The current situation in the business technology realm is something I’ve started calling the “infrastruggle.” Simply put, challenging economic conditions have put the screws to IT budgets just at a time when enterprise executives are being required to deliver improved service levels, reduce downtime (even for maintenance), ensure data is managed per legal and regulatory requirements, and innovate in ways that will enable IT to turn on a dime in response to changing business priorities.
That’s not the infrastruggle. That’s only the backdrop.
The infrastruggle refers to the battle between technologies—in too many cases, only slideware—that’s being waged by vendors seeking to help consumers cope with the pressures they confront, as well as to gain access to their constrained budgets with products and services that lock the consumer into the vendor’s products while locking out competitors. Marketing machines are in overdrive, extolling the benefits of server virtualization and clouds. Not surprisingly, marketing claims don’t always align with technical reality.
Everyone has heard how server virtualization is the “natural evolution of computing” given the commoditization of hardware. Server hypervisors, we’re told, enable us to abstract software away from hardware so applications and workload can move effortlessly between different hardware kits. It’s the ultimate consolidation tool, saving energy, reducing server sprawl, and increasing labor efficiency by enabling more to be managed by fewer administrators.
Only there are a lot of gotchas when you peer under the covers. For one, x86 server hypervisors aren’t as robust as the brochure suggests. The basic hypervisor license, which often ships with the server for free, lacks the high-availability features users want, requiring the purchase of more expensive enhanced licenses. Another hard learned lesson is that all applications aren’t hypervisor-friendly. You may need to hire a consultant to help sort that out—for an additional fee. Then, comes the realization that administrators require specialized training, which means more cost. Finally, after several workloads have been propped up in a single physical box, the impact on your LANs and SANs begins to appear. Major changes are required to get anything close to acceptable I/O throughput—a huge undertaking and expense.
Add to the mix the hypervisor wars—efforts by “clubs” of hardware and hypervisor vendors to create approved and certified hardware/software stacks that exclude other hardware and software vendors—and you start to see the self-interest of vendors rearing its ugly and inevitable head. The leader in hypervisors has introduced proprietary and non-standard SCSI commands into its dialog with storage gear. Only certified kits that understand the commands will work with the hypervisor. The same vendor is about to add a “storage hypervisor” to its server hypervisor software that will further constrain whose gear will work with their stuff and force those with SANs to segregate the friendly arrays from those that aren’t supported.
Not surprisingly, the story has two potential conclusions. Some will stay the course, figuring that real money has been invested and careers are on the line. They will throw out more money, trying to make hypervisor computing work. Others, and according to recent surveys this group includes a significant proportion of prominent firms, will abandon server virtualization initiatives when they’re less than 20 percent complete. This second group accounts for why leading market watchers are saying that server hypervisor technology is only deployed on about 17 to 20 percent of the server install base, with few encouraging signs of accelerated adoption. That’s way off the market penetration targets we were hearing from the vendors in 2005.
Clouds are evolving in much the same manner, which is to be expected since so many clouds are building their offerings using specific server virtualization software stacks. All that external clouds do in this situation is further complicate matters by trying to deliver compute services across unpredictable and clunky WANs. Still, the hype is extreme. One recent analyst touted 356 percent growth in cloud adoption. Only after reading the fine print in the survey appendix do you realize this reflected nine additional votes for clouds over last year’s survey.
Isn’t it about time for a real discussion of the technical requirements for meeting business needs? Imagine what we could accomplish if we spent the money on infrastructure and data management instead of hypervisors and clouds.