Wowza! this makes the Fusion-io solution seem like a ‘good try’ but Texas Memory Systems comes through again with amazing throughput and more!
Alright, crashing! but yowza! ;)
Firstly, let me commend Beth Pariseau for this great article and interview with Fusion-io, including the deep-dive discussions with them!
Now that the formalities are out of the way… time to be not so nice ;)
I like the idea of the Fusion-io devices, I even mentioned them last June, so I’ll call it like it is!
What Fusion-io brings to the table is indeed a nice comprehensive solution to provide high-speed data access in a very small foot-print. This is indeed a fact, I’m sure none of us will argue that they are indeed providing you to have very FAST storage operating in a distributed model, which works out perfectly in a one-off scenario in ways hard to even address!
The business challenges which Fusion-io solves is the difficulty of getting high-speed disk closer to very specific applications (such as grid-computing in ‘x’ number of boxes or OLTP) and it does a pretty good job of that, allowing me to insert their solution into my existing commodity servers!
However, for the one challenge this solution solves it still leaves all of the other pressing issues as this forces us into a distributed fashion, almost contradicting the consolidation efforts which Virtualization, FCoE and and Virtualized storage bring into the Datacenter.
So, I commend the effort, however the implementation of distribution not only increases my risk but it raises question of my scalability of this as a long-term viable implementation. Here’s a top down list of challenges addressed and non-addressed with this implementation.
Infact, when you think about it, the Fusion-io introduction is a clear replacement for DAS in the current datacenter, but it lags so behind conventional (and even archaic) models of SAN implementations that it’d be hard-pressed for any Data Center or Enterprise Architect to use this in any extensive deployment with the lack of scalability, DR/BC sensitivity, HA application and short and long-term backup and archival.
On it’s own, it’s a challenge to see it last and take a significant portion of the Enterprise Storage market as a whole but as a niche player it is king. With offerings like TMS – RAMSAN, the NetApp V-Series RAMSAN Bundle, and other SSD/EFD solutions premiered by the larger storage vendors this will not only continue to be an aggressive play in the future but will set a precedent of things to come.
It just goes to show, storage is dumb – It is how you use it and the intelligence into managing it which is the clear differentiator, and these differentiators will set the dogs apart from the wolves. (or lolcats if preferred)
Fusion-IO has released the ioDrive rather recently, and apparently is backordered! oh my!
This thing looks cool on the surface!
Although it has tiny sizes of 80, 160 and 320gb, nonetheless the possibilities seem rather cool. This can definitely be a great boundary for high-speed disk for small data-sets.
Certainly I have initial concerns around the raid-ability of the disk and the potential losses, albeit it does predict protection from moving disk components – Nonetheless if you’ve never experienced solid state disk failures (As I have) you’re likely to find them to be a realistic problem to need to address.
I’ll personally be watching this one going forward, as they seem to be breaching a boundary of availability and feasibility in the SSD market, especially with the practicality and the sizes of SSD being even easier to deploy (and cheaper!)
If their product works as well as it is proposing to operate, I don’t imagine they’ll be able to survive on the open market for long without getting snatched up!
