The two major classes of users are technical customers and enterprise customers. The former comprises scientists and engineers who need to perform number crunching; examples include climate prediction, protein folding simulations, oil and gas discovery, defense and aerospace work, automotive design, financial forecasting, etc. The later category encompasses the corporate data center that stores customer records, inventory management, and employee details.
Technical customers usually require the Message Passing Interface (MPI), whereas enterprise customers require an SQL database. Because a number of MPI and database implementations have been created with sockets or DAPL, it is possible for a company to target both classes of customer by delivering a quality software stack.
Unfortunately, many smaller HPC companies often target technical customers in the hopes of becoming famous, and in the process neglect enterprise customers who would make them rich. Enterprise is a much bigger market than technical simply because storage is required more than processing.
This blog should not be considered a source of stock market advice, though the trends discussed here so far point to which companies should do well and which will probably do poorly. Mellanox, Myricom and Linux Networkx are by far the most respectable for understanding the dynamics of the marketplace, whereas Dolphin, Quadrics and Cray seem to have forgotten the lesson from above.
The most appalling example is PathScale, which has completely reversed its original strategy of only providing MPI and then only on HTX; they are now in the process of developing a complete OpenIB stack along with a PCI Express adapter. PathScale was recently bought out by QLogic for just under $110M. Quadrics, which according to the Top500 has a much bigger market share than PathScale, only posts $20M in sales annually. It seems unlikely that a reasonable use of discount cash flow analysis would price PathScale as being worth so much, especially since the InfiniBand market (and indeed cluster interconnects in general) is leaning towards commoditization. It would appear that QLogic made an ill-advized investment.