From Computer Business Review, a sister publication
Siemens Nixdorf Informationssysteme (SNI), the computer division of electronics and engineering conglomerate Siemens AG, has spent the best part of three years trying to shake off the image of a typical German industrial giant: bureaucratic, reliable but dull, efficient yet with a clubby civil service culture. Pyramid Technology could not be more different.
By Graeme Burton
A product of Silicon Valley entrepreneurialism, the San Jose- based high-end server manufacturer has a sand-filled volleyball court in the middle of its car park and lines the corridors of its offices with pocket-sized packets of jellybeans. It is unsurprising, therefore, that the marriage of the two companies a little over two years ago was viewed by many in the industry as one of the greatest personality mismatches of all time. But it was a union driven by necessity – both companies had something the other desperately needed. SNI had money and thousands of customers across Europe but few in the US; Pyramid had highly- regarded Unix server technology, a strong American salesforce, and an entrepreneurial image unheard of in Paderborn, Germany, SNI’s home town. Gerhard Schulmeyer, the newly installed chief executive of SNI, was looking for ways to get a foothold in the US computer market and was desperately trying to give his company an injection of Silicon Valley entrepreneurialism to help it stay in contention in the global computing market. SNI saw Pyramid as its ideal partner because, not only would it provide that all important Silicon Valley base, but the two had a long record of cooperation. Failed in Europe SNI was an original investor when Pyramid went public in 1985 and already handled sales of Pyramid machines in Europe (along with UK-based Fujistu subsidiary ICL). Second, they had a broadly complementary product line-up – both continue to use the MIPS RISC chip in their Unix servers in preference to Intel. And third, Pyramid was relatively inexpensive. Now, however, more than two years into the partnership, there are suggestions that the merger has failed to meet its goals. Pyramid has lost ground on its competitors, say analysts, and has failed to exploit the European opportunity. SNI meanwhile has taken greater control, closing Pyramid’s production line in San Jose for example, and moving output to Germany. And staff have left, claiming Pyramid is no longer a ‘cool’ company to work for. Certainly, SNI has taken Pyramid to its core. During 1996, the company reorganized its entire server business, passing control of all server sales and marketing to Pyramid in America, Britain, Japan and Korea, with SNI handling both company’s server sales in the rest of the world. Subsequently, Pyramid chief executive John Chen has been given the task of overseeing SNI’s entire server sales business, Open Enterprise Computing, and has been appointed to the board of the company. This move has made it difficult to distinguish the performance of the two companies’ products and, say competitors, covers up the fact that Pyramid is still failing to win much new business in Europe. The key reason why Pyramid’s board chose to sell out to SNI, was to do just this. Pyramid’s aim was to exploit SNI’s European corporate customer base – 4,000 in Germany, Austria and Switzerland alone. If we could capture just one tenth of those users, that’s more than we could ever manage on our own in five years, says Chen. That does not, however, appear to be happening. According to David Charmers, head of European technology at arch rival Sequent Computer Systems Inc, Pyramid is no longer its major competitor. Although the companies competed head-to-head in the high-end SMP symmetric multi-processing server market for many years, recently Sequent has accelerated ahead, reporting turnover for 1996 of $595m, compared to Pyramid’s $275m. Charmers also criticizes Pyramid’s technology strategy. I think they have failed to make the move from a fairly good SMP system to the next generation. In our view they took the wrong decision to build massively parallel processing (MPP) boxes, he says. Pyramid is notable in the high- end Unix server market for two reasons. One, it has so far stuck steadfastly to the MIPS RISC chip while a number of its rivals have defected to Intel. Two, it is also decidedly lukewarm in its enthusiasm for NUMA non-uniform memory access, the technology which is emerging as a standard design for a new breed of high end servers called shared memory clusters SMCs. While Pyramid was struggling to deliver its RM1000 MPP server, for example, Sequent was revamping its product lines around cc-NUMA cache coherent- NUMA technology. Recently both Hewlett-Packard and IBM have also committed themselves to the technology. Chen rejects the criticism. I think it [MPP] will break out, riding on behind the data warehousing applications world. MPP offers two things. It runs certain data-stream oriented applications extremely well from a technical point of view and it also provides resiliency and ‘re-tryability’ as part of the architecture. Database companies like Oracle and Informix need that. Furthermore, Chen believes that NUMA, despite the growing number of adherents in the industry, is a flash-in-the-pan which will burn out as applications taking advantage of the MPP architecture grow in number. The NUMA architecture still does not solve the ultimate scalability and availability problems. The SMP/MPP integrator we’ve been working on will solve that. We absolutely totally believe that we are on the right track. says Chen. Chip architecture problem Pyramid’s strategy is to wait for the market to come around to its way of thinking, a move it could not have considered without the financial support of SNI. But whether Siemens AG, Pyramid’s ultimate owner, is in the mood to wait for the MPP systems boom is another thing entirely. One rival says there is as yet little evidence of Pyramid products among SNI’s customers. Perhaps the biggest problem facing Pyramid is its chip architecture. Outside of Silicon Graphics, owner of the MIPS chip technology, SNI and Pyramid are the only companies using MIPS as the basis for high-performance computers. They’re lacking economies of scale because they use MIPS instead of Intel. That’s not going to kill them tomorrow but the longer they stay off a volume architecture the bigger the problem will become, says Sequent’s David Charmers. Even Chen admits that Pyramid has not been able to grow sales of SNI’s Primergy Microsoft Windows NT servers in the US which are based on Intel processors. Observers believe that the pressure in the high end Unix market will force alliances among the smaller players like Pyramid. Chen says Siemens wants a big information technology position in the US. While Pyramid is a good starting point, it certainly couldn’t be an organic end-point.