The Itanium Solutions Blog

A recent press release sent some shockwaves through the IT user community. Now, thanks to the collaboration and innovation of mission-critical computing giants NEC, Intel and Microsoft, the first servers with on-the-fly, dynamic partitioning capability was introduced: the NEC Express5800/1320Xf.

Before we get into how it works, let’s be clear on what this means to you. With the NEC Express5800/1320Xf server you can hot add and hot replace processors and memory in your server on demand - without having to stop any of your applications. So if you need more processing capacity on short notice, or if you need to quickly replace failed modules, you can do this without impacting your user community.

It kicks the whole idea of “high availability” up to an entirely new level.

For Microsoft, the recent releases of Windows Server 2008 and SQL Server 2008 were critical to making this happen. The new edition of Windows Server, a major upgrade from the 2003 edition, now supports hot adding of processors, memory, and I/O host bridges, plus hot replacement of processors and memory. SQL Server 2008, with its new “Resource Governor” feature, allows real-time management and allocation of “application resource pools” to keep your database applications running at optimum performance, regardless of peak system demands.

This only works on systems using x64 and Itanium processors, but for good reason.

Intel recently introduced a number of new high-availability features in their Itanium 9100 series processors, including the new Demand-Based Switching feature, which allows dynamic allocation of processor power as needed. Add to this to the 9100’s new Core Level Lock-Step feature which allows the dual-core processors to automatically mirror the operations of one core by the other core, plus the existing Socket-Level Lock-Step, Enhanced Machine Check, Intel Cache Safe, and Advanced Error Detection and Correction technology, and you have a very powerful combination of improved features for NEC and Microsoft to use to full advantage in this new release.

We capitalize on all this using our own unique, cellular-based hardware architecture, allowing for dynamic logical reconfiguration of the Itanium processor and memory partitions, a proprietary approach to error correction and error management, very large cache architectures, and a critical “dedicated cache coherency interface,” as they describe the linkage between data caches. Our new server also supports partitioning approaches that allow multiple Operating Systems to work independently of each other simultaneously. It also uses cross-bar architecture to allow real-time rapid reconfiguration of hardware cells and I/O bridges on demand.

As all of us know, IT organizations have been working hard to consolidate servers into fewer, more powerful systems. The upside of this has been a dramatic increase in cost savings for the enterprise. The challenge is trying to achieve this with the expected high level of availability that today’s computing business demand. So when a product can dynamically reallocate system usage to support unexpected rapid surges in demand — or in response to localized system failures - it’s going to generate quite a bit of attention. Microsoft describes their recent innovations as “bringing the mainframe to the mainstream.”

For more on this, check out this paper which takes an in-depth look at the Dynamic Hardware Partitioning issue.

Posted in Computationally Intensive Applications on April 23rd, 2008 by Mike Mitsch |