About a month ago, IBM proclaimed that it had surpassed HP as revenue leader in high-end, eight-way, Intel-based servers. This was based on Gartner Dataquest results and, while the cheery news probably caused IBM's stock to tick up a bit, it means that IBM was previously behind HP in revenues. (But based on IDC results, HP claims it took over the number one position in high-end, worldwide Unix server revenue with a 35.6 percent market share. "This was achieved by gaining 15.8 percent market share sequentially and increasing revenue by 33.9 percent. IBM lost 7.8 percent market share sequentially with a 57.9 percent drop in revenue.")
Reality check: IBM's revenue announcement doesn't necessarily mean it sold more units than HP. It could also mean either that IBM increased the cost of its units or that HP decreased the cost of its product. That would make this announcement meaningless in terms of market share, and more about the cost of goods. Flawed as this logic is, let's assume that sales are actually a real market indicator.
A few days ago, IBM announced that its ASICs (application specific integrated circuits--thingies that define the interconnection of a set of basic circuit building blocks) were being used in the Cray X1 supercomputer. Mind you, it's not IBM's supercomputer, it's from Cray--the company that once was king and then almost disappeared, the company that's been on the backburner for a while now, and the company that's building a supercomputer with AMD Opteron CPUs. IBM's parts do play a role in the X1, but, when you get to that nameplate on the front, there are four letters on it, not three.
Reality check: It wasn't but a few years ago that Cray's best effort, the SV1, couldn't hold a candle to the super-scalar MPP systems from other vendors, especially from IBM.
Shortly afterward it was reported that an IBM 32-processor "Regatta" (which uses the dual-core Power4 CPU) server smacked down HP to gain top benchmark honors from the Transaction Performance Council. The TPC-C test clocked the IBM system at 428,000 transactions per minute, a whisker ahead of the 423,000 scored by one of HP's 64-processor Superdomes. Most telling, however, was that IBM did the deed by bumping memory from 256GB to 512GB. On a smaller scale, that's an old trick known to desktop PC owners. It reduces the time spent polling to/from disk which, naturally, speeds things up. The IBM server was not intrinsically faster for any other reason.
Reality check: The fastest supercomputer honors, according to TPC results, belong to a 128-way Fujitsu juggernaut. But, since it doesn't run Oracle, as do the IBM and HP systems, it's not considered a contender in the benchmark ratings.
Reality bites. IBM is under siege and it may be vulnerable in too many areas and facing too many competitors to consistently maintain its once solid leadership position. An exception in this soft slide might be IBM's service organization. While JPMorgan Chase recently replaced a fleet of IBM portables with Dell equivalents, it also signed a $5 billion contract with IBM to provide outsourced service.
That bright light at the end of the tunnel lets IBM offer clients a huge TCO benefit, no matter whose hardware is actually on site. It may also necessitate a violent shift in the leadership paradigm we use, pushing it towards an overall activities indicator rather than a simplistic, dollars-for-hardware tally. Better still, as IBM's success in service outsourcing becomes more evident, it will spur more competition in this arena--and more competitive prices. We may see 2003 ushering in a significant plummet in real TCO.