In my discussions with customers, I often ask the "V" question: how are you doing with virtualization?
Occasionally I get an interesting response: they're an entrenched IBM customer, and they point to the use of virtualization on their mainframe, and perhaps their big AIX boxen, and say they're largely virtualized. Asked and answered.
I need to ask them a better question: how are you doing with pooling of resources?
The Bigger The Better
Any time you pool resources, you're angling for a better outcome. Lower cost-to-serve through scale efficiencies. The ability to load-level across multiple, shifting demands. Being able to react quickly to new and unforseen demands. Efficient processes that manage resources in the aggregate, rather than individually.
Of course, alll those thoughts apply to the IT domain, but there's more to be considered. Moving workloads (and their data) where they need to run for cost, compliance or performance reasons. Implementing consistent controls and processes that secure and protect information resources.
Need evidence? Consider other forms of infrastructure: like power generation, global networking -- even how airlines operate. Pooled and aggregated infrastructure models are the norm. And IT infrastructure isn't all that different.
One of the more powerful thoughts behind cloud (and virtualization) is the construction of progressively larger and larger pools of resources. Generally speaking, the larger the pool, the better: greater efficiencies, lower cost-to-serve, etc.
The ideas behind pooling network resources have been around for a while. IT practitioners have long realized that aggregating all network demand (voice, data, video, etc.) makes far more sense than constructing individual and isolated networks around each and every use case.
In the storage world, people have been continually interested in constructing progressively larger and larger dynamic storage pools for many of the same reasons. If they can span distance, so much the better.
Hence all the interest in EMC's FAST, VMAX, Atmos and newer VPLEX Geo capabilities.
In the compute world, VMware has sparked widespread interest in using virtualization to pool compute (along with network and storage) to create progressively larger and larger (and hence more efficient and responsive) pools of infrastructure resources.
And if your "owned" pool isn't big enough, there are plenty of service providers arriving on the scene who'll grant you access to their pool :-)
Building The Bigger (and Better) Pool
Lots of small pools aren't ever going to be as effective as a few bigger ones. And that's what I think I'm finding when I speak to these IBM-centric customers: perhaps they're confusing the underlying enabling technology (e.g. virtualization) with the ultimate goal (constructing progressively larger and more capable pools of resources).
A single AIX server (or mainframe) does create a certain sized pool: although it's typically constrained to the boundaries of a single server and a single location. Better if one could pool *all* the AIX resources (or mainframe resources) with the same fluidity as being achieved in a VCE stack (think Vblocks).
But even if this comes to pass -- let's face it -- the candidate workloads for this sort of one-off pool is rather constrained as compared to other more mainstream alternatives.
For example, most flavors of virtualized desktop computing wouldn't be a candidate (that's usually many thousands of potential compute images), nor would the vast Microsoft landscape (Exchange, Sharepoint, SQLserver or anything else) be a candidate workload for the pool. Arguably, many forms of Linux could run on an imaginary IBM-powered virtualized pool, but there's also some work associated with that proposition.
I also could construct an argument that -- due to "Intel economics" -- adherents to this particular world view will find themselves with a progressively smaller technology and vendor ecosystem to work with (Lotus Notes, anyone?) as well as not being able to benefit from the rapid decline in commodity technology prices.
Indeed, the technology envelope around Intel-based virtualization (think VMware and everything else) are fast growing capabilities that can't be found anywhere else than -- perhaps -- modern mainframes. And not everyone can afford to run everything on a mainframe :-)
Finally, there's not likely to be a rich marketplace of external service providers who are willing to provide "as a service" offerings that are compatible with the internal pooled environment. Other than perhaps IBM themselves :-)
The Cup Is Half Full, Though
I'm unlike to win many adherents to this particular strategic view from the customers who are heavily invested in IBM server technologies. The legacy is just too daunting for many to consider. Progress will be likely slow and incremental, rather than the more accelerated adoption and associated transformation we're seeing from customers with different landscapes.
Their adoption of Intel (and VMware) powered internal and external clouds will likely be driven by meeting the needs that the IBM approach can't deliver on. Things like virtualizing thousands of desktops. Or virtualizing the Microsoft landscape. Or a few other interesting scenarios :-)
And, from a purely selfish EMC perspective, we're more than happy to create progressively larger "storage pools" that are dynamic, virtualized and geographically distributed that support all manner of IBM servers: AIX, iSeries, zOS -- the works!
So, how are you doing on the pooling thing?
Comments