Features:

IS SUN LOOKING TO PURCHASE CRAY?

There are rumors circulating that Sun Microsystems may be considering an acquisition of supercomputer maker Cray. Such an endeavor would bring Sun back to its roots in IT. Timothy Prickett Morgan, in the Computer Business Review Online, has presented a compelling case regarding the advantages of a Sun/Cray merger.

A Bit of History

Before garnering its place in the server market, Sun provided UNIX workstations for academic and government research institutions. Scott McNealy, Bill Joy, Andy Bechtolsheim, and Vinod Khosla helped Sun carve a distinct position in the HPC market. Surpassing its workstation involvement is Sun CTO Greg Papadopoulos's work in creating the CM5 massively parallel supercomputer from Thinking Machines.

Sun's former chief strategy officer, Mark Tolliver, held the top marketing job at Maspar and worked at Hewlett-Packard, a Sun rival. In addition, Mark Canepa, Sun's storage unit manager began at HP, running the Apollo workstation unit that HP acquired in order to go after Sun's fledgling workstation business, eventually driving the HP acquisition of vector supercomputer maker Convex. (The Convex machines are at the heart of the HP Superdome servers.)

The SGI decision to buy Cray in 1996 created two competing lines. Sun adopted the Sparc-based Cray 6400 supercomputers from Cray Research and also contracted Clark Masters, who ended up leading Sun's enterprise server unit for years. Shahin Kahn, Sun's chief competitive officer and the general manager of its cross-divisional HPC business unit, also joined Sun after marketing at Cray Research.

These developments proved crucial to Sun's entrance into the supercomputing world. The new people understood the Sparc chip and large systems well and helped Sun to sell thousands of the newly named "Starfire" Enterprise 10000 supercomputers.

In April, many Sun executives were re-organized, with many losing their strong positions within the company. Yet Masters and Khan kept their managerial jobs. Recently, Robert Youngjohns, Sun's head salesman, tapped Masters to take a point position selling Sun's hardware, software, and services into government organizations. And now, Kahn is rumored to be leaving Sun, possibly to make room for a Cray acquisition.

Why Would Sun Want To Buy Cray?

Cray, much like Sun, has discovered the importance of differentiation in the market with the adoption of AMD's Opteron processors. For instance, Cray won a $90 million contract for its work on "Red Storm," a parallel Linux supercomputer for Sandia National Labs. In addition, Cray purchased OctigaBay, an upstart supercomputer maker which has created an Opteron supercomputer that runs Linux and can scale to over 12,000 Opteron processors to deliver 58 teraflops of number-crunching power, using the 2.4 GHz Opteron 246 processors.

Sun, if it were to acquire Cray, would be purchasing the other half of Cray that wasn't bought from SGI in 1996. SGI sold this half to a failing Tera Computer, which ended adopting the Cray name. The Cray X1 and an increasing demand for more traditional and reliable supercomputers by governments worldwide have created higher levels of stable business for the company. To top it off, Cray has been contracted by the U.S. Defense Advanced Research Projects Agency (DARPA) to assist them in reaching petaflop computational power by 2009. This makes Cray a very attractive asset. Sun and IBM were the only other two companies contracted for such work.

Cray's "Cascade" project is also on the table. Though details are shadowy, Cray says it is designing a supercomputer to include new processor designs to improve memory hierarchies. Such processor-in-memory technologies are hoped to provide larger amounts of sustainable memory bandwidths and lower latencies. The Cascade machine will also try to include shared memory and distributed memory programming models while trying to allow existing applications to port to the new Cascade architecture.

Sun, working on its own project called "Hero," plans to use DARPA's $49.7 million to create development tools that will mask the underlying parallelism of a machine with hundreds of processors, making it seem like a large, single processor. In addition, Sun is working on the benchmarks that measure the programmer productivity and performance of the DARPA prototypes.

Above all, Sun can afford to purchase Cray. Sun recently received $2 billion in a Microsoft settlement and could afford to pay up to $650 million for Cray, which took in $237 million in sales and $21 million in profits in 2003. Most of the money would fall to the bottom line if Sun bought Cray and could double its supercomputer sales.

Though Cray is having some profit dips due to their focus on Red Storm, OctigaBay, and X1E, the company has thrown off more profits than Sun in the last year. An acquisition of this nature could benefit Sun in many ways.

This story is strictly based on speculation and rumor. And though details are thin, HPCwire will continue to monitor the story as it develops.