As I joined the Tokyo-subway crush of identically backpacked conference attendees at VMworld 2011 last week, I calculated that this year's population -- in excess of 19,000 -- represented a fivefold increase over the first VMware event in 2005. The crowds clearly indicated that VMware has become the center of a thriving partner ecosystem from which the next-generation data center is emerging.
The list of partners is a who's-who leading players -- with a heavy focus on storage tech. Four of the five "diamond" partners are among the largest storage vendors (Dell, EMC, HP, and NetApp) while the fifth, Cisco, is far and away the biggest converged data center player.
As you might expect, VMware itself promoted a raft of new product revisions, including vSphere 5.0, View 5.0, Site Recovery Manager 5.0, and vCloud Director 1.5. Taken together, the enhancements in this spate of product releases is nothing short of amazing -- with substantial improvements across the board, especially in storage. (I'll dive into some of these features over the next few weeks, so stay tuned.)
Grumbling in the VMware ranks
Everyone understands that VMware's technology leadership comes at a premium price, but how long the company can resist competitive pricing pressures is an open question. The much-publicized customer push-back over the new vRAM-based licensing scheme used for vSphere 5.0 is an early indicator of trouble on the horizon. It was top of mind for many of the attendees I spoke with.
I don't think that either of these partners will go out of business as a result -- far from it. VMware's releases are new, more expensive, and not yet as well-polished as those of its third-party counterparts. But the fact that VMware appears comfortable stealing their lunch money must give them pause.
VMware and the wide-area data center
As usual, a loud buzz persisted about forthcoming VMware products and features. The most interesting was an increasing focus on geographically stretched cluster deployments -- including the new VXLAN standard and an emphasis on application and storage interdependencies across VMware's spectrum of products (particularly VMware's redesigned HA and SRM).
If you're looking for the next big thing to come out of VMware and its core storage and networking partners over the next few years, the stretched cluster is where it's at. Currently, organizations seeking to provide both disaster avoidance and disaster recovery need only look as far as the vMotion and High Availability features baked into vSphere. However, if you want to offer those same feature sets across multiple sites, things get significantly more complicated and risk collapsing under their own weight.
vSphere 5.0 extends the WAN latency requirements for intersite vMotion from 5 milliseconds to 10 milliseconds (a huge jump). As a result, metro-vMotion becomes a reality in many use cases in which it wasn't previously, and it eases site-level disaster avoidance -- think nondisruptively shifting compute loads from one data center to another to avoid the effects of hurricane Irene. Likewise, Site Recovery Manager has made site-level disaster recovery, such as recovering from an earthquake, relatively simple to implement.
But implementing site-level disaster recovery and disaster avoidance at the same time is extremely complicated, generally involving a great deal of storage tweaking, networking voodoo, and HA scripting that most organizations would find far too involved to maintain. The challenges in this space involve much more than allowing vMotion over distance or fielding active-active geoclustered storage. Chief among the issues are the 30-something isolation/partitioning scenarios that can variably prevent those two data centers from communicating -- either at a network or storage level -- and bring down the whole house of cards.