At Silicon Graphics' Whidbey Research Laboratory (WRL), we have realized the importance of satisfying customer demand for commodity applications on commodity operating systems, more specifically, Windows NT.

Clearly though, we have to maintain the superiority IRIX affords us in the high-performance compute, database, and other markets while satisfying these new customer requirements.

Our current research, therefore, hinges around finding a way to integrate Windows NT into the IRIX environment seamlessly. We have focused on the following three areas:

- Hiding Big UNIX Iron in an I/S NT shop.
- DSM vs. Wolfpack, who's to win?
- Vector processors on NT, why not?

Servers are moving back into I/S and TCO is of paramount importance. Our research in this area focuses on hardware and operating system configuration architectures that permit the complex administration of large UNIX machines into the NT administration framework.

In our second area of interest, we are trying to understand whether the apparently divergent approaches of clustering vs. DSM, as apparent in NT and UNIX respectively, are truly incompatible or whether they can coexist synergistically.

NT has had a successful history of running on commodity processors. HPC research still requires special purpose hardware architectures, however, much of the software that HPC users use is aimed at the NT/Win32 environment. Does it therefore make sense, to push NT as the base API on HPC machines?

Note that the above are the unconstrained thoughts of the members of SGI WRL and should not be construed in any way as Silicon Graphics policy, intentions, or interest.

---- Akmal Khan
akmal@sgi.com