If blade servers are to become the revolutionary IT tool they are touted to be, system suppliers must deliver a new generation of management tools, and software companies need to rework their licensing models.
The management tools needed for the successful implementation of blade servers are evolving rapidly. Established players in the server management market such as Hewlett-Packard and IBM have a large head start over their competition when it comes to managing individual boxes - their many years' experience designing tools that are closely integrated with their hardware offerings being their greatest advantage.
These two companies also have a strong background in multi-vendor management environments, which may give them a leg up when managing the blade ecosystem of hardware plus software. But server-based agents monitoring internal temperatures and other minutiae only address a small portion of blade server management needs.
Software that handles the grunt work of provisioning - the loading and configuring of applications and operating systems - will make or break blade servers for the enterprise. With the right provisioning software, a blade chassis can function as a well-orchestrated whole.
This involves a lot more than cloning a disc image and blowing it out to server blades when needed; a cloned image of an OS and one or more applications still needs to be tuned for the specific operating conditions it will face. Those who treat provisioning tools as a separate category of blade management software are missing the boat.
Provisioning rightfully encompasses the entire lifecycle of the blade and the apps it runs. Imaging and server management are not peers of provisioning; they are vital pillars of a well-managed and properly provisioned blade system.
Early adopters are resorting to a mixture of home-grown tools and the first generations of blade provisioning software - but there is still a long way to go, even if the roadmaps become clearer. Sun's vision, as expressed in its N1 family of provisioning and management products, may be the best considered, but it will have to grow beyond simply managing applications running Solaris on Sparc or x86 blades if it is to reach its full potential.
Perhaps the best model for full lifecycle provisioning may be that of Opsware (formerly Loudcloud), which supports a half-dozen operating systems (AIX, HP-UX, Linux, Solaris, and Windows NT/2000); a variety of web servers; and a plethora of application, database, integration, and other services.
Software licensing schemes must completely transform themselves if blade servers are going to be as flexible as their proponents say they are. Although some enterprise application companies have begun to move away from client- or seat-based pricing, the processor-based licence is only the first step. Ultimately, licensing metrics for blade software have to be based on the number of processors in use per unit of time.
This licensing aspect is important because one of the salient features of blade technology is flexibility. The idea that the blades in a chassis can be readily repurposed from running a database application to running an ERP system to hosting a web server is at the heart of blade computing. So long as software licences are written to charge the customer based on the maximum number of CPUs that will ever be used for the application, there's no incentive to reconfigure a blade once it is running a particular application.
Software vendors that fail to step up to the plate on blade-server licensing issues can expect to be prodded by their customers, who have every reason to demand more flexible terms. Ultimately, any vendors refusing to adapt to the new reality of licensing should expect to see their customers deserting in droves to vendors who "get" blade computing.
Blade servers have the potential to be the most important thing to happen to the server room since the KVM switch. But without provisioning tools that give system administrators the ability to repurpose servers based on the workload of a given day, or a given hour - and more importantly, without the software licences that permit this kind of flexible model - customer savings will be limited at best.
Meanwhile, a server blade standard could spur innovation market, which could, inevitably, lead to the creation of a hypothetical Server Blade Industry Association (SBIA), which would not only promote the blade concept, but also would certify hardware and software as conforming to the specification. Through logo programs and public plug-fest shootouts at Networld + Interop, this SBIA could do for blade computing what the Wi-Fi Alliance has done for popularising wireless networking.
Yet as tempting as a blade standard or even a blade association would be, this is still the time of innovation. Vendors must be free to experiment, without limitations. In a year or two, when patterns of blade-server usage begin to emerge, and when early adopters begin to hit the wall on performance, scalability, and expandability, only then will it be time to begin the debate about industry standardisation.
PJ Connolly and Alan Zeichick write for InfoWorld