API NetWorks and Cray Inc. Announce Strategic Alliance to Introduce First Move of Alpha Linux Into Large-Scale, Mission-Critical Clustered Environments

API NetWorks and Cray Inc. announced a strategic alliance under which Cray will offer large-scale Alpha Linux clusters built on API NetWorks server technology, backed by Cray's integration, service and support, marking the first move of Alpha Linux into mission-critical supercomputing environments.

CONCORD, Mass. — January 29, 2001 — API NetWorks, Inc. and Cray Inc. today announced a strategic alliance to bring Alpha Linux clusters to customers who run large, mission-critical technical computing workloads. Under the agreement, Cray will deliver clustered systems based on API NetWorks Alpha servers, including the dense 1U CS20, and will provide the system integration, software environment and field support that its supercomputing customers expect. API NetWorks will supply the server hardware and engineering expertise in Alpha platforms.

The alliance pairs one of the best-known names in supercomputing with the company that has done the most to turn the Alpha processor into a commodity building block. For customers, it means an Alpha Linux cluster no longer has to be a do-it-yourself project; it can be bought, installed and supported as a production system.

Why this alliance matters

Until now, Alpha Linux clusters have been adopted mainly by research groups and technically adventurous organizations willing to integrate and support the systems themselves, often with help from specialist integrators such as Linux NetworX. Large government, defense, weather and industrial sites have been more cautious. Their workloads run around the clock, their deadlines are fixed, and they require a vendor who will stand behind the entire system, not just individual components.

Cray brings exactly that. The company has decades of experience installing and supporting supercomputers at the world's most demanding sites, and a customer base that measures performance in sustained results on real applications. Its familiarity with Alpha runs deep: Cray's massively parallel T3D and T3E systems were built around earlier generations of Alpha processors. Extending that relationship to Linux clusters is a natural step.

What the alliance covers

• Hardware: API NetWorks Alpha servers, beginning with the dual-processor CS20 1U system based on the Alpha 21264, as the compute nodes in Cray-delivered clusters.
• Integration: Cray configures, tests and delivers complete clusters, including racks, interconnect, storage and management infrastructure.
• Software: a Linux-based environment with compilers, math libraries, MPI and cluster management tools suited to technical computing.
• Support: Cray provides service and support for the delivered systems, giving customers a single point of accountability.
• Engineering cooperation: the companies will work together on configurations and future platform requirements for large clusters.

Division of responsibilities

Technology behind the clusters

The CS20 was designed for exactly this kind of deployment. It fits two Alpha 21264 processors into one rack unit, so a single rack can hold dozens of dual-processor nodes. Each processor has a large dedicated backside cache and strong floating-point performance, and 64-bit PCI slots allow a high-speed cluster interconnect to be added without bus bottlenecks. Because every node is identical and remotely manageable, large systems can be installed and maintained from a central console.

Running Linux on those nodes gives customers access to a fast-growing body of open-source tools, as well as the optimized Alpha compilers and libraries that have long made the architecture attractive for scientific code. The 64-bit address space removes the memory limits that constrained 32-bit cluster nodes of the period, which matters for large simulations and data-intensive analysis.

Target environments

1. Government and defense laboratories running large-scale simulation.
2. Weather forecasting and climate research centers.
3. Automotive and aerospace engineering groups using crash, structural and airflow codes.
4. Energy companies processing seismic data.
5. Universities operating shared research computing facilities.

The question customers kept asking was not whether Alpha Linux clusters were fast, but who would stand behind them. With Cray delivering and supporting these systems, that question has an answer.

An API NetWorks spokesperson

Market context

Early 2001 was a turning point for supercomputing. Commodity clusters were climbing the TOP500 list, and even traditional vendors recognized that many customers wanted cluster economics without cluster risk. Cray itself had re-emerged as an independent company in 2000, when Tera Computer acquired the Cray business and adopted its name, and was building a portfolio that spanned vector machines, multithreaded architectures and clusters. Adding an Alpha Linux cluster line gave it a price-performance option for customers who did not need a custom architecture.

For API NetWorks, the alliance validated a strategy pursued since the company's founding: making Alpha available through partners who add systems, software and service, rather than competing with them. It also created a channel into accounts that API could not easily have reached on its own.

About API NetWorks, Inc.

API NetWorks, Inc., formerly Alpha Processor, Inc., is headquartered in Concord, Massachusetts. The company was founded in 1998 as a joint effort of Samsung and Digital Equipment Corporation, later Compaq, to bring the 64-bit Alpha RISC architecture into mainstream markets. API designs and markets Alpha motherboards, processor modules, high-density rack servers and system-level building blocks for Linux and Tru64 UNIX, and supplies them to OEMs, system integrators and cluster builders worldwide. Through its API NetWorks unit the company is extending its high-performance design expertise into interconnect silicon and platforms for networking and communications equipment.