There's been a rash of coverage in the trade press recently about the role that lack of standards plays as an inhibitor to the blades market.

On the face of it, standards for blades are attractive, offering the ability for third parties to build a single version of things like switches, and offering the ability to source blades from multiple vendors, with the implication that you can select from the best new implementations while maintaining much more independence from vendor lock-in. And on this last point, some of the commentators are correct - commitment to a specific blade architecture implies a higher degree of commitment to a vendor than does a rack-mount server purchase, because of the fact that the infrastructure is unique. But then again, it's not hugely different than committing to a specific brand of rack mount server. The main difference is that the commitment is now modulo 8, 16 or 32, rather than one (and yes, there is some initial startup investment in training, etc).

But before we run away in the direction of standards, we need to answer a fundamental question - do we loose more than we gain by designing and building standard blade servers, and how different is the real data center environment today when considering vendor lock-in in an environment of standard servers?

Standards versus innovation, the balancing act

The vision of a standard blade is compelling, but there are real tradeoffs. The first, and most important over the long term, is innovation. A standard blade implies more than just a standard form factor and connector. It assumes similar behavior, which in turn implies similar resources, or at least a consistent logical presentation of resources (think of a HAL on steroids), which will be very difficult to implement if the underlying system architecture changes radically. And change is certain. Blade servers have gone through multiple generations of fundamental architectural change since their commercial introduction in approximately 2001, and can be expected to go through additional change over the next decade as architectures evolve. The evolution will involve fundamental refactoring f the architectures, particularly the I/O and networking infrastructure and power management. Industry analysts have pointed out future innovations in virtualized I/O architectures, for example, that will require some fundamental re-engineering of future systems. To attempt to force future architects to accommodate a standard blade when considering these fundamental changes is a major inhibitor of future customer value over the long term.

The issue of standardizing blades and their behavior goes deeper, since, as mentioned above, there will be some expectation of consistent behavior. This will in turn begin to infiltrate management architecture, and will potentially ripple through the entire software stack, forcing all vendors into a least common denominator management stack. Vendor management stacks are differentiated for a reason - both their designers and their customer believe they add value, and the future evolution of dynamic data centers with virtualized environments demands that users and vendors have room to innovate. The path to the next level of data center architecture will be a messy decade-long process, almost certainly marked by some spectacular successes as well as some embarrassing failures, but the handicap it now by attempting to freeze one of the most important centers of architectural innovation is premature and short-sighted.

Standards and Vendor Lock-in

The notion of a truly standardized blade that users can buy from multiple sources and plug into their enclosures from any vendor has a lot of attraction at first blush. But wait! We have been buying little rack mount servers for more than a decade, and they have about the most standardized interfaces possible - a few Ethernet ports, some SCSI connections, a power plug, and a rack format that has been with us since before World War II. Why don't users mix and match servers freely in their environments, and why haven't we been able to buy standard boards that can plug into our Dell, IBM and HP rack servers, which after all, should have been easier to standardize than the more complex blade system architectures? The truth is that even in something as supposedly standardized as a rack mount server, vendors continually innovate, and that there is considerable value in the entire management environment, as well as in the totality of the vendor relationship. These last two points are particularly critical when considering blade servers. With management costs increasingly dominating the operations picture, the value of a consistent management environment and a smooth vendor relationship probably eclipses any small theoretical advantage in cost that might be obtained from multiple sources of server boards.

Furthermore, the notion that independent standard vendors will lower prices significantly has potential flaws. The market for blade servers is already extremely competitive, and with Dell now entering with a competent product, will become more so. And our customers are not stupid - they understand that major new projects represent important junctures for vendor competition, and behave accordingly. Another hole in the competitive price scenario is the pattern of volume in the industry. Right now, HP and IBM account for over 80% of the market volume, and have developed accordingly efficient supply chains to support these volumes. It's hard to make a case that an ecosystem of multiple smaller sources could do much better.

Lets Go to Fry's and Buy a Blade

But suppose, for a moment, that everything I say is so totally wrong (my kids think that all the time) that within six months we have a standard for blades, and you are operating a data center with a nice set of enclosures, hopefully from HP, designed for these standard blades. Your sales rep convinced you that your initial installation should at least have the vendor's own branded blades. Everything works fine. Now its time to add some capacity and your boss gives you a shopping list and you head down to Fry's. In the blade aisle you have the branded blades from HP, and some really great bargains from the same people who sell you modems, ATX motherboards and chassis. Some of them are cheaper than others. But they are all standards-compliant, meeting the MBDB (Minimal Brain Dead Blade) industry standard. But hey, the standard never said anything about build quality, signal integrity above a certain minimum, and a host of other subtleties that go into making a true enterprise quality blade.

Take a couple back to your data center, along with maybe a couple of power supplies and a new switch, plug them in, and see what happens. Seems pretty OK to me. And when it fails, I'm sure all the multiple vendors as well as your own internal engineering groups and users will be quite understanding and more than willing to work together to solve your problem.

OK, so I'm engaging in a little reduction-ad-absurdum argumentation, but the germ of truth is there - third party blade suppliers will need to be qualified even more rigorously than the primary vendor's equipment, and support arrangements will need to be modified. Most organizations today do not want to cope with the support of mixed vendor environments except across fairly coarse project, application or infrastructure function boundaries. To do so within a single enclosure is an almost insurmountable burden. It is also an activity that adds nothing in fundamental business value while consuming resources that almost certainly would be better employed elsewhere.

Absent Standards, Where Do We Go From Here?

It's a very Dickensonian world we live in, as in "It was the best of times. It was the worst of times." Vendors will continue to innovate and compete. Successive waves of products will get better and will not be standardized for the foreseeable future. Users will continue to be forced to make major commitments to complex infrastructures, and will have to spend the time and resources to understand the new technologies. But the solutions they get will deliver real business value, and competition will keep vendors as honest and efficient as they can be. Not a bad situation for consumers of technology.

So in the final analysis, will standards really limit the growth of blades? All of the industry observers agree that blades will remain the fastest growing segment of the server market, and at HP, we are still seeing blade growth well in excess of the industry forecasts. Clearly something is working here, despite the lack of standards.