Why Nehalem-EX matters
The x86 architecture has been "growing up" for many years. IT has been giving it more and more responsibility as it does. The just-launched "Nehalem-EX" products essentially graduates Intel's Xeon server line into Big Iron adulthood.
Every new IT product, it seems, is "revolutionary." A "breakthrough." "Transforms the economics of computing." Yeah, right.
Very few products really are, or do, these things. There are, however, a handful of products each year that are excellent milestones to the progress IT is making, and that are themselves impressive and important accomplishments that move the industry forward. Intel'sis one of those. Why?
Completing the set
All together, Intel's "Nehalem" processor generation is pretty darn impressive. It's distinguished by its aggressive manufacturing density (45nm and 32nm), large core counts (up to eight cores per socket), simultaneous multi-threading (aka "Hyper-Threading"), large memory capacities, high I/O bandwidth, and built-in support for power optimization, virtualization, and 64-bit computing. (That list doesn't even come close to being exhaustive.)
Nehalem-EX (officially, the Intel Xeon 7500) completes the set with a high-end, server version. It ups the feeds and speeds, of course. More cores! More sockets! More memory! More widgets! Most important, it makes it straightforward for server vendors to create extremely powerful x86 servers. 4-socket/32-core servers are easy peasy. 8-socket/64-core servers-enormous by any historical standard-are not much harder. Vendors like IBM and NEC that have majored on scale-up will go even higher. At the same time, Nehalem-EX adds the I/O bandwidth and reliability/availability features needed to feed and care for such large resource pools. Sure, there may be more variants of Nehalem to come (preview: Even more cores! Even higher frequencies! Even more bandwidth!), but the entire range is now in play.
Scale up, meet scale out
In the old days, the servers that scaled out for distributed and networked applications were not the same ones that scaled up to handle large enterprise apps and databases. They needed entirely different designs, components, software environments, interconnects, operational skills--you name it, it was different. Not so much any more. Nehalem-EX lets server designers scale from a few processor cores to scores upon scores of cores. Sure, OEMs will design mission-critical systems somewhat differently from the "average" server. Users will still configure and operate somewhat differently. But many of the components, skills, and software choices can now be the same. When it come to optimizing the economics of computing, that's a very big deal.
We are deeply within the virtualization age. For a growing number of enterprises, "100 percent virtualized" is the goal and target. Just about everything Nehalem does--all those cores, all those GB of memory, all the built-in virtualization acceleration--optimizes for a virtualized world. Sure, you can run those resources physical, if there's a specific reason to (for example, in high performance technical computing). But Nehalem EX's advances at the processor and server level, coupled with the latest virtualization environments, mean there's little need to run physical for even demanding workloads. This powerfully supports IT's drive toward ubiquitous virtualization, and the flexibility, availability, and economic benefits that entails. Much datacenter rationalization, consolidation, virtualization, and optimization will result.
Active, real time, front and center
Scale-out computing grew up running relatively lightweight workloads in distributed enterprises, serving Web pages to the Internet, and filling out HPC clusters. But its tasks have evolved. Enterprises, for example, now entrust substantial workloads--including core database, ERP, CRM, supply chain, order taking, and revenue collection functions--to x86 infrastructures. The Web is no longer static. Web pages are dynamically customized and personalized, the results of dozens of active components run in real time as the user browses. Web applications now rival desktop applications in functionality and interactivity; that demands constant back-end processing. The HPC style of analytics jobs that were once the province of secondary engineering teams are now the core of product design across a range of industries. In many cases, business analytics run 24x7, driving real-time business decisions. In short, there's vastly more workload to be run, and it's vastly more important to businesses. Much of this active, real-time, front-and-center workload runs on x86. Industry trends and the virtues of the Nehalem generation mean that even more will run there.
Seldom can you point at one product as the unique moment when large-scale forces and trends transpired. But we can point at milestones along the way. Nehalem-EX is a natural point at which to look at Intel's Xeon line and say "it's grown up now!"