For a number of years, at least one major storage array manufacturer has had the ability to run applications in their array. It has appeared on the road-map a few times but never happened. Use-cases were using it as a database offload engine, back-up services, email-archiving and a few other infrastructure related applications; nothing startling but potentially extremely useful.
As other arrays start to gain this ability, at least in theory; I wonder if we will see this idea percolate back up again. For example, if you look at the V-MAX; it has a huge amount of processing power and certainly once it gets on the processor upgrade tread-mill from Intel; I suspect we will see something from that direction.
Actually, if you look at the idea of the software mainframe from VMWare; couple that with V-MAX and with some of the advances made in high speed connectivity; you might have an interesting server/storage cluster.
Of course, if you start looking at things in this light; Oracle’s acquisition of Sun becomes a whole lot more understandable. Joined-up thinking allowing platform plays. Also HP’s courting of Dave Donatelli looks an extremely sensible move on their part; they certainly need someone who can drive this initiative in HP.
We’ll probably need to come-up with some new terminology to describe this because it will have characteristics of both DAS and networked storage.
BTW, we already have done this in the media space with some storage products able to run transcodes in the ‘array’. Perhaps it is time for this to become more mainstream.
Agree, couple of interesting examples right now :-
1) Oracle Exadata – app off-load engine
2) HP Extreme Storage ExDS9100 – with more CPU than it’s knows what to do with on ‘local’ blades
Interesting being able to run Hadoop style apps within the ‘data zone’ and then have the only IO back to the app servers being control metadata and results…
Interesting to be able to run protocol gateways / converter appliances within the storage ‘data zone’, or to be able to run add-on appliances (AV, security, mngt, classification, async transformation processes etc) within the ‘data zone’
Starts to raise the questions of which is the compute Vs which is the storage, could bring a whole new meaning to VSAs, and also easier ways to plug-in functionality and isolate changes etc?