If you are like me, you are still trying to get your head around the whole software-defined storage thing. Some folks insist that it is an entirely new architecture, but I remember running into one of the folks who manages IBM’s System Managed Storage at their IBM Edge show last year who shook her head wearily and noted that IBM SMS was doing what SDS is supposed to do all the way back in 1993. Once more, everything old is new again.
At IBM Interconnect 2016, I had the wonderful opportunity to chat with IBMers about their go-to-market strategy for software-defined storage. Specifically, I met with Diane Benjuya, Marketing Manager for Spectrum Accelerate and Doug Petteway, Senior Offering Manager, Spectrum Accelerate, regarding their forthcoming products. Anyone familiar with IBM wares will recognize the root technology, XIV, which IBM purchased from its Israeli developers a few years ago and fielded as a grid storage array. Over the years, it has become quite popular, according to both Benjuya and Petteway in the video interview below — so much so, they say, XIV software is being ported to an x86 server to provide the basis for a Hyper-Converged Infrastructure appliance (hypervisor+server+storage+XIV).
Here is the interview.
While I thank Diane and Doug for their time and enthusiasm (despite tremendous jet lag and the last minute invite), I was sorry that the interview left me with more questions than we had time to address properly. I wanted to understand how IBM arrived at its definition of that amorphous term, software-defined storage. Like many other products in this space, they seemed to be using the term to describe a server-side stack of storage value-add services customarily found on an enterprise array controller. Nothing more. I had hoped for a more robust model, especially since IBM has some storage virtualization technology in SAN Volume Controller that might help to differentiate IBM’s SDS stack from most of its competitors. VMware excluded storage virtualization from its definition of SDS and the whole industry seems to have been unwilling to deviate from that definition. However, being able to carve volumes of storage from a virtual pool — which you can do with a product like SVC from IBM or SANsymphony-V from DataCore — would seem to me to be very much in line with the true spirit of SDS.
Clod Barera seemed to think so, too, when I interviewed that IBM Distinguished Engineer at an IBM event a couple of years ago. (The video interview is here.) Clod was speculating at the time that XIV software might make a really good complement to SVC, turning the whole thing into what we now call an HCI appliance. But, this time around, the Spectrum Accelerate folks didn’t want to go there.
Also missing was any discussion of the possibility of adding file and/or object system elements to the IBM Spectrum Accelerate stack, which seems to be a potential evolutionary path of SDS going forward, or anything like DataCore’s I/O handling parallelization, which seems to make a lot more sense in speeding up RAW I/O than does throwing a lot of flash at the problem.
It will be very interesting to see the Spectrum Accelerate story unfold over time. Given the great patent work already done on XIV, and IBM’s extensive IP in storage generally, there is a lot of technology innovation they could throw at SDS without a lot of extra work.
For the record, I was IBM’s guest at IBM Interconnect 2016, where I received payment to live tweet and blog at the show. These video interviews are, however, my own work. They have not been edited in any way by Big Blue.