In his keynote address at VMworld, Pat Gelsinger, VMware CEO, said that virtualization “is partly to blame, as we hide the I/O patterns from the underlying storage … changing the characteristics of the storage environment itself.”
While the statement is correct and it’s great to see VMware acknowledge this–it’s hardly fair to blame VMware. The real problem lies in the disjoint between two fundamentally different architectures.
VMware has created a shared-services architecture that virtualizes the underlying physical assets. Although architected almost 25 years ago, storage still relies on this original design, which was created to be a subsystem to a single physical server.
While VMware has moved the world forward to create the ultimate in flexibility and asset utilization, storage has not moved an inch forward in the modern data center. It has simply grown larger and more complex (some might say “fatter”), while continuing to be based around the same architecture.
In a separate session, Chad Sakac, SVP of Global Systems Engineering for EMC, asked whether anyone in the audience had noticed the burgeoning number of storage startups out there. Then he answered his own question by saying they’re “multiplying like rabbits.” He then went on to state that there are three (and only three) storage architectures, and that every product sold by these startups utilized one of them.
And this–in a nutshell–is the real problem. VMware is aggressively moving infrastructure to a software-defined data center. EMC, claiming to be software-defined storage, clings to the past, pulling out dated and failed products to be repackaged (ViPR–subject of another blog). There’s a conspicuous dismissal of the vast amount of investment in storage startups who are addressing the architectural mismatch between storage and virtualization.
The bottom line is: storage in the virtualized data center is broken.
On the one hand, you have the CEO of VMware (who, by the way, is a former EMC executive who understands storage well) acknowledging this fundamental problem now known as the “I/O Blender,” and EMC, on the other hand, ignoring the wave of storage startups who are redesigning storage from the ground up to fit virtualized data centers. EMC clings to its outdated view of storage architectures and conveniently dumps each of these storage startups into one of three buckets, while chanting that there’s nothing new out there.
Although EMC continues to ignore the problem (or, perhaps, is incapable of fixing it, considering where the company started 25+ years ago), I can speak on behalf of the many storage startups and say that this is not the latest whim of investors throwing money at a trend, but, instead, a change that is long overdue.
Virtualization is driving this change, and virtualization is an unstoppable force that will continue its way through the entire infrastructure stack. While you can simplify things and say that the virtual machine is still just a server, or that a virtual desktop is still just a desktop, try asking the server and desktop vendors if anything has changed. It’s easy to simplify things into neat little PPT slides that make everything look the same… but ask Dell and HP how their desktop and server businesses are doing now.
This is a disruptive change. Big Storage (traditional storage vendors like EMC) are suffering from the innovator’s dilemma. Storage will be virtualized. It will be defined in software running on x86 hardware–changing not only the architecture of storage, but also its economics, its distribution model, its capabilities, and its manageability in ways that the “Big Storage” vendors cannot even imagine.
Kelly Murphy,
Co-Founder and Chief Strategy Officer, Gridstore
The post From VMworld 2013: VMware CEO Acknowledges the I/O Blender, While EMC Clings to the Past appeared first on Gridstore.
