VMware is in an enviable but tricky situation: The company must work closely with hardware partners, keeping these prime sales and promotional channels happy and supportive. But VMware must also innovate around proprietary OEMs, subverting their products with integrated software before a rival steps up with an integrated alternative. This two-prong “coopetition” strategy has worked marvelously so far, but is it sustainable?
Looking Backward
The server virtualization strategy is all about continuity of legacy solutions. VMware’s core market is the evolution of existing systems to more-flexible infrastructure, and the company has been remarkably successful there. Virtualization allows critical open systems applications like email, databases, and CRM to escape from the bonds of hardware, improving responsiveness and availability and making disaster recovery possible.
VMware is the heart of today’s virtual datacenter just as Microsoft, Sun, IBM, or HP dominated in the past. Any hardware or software vendor wanting to “play” in this market must be VMware compatible above all else. Without a precious HCL listing, customer acceptance is severely restricted. This is how much customers rely on VMware: They would rather switch hardware or software providers than venture outside vSphere!
As kingmaker of the datacenter, VMware has been remarkably well-behaved. The company has aggressively implemented features that enhance the value of OEM products. VMware has worked to open about roadmaps, to share access to technology, and to support this broad OEM market, although not always successfully.
Consider the case of enterprise storage. Many of VMware’s best features (HA, vMotion, DRS) rely on shared storage. Yet the “I/O Blender” inherent in the hypervisor’s presentation of storage interferes with the ability of enterprise storage array performance and features. So VMware spent years adding and enhancing VAAI to enable these products to function properly in virtual environments. With vSphere 5.1, third-party FC, iSCSI, and NFS arrays are marvelously efficient and effective thanks to VMware’s integration efforts.
The I/O Blender: How Virtualization Challenges Storage
Interested in the I/O Blender? Here’s my presentation on the topic from the 2013 Melbourne VMUG User Conference!
It’s a good thing, too! Imagine the upheaval that would result if VMware shut out important datacenter companies like HP, IBM, Dell, NetApp, HDS, Microsoft, Symantec, and the like. Sure, EMC or VMware would likely surge in the short term, but it would undermine the whole vSphere datacenter vision. Before long, existing competitors (Microsoft, Citrix, Red Hat, etc) would sideline VMware with the help of these disgruntled monsters.
Disruptive Innovation
But VMware faces another existential threat in the form of converged infrastructure and cloud offerings. Nimble startups and innovative challengers offer an alternative vision: Rather than “fat PCs and storage arrays”, they suggest a datacenter full of cheap commodity hardware running scale-out software. Whether a converged virtual infrastructure solution (think Nutanix, Scale Computing, or Amazon EC2) or a cloud platform (Microsoft Azure or Google App Engine), these alternatives threaten the entire ecosystem.
If VMware doubled down on traditional datacenter hardware, their dominant position would certainly help slow the charge of software-driven alternatives. But the cost savings and productivity improvements inherent in converged infrastructure and cloud platforms are hard to ignore. Eventually, vSphere+hardware would look as stale as the monolithic data centers of the 1990’s and the market would move on.
Instead, VMware has wisely worked to join this trend. Their cloud platform and application moves haven’t worked out so far, but moves into software-driven converged infrastructure look quite promising. Many of VMware’s moves in this area have largely gone unnoticed: Since ESX 3, they have integrated many features (HA, cloning, replication) that traditionally required specialized hardware or software. In most cases, VMware’s offerings are limited in scope or functionality, leaving room for OEMs, but these restrictions are being stripped away.
Other VMware products are likely more galling to traditional partners. The vSphere Storage Appliance (VSA) in vSphere 5 meant shared SAN or NAS storage was no longer strictly required for smaller environments. But the vision of VMware Distributed Storage (VDS) laid out at VMworld 2012 is downright shocking: It’s a real full-featured storage array killer for the large shops that today shell out millions to EMC, NetApp, HDS, and the rest.
Two Strategies at Once
VMware is playing both angles at once. Their close partnership with hardware OEMs provides a great sales channel with plenty of customer touch points and keeps these legacy providers in check. But at the same time, disruptive technologies like VDS and Nicira promise to move the vSphere-centric datacenter out of this world of lumbering, expensive dinosaurs. This two-prong strategy is the best choice VMware could make, though it’s certainly making as many enemies as friends!
The Nicira acquisition fabulously illustrates this dual approach. On the one hand, Nicira allows VMware to continue to work with “close friends” like Cisco, Juniper, and Brocade. After all, Nicira uses standard protocols and interoperates with big iron networking equipment. But the Nicira technology could also lead to further commoditization of networking hardware, with companies like Dell and HP (not to mention Broadcom) winning big as the old dinosaurs stumble. VMware wins either way (or both!)
This last point is critical to understand: The datacenter market doesn’t need a single winner. Traditional big iron can continue even as converged infrastructure and cloud platforms take off. VMware will continue both to integrate with, and innovate around, the big players. And they’ve got a stake in the cloud, too! The datacenter of tomorrow will likely include all three architectures, along with monolithic UNIX and mainframes, too!
Stephen’s Stance
It can be confusing to watch VMware roll out integration with OEMs and undermine them in the next breath, but this strategy makes perfect sense for everyone. Although the big guys would love their current products to remain dominant, they understand that times and technologies are changing and VMware has to change as well. The smart ones will adopt a similar enhance-and-undermine strategy, though many likely will not make this transition. It’s just the nature of the business.
“The datacenter of tomorrow will likely include all three architectures, along with monolithic UNIX and mainframes, too”. I agree. The part that I’m not sure is the datacenter of the day after tomorrow 🙂