Mainframes today elicit both passion and derision. Some people swear by them and others declare their demise. Before I weigh in on the future, I would first pose the question "What is a mainframe anyway?" When I looked up the "definitive characteristics" in wikipedia*, I was encouraged to see the characteristics I expected were called out:
"Nearly all mainframes have the ability to run (or host) multiple operating systems and thereby operate not as a single computer but as a number of virtual machines. In this role, a single mainframe can replace dozens or even hundreds of smaller servers, reducing management and administrative costs while providing greatly improved scalability and reliability."
My response to this description of the mainframe is best described as déjà vu. I could be reading any of a myriad of white papers on the next generation data center, including some of my own. Similarly, I could be reading the marketing copy from leading virtualization solutions in the industry. IT managers of the world are collectively, if not knowingly, evolving their data centers toward the next "mainframe" generation, only this time, the future mainframe is not an expensive and proprietary appliance. This new "mainframe" will be a collection of "pools" of virtualized resources, built on heterogeneous industry standard compute, storage and network resources that can be aggregated into "servers" of almost any size. Each "server" will have precisely the mix of resources needed for its workload. Servers will be "provisioned" as needed, ideally by policy without operator intervention, and when no longer required, resources are returned to the pools. Intel has described this future architecture as "Dynamic Resource Management". DRM is one of the key pillars that future data center products and technologies are measured against at Intel.
The benefits of DRM are huge, great utilization of resources and reduced cost of operation, but there are still many barriers. I don't believe the DRM vision is unique to Intel, but I do believe that Intel is uniquely positioned to enable this future vision in a way that avoids today's proprietary vertical solutions. IMHO Intel already has the leading architecture and technology roadmap for data center virtualization with Intel VT, VTD, & VTC. Realization of an open DRM will require more than technology leadership, it will require innovation, cooperation and standards for platform management and metering in the ecosystem. The innovation is happening, just last week one of the leading server OEMs announced an enhancement in virtualized storage that will improve their capability to pool storage resources. Intel has a history of driving open standards ( usb, pci, 10gbe, etc...). I expect Intel to be at the front of the industry players making an open "Dynamic Resource Management" architecture real.
Some say DRM is already here today, just pick your favorite virtualization or grid vendor. My position is that today's DRM solutions are still very vertical stacks they don't really have the openness we want. I also don't believe most people ready to bet their production systems on these new environments. I do look forward to the discussion.