IBM has announced that it is investing in development, marketing, and sales-channel programs to deliver industry-specific HPC cluster solutions at the departmental, divisional, and workgroup level in the mid-market. The initial vertical focus includes life sciences, Computer Aided Engineering (CAE), and finance.

The initiatives include pre-architected solutions based on Linux and Windows Compute Cluster Server 2003 with support from IBM; a partnership with Microsoft around key HPC solutions that features four new benchmarking and tuning centers; the HPC ValueNet for IBM Business Partners; and a new On-Demand Test-Drive Facility. Working with Accelrys, IBM is optimizing System x and BladeCenter configurations for Discovery Studio, a suite of solutions for protein modeling and computational chemistry for Linux- and Windows- based environments. IBM has optimized hardware configurations for CAE applications from ABAQUS, ANSYS, Inc., ESI Group, LSTC, and MSC.Software. In conjunction with Microsoft, IBM is offering a clustered solution featuring Office Excel 2007 that will allow clients to use parallel processing on clusters, thus freeing up client machines from long-running calculations. The HPC ValueNet for IBM Business Partners that offers joint marketing plans and incentives; sizing guides with predefined cluster configurations for standard and enhanced configurations; accelerated delivery and competitive pricing through IBM's Express Seller program; and field support from IBM's Cluster Enablement Team. The On-Demand Test-Drive Facility provides customers the opportunity to test-drive their HPC applications in IBM's Deep Computing Capacity on Demand centers with access to over 20,000 processors.

When most think about HPC or clusters, thoughts of massive computing solutions come to mind. Seismic research, astronomy, financial derivatives calculations, etc. are just some of the conventional environments where clusters of high-performance systems may be found. While it is certainly the case that massive HPC solutions can be found doing these very tasks, the reality is that clustered computing can be applicable to a smaller-than-galactic-scale undertaking, something that might be akin to the needs of a mid-sized enterprise. This scenario should sound familiar as it shares many similarities with consolidation initiatives that a few years ago were solely considered in the realm of large organizations. Whether it is applying multiple physical resources to one workload or applying a single physical resource to multiple workloads, we see that the suitability of the technology is ultimately dictated by business needs, not by any perceived aristocratic position of the technology. If server sprawl in the datacenter is the middle-of-road solution, then either roadside shoulder offers a more compelling value proposition.

With these initiatives we see Big Blue looking to expand the areas where clustering may be relevant for the less-than-Fortune-500 entity. By focusing on specific verticals that are HPC-cluster-friendly, IBM has the opportunity to stimulate demand through a pre-designed and tested approach that should translate into less deployment risk and reduced deployment time. Both of these are key considerations to channel partners and their customers. Another way to view this is that it is a more efficient way to deploy clusters, which historically have tended to be custom-derived or heavily integrated affairs. Scaling down the HPC cluster to meet the needs of the mid market may make available potential new customers that previously would have forgone a clustered approach due to the complexity, cost, and time to deployment. It might seem counter-intuitive to be scaling down that which was designed to scale up, but we need to remember that business needs, not the technology, is what drives deployments. Given the computational ability of systems in the part, massive scale was the only means to garner sufficient processing power, but with today's more-powerful-than-ever CPUs, scale no longer needs to be massive to meet the business imperative.

Another nod to business practicality is the emphasis on Microsoft Windows Server Clusters, and the HPC ValueNet. Mid market enterprises have considerable investments in Windows solutions, and tend to purchase their technology through business partners. We believe IBM's focus on building out the ecosystem for partner-delivered clustered solutions, especially those with Windows as well as Linux offerings, is a wise move. Armed with the expertise that will come from the benchmarking and tuning centers as well the Cluster Enablement Team, channel partners may find themselves more comfortable with some new approaches to meeting the needs of their customers while knowing that IBM is standing in the wings to assist, if the need arises.

Overall, we are pleased to see the notion of HPC clusters being positioned as a viable alternative for the mid-market. If IBM can successfully engage its channel partners to lead with and deliver these solutions into the mid market, we may see the emergence of another mainstream alternative to server sprawl. For a mid market firm, the opportunity afforded by server consolidation though blades with virtualization technologies combined with clusters for compute intensive workloads may prove tantalizing. To our way of thinking, the largest challenge will not be the technology, but overcoming current perceptions of what HPC clusters are about and cultivating channel partners' and customers' mindshare to reconsider how HPC clusters could be relevant in their organizations.