to be quite a buzz word at the moment. RAMSAN have been doing very expensive, very high performance RAM based SAN (hence the name oddly enough) for awhile, as have a few other vendors. EMC recently announced a SSD based disk shelf, and lots of other vendors have been talking about it. So where does it fit?

My opinions here, we get it to fit in 2 places that compliment some of the other buzz words at the moment.

Flash drives are technically different to SSD and are closer to RAM, so draw a constant power requirement. These have always been expensive and potentially always will be, but they yield huge performance stats.

SSD is quite a mature technology, but finally the drive manufacturers are packing in the capacity. My big chunky Alienware laptop has a 32g SSD drive which cost an extra £200ish. I see a huge performance increase in hibernate speeds, but standard Windows running is much the same. We’re heard announcements of enterprise ready 1TB SSD drives, but these still  cost a fortune.

So given their price, where do these fit? People are now giving us Tier-0 storage! I’m not sure if I would tier this storage at all, these technologies can go up a level and influence the system performance as a whole. If we compliment the memory based cache implemented in all storage controllers with an SSD based disk array, we can grow our cache significantly.

Do we need extra level-2 cache? NetApp for years having been telling us that cache is not the root of all performance woes, and have proved year-in year-out that they can out perform systems with double or triple the cache levels. But then NetApp got involved with de-duplication.

So with de-duplication we have less data blocks on disk, but potentially several blocks accessing the same data blocks, requiring several disk reads still. We are now seeing the cache getting tuned to support de-duplication blocks, but the more we can store this in cache, the more performance we will get.

So put in a SSD based disk shelf that attaches direct to the read cache and you’ve got some amazing performance potentials! Imagine all your Windows VM’s reading the same de-duplicated blocks off the storage, but all this is stored on SSD based storage (all the usual RAID-DP configured, so performant and resilient!). This is significantly slower than level-1 cache, but is going to massively out-perform your sluggish 15k FC spindles.

Lets face it, unfortunately hard drives aren’t getting any quicker, just bigger! So I can see these not necessarily dropping down a tier, but being utilised differently.

Now throw into the mix a transparent storage tiering system, one of my favourites, Acopia. A great technology that sits in front our all your storage arrays and virtualises all this storage. You then have SSD, FC 15k, SATA all within the same namespace and totally transparent to the end user or application, but used intelligently and cost effectively. You have dynamic storage tiering and blinding fast access times.

Step forward into the future and we see this at a block level. I want talk about ONTAP 8 as I’m not allowed to just yet, but lets just start to think about an Acopia type product that gives your entire storage array (all x number of them) a global namespace across CIFS, NFS, iSCSI and FCP. You can now tier your databases, and virtual machines across different levels. You’re directors Exchange database sits on SSD (and they can afford that), your data warehousing sits somewhere between SSD and FC disk, but your HR documents and Sharepoint repository sits on SATA disk, but all referenced from one Global Namespace.

Imagine if we can do block level storage tiering? Alas, I’m dreaming now! Maybe in 5 years time someone will look to roadmap this, but for now I’ll just get quietly excited about SSD drives and the potential they add into the storage world and their potential uses.

Tier-0, Level-2 read cache, whatever it will get called, it’s going to change the way we access storage and the way we tier our data.

General , , ,

  1. No comments yet.
  1. No trackbacks yet.

This site is not affiliated or sponsored in anyway by NetApp or any other company mentioned within.
%d bloggers like this: