Then came two bits of good fortune: the advent of Fibre Channel SANs and the hubris of competitor EMC. The first brought mainframe storage attachment concepts (though not networked storage) into the distributed-systems world. The latter alienated many high-end storage users and paved the way for HDS to get in the thick of things.
In 2001, HDS owned about 17 percent of the high-end storage market, compared with EMC's 74 percent and IBM's 9 percent, according to Merrill Lynch. By the beginning of 2004, however, the race had tightened, with 42 percent of the market going to EMC, 37 percent to HDS and 27 percent to IBM. Projections show the three closing to within a few percentage points of one another, driven by the pending release of EMC's Symmetrix 7 products, IBM's Power5 microprocessor-based Shark arrays and HDS's just-announced Lightning-3 TagmaStore arrays.
There have been challenges along the way, of course, including self-serving storage product characterization schemes by leading analyst firms. For example, when Gartner Group classified storage platforms as either "modular" or "monolithic," arguably to promote differentiation among vendor clients, HDS found its Lightning and Thunder arrays pigeonholed in a way that made differentiating them from competitors' wares difficult. In response, Yoshida co-authored a white paper to clarify HDS' product classifications, borrowing the ideas of "core" and "edge" from the networking space. Marketing content aside, the paper is an informative model for developing purpose-built storage capabilities behind specific applications.