VMware is jockeying to become the controller of the “Software-Designed Datacenter” of the future. But, just as the Wizard would be nothing without the consent of Oz, “Software-Defined anything” needs more than a controller. VMware must entice, cajole, and coerce an entire ecosystem of server and storage providers to create and implement remote control mechanisms. And the example of Software-Defined Networking (SDN) isn’t entirely hopeful.
Disinclined to Acquiesce to Your Request
“Software-defined” implies a separation from decision-making (software) from action (usually hardware). Whether we are talking networking, storage, or the datacenter as a whole, the road to software-defined-ness requires both a controller and support from those who would be controlled.
But consider a typical VMware infrastructure: Why would HP invest in a protocol giving an outsider total control over their servers? Why would NetApp relinquish their seat at the table? What about Cisco?
Abstraction inevitably erodes the power of individuals, and “Software-Defined anything” threatens the vendors involved. Enterprise storage vendors are willing to play along with VMware-centric moves like DRS and VAAI because they believe it can drive differentiation and sales. Demand from “hypervisor hugger” customers merely supports this decision. No sales, no investment.
HP, IBM, and Dell are practically tripping over themselves to prove that theirs are the ideal products with which to build virtual infrastructure. Although they all look forward to “stack” sales and services revenue, server vendors would certainly love to go back to the time when HP-UX, AIX, and Solaris locked buyers into the HP, IBM, or Sun ecosystem. But those days are gone, Dell is lunging through the newly-open datacenter door, and VMware is everyone’s ticket to the future. Only Oracle, with their alternative datacenter vision, is tepid on VMware.
VMware’s success puts the whole IT industry in a terrible position. Even if daddy EMC (and his questionable BFF, Cisco) is overlooked, no IT company wants to surrender control and become a maker of invisible and interchangeable components. The road of commoditization leads to China, not Austin or Armonk.
Beware! This VMware Software-Defined Datacenter video contains Buzzword Bingo!
Kicking and Screaming
In the networking space, the general concept of SDN has enthusiastically been embraced, even as specific protocols like OpenFlow are dismissed as being of limited utility. After all, if Ethernet switches were simple forwarders, why not buy HP or Huawei?
Cisco wisely adopted an “ours is even better” response to OpenFlow. They promise OpenFlow support, sure, but Cisco is really pushing a whole range of control APIs under their SDN banner. Future Cisco-centric solutions can be entirely software-defined, but the definitions will be Cisco-centric. This allows the company to claim their own custom-fit SDN crown without losing control.
One expects the same sort of “frenemy” response to VMware’s Software-Defined Datacenter vision. Server and storage vendors may indeed give lip service to accepting VMware control, but they will never give up their differentiation. Instead, they will embrace and extend VMware’s vision and protocols, retaining the best capability for their “stack” customers.
What Will Users Do?
This is not a positive vision for the future. VMware will continue to drive “proprietary standards” for datacenter control, positioning vSphere as the center of the universe.
But big-name hardware vendors will drag their feet, rushing out bare-minimum implementations. They will resist handing too much control to VMware, yet will continue marketing their current love-fest approach to the folks from Palo Alto. Prepare for cognitive dissonance as actions fail to match words!
But many other companies will embrace VMware’s whole vision. EMC (and perhaps Cisco, if their marriage lasts that long) will of course be first in line. Next come the Asian challengers, who will see VMware’s vision as a way to drive greater sales and margins. Startups, too, will work to implement SDD, since it gives them a wedge into the enterprise market.
This leaves users in a difficult position. They can prepare to do battle with their chosen server, network, and storage vendors, pushing them to completely integrate with VMware. They can switch to a new supportive vendor (EMC would be happy to cash those checks!) Or they can consider an alternative vision, with less reliance on VMware for core datacenter automation.
The Software-Defined Datacenter is a great concept, but it just won’t work. The big enterprise companies will never allow VMware (and daddy EMC) to commoditize them out of existence, so useful implementations will be rarer than ruby slippers. The best we can hope for is point enhancements to enable greater virtual machine mobility through SDN and improved storage integration.