This is not a simple question to answer. It likely will change a few times over the course of time. But we can be reasonably sure that their won’t be widespread installations of Irix, AIX, HP/UX and others of their ilk. There are many reasons for this, technological, legal, business, marketing, and so forth.
Looking at the top 500 list, it isn’t a high risk bet that Linux will remain in some form or the other.
So who are the contenders? Who would like to be a contender?
First, Microsoft really wants Windows everywhere. So far Windows has largely failed to penetrate these systems in any meaningful way. I am sure there are pockets of a few clusters here and there. I have asked our customer base about this repeatedly, whether or not they want to see a Windows based cluster. I haven’t heard a “yes” to date. I posted a question on the beowulf list and got about 1 dozen responses, 9 of which were from vendors. There is an opportunity for Microsoft to add value. Where and how is the question.
Second, MacOSX. OSX is a unix-like system. Courtesy of some IMO poor marketing decisions, it will only run on Mac machines. If Apple thinks I am going to drop my AMD boxen for a “Core Duo”, well … heh … no. AMD64 machines seem to be the hot commodity in supercomputing these days, so its not hard to figure out at least in the short term what will happen. Longer term, again, few groups will buy hardware from a single source for a significant premium when comparible and generally faster boxen are available for less. Some of us used to say something about this at SGI. You can’t paint a box purple and charge 3x the price of a comparible box, and expect it to sell.
Solaris. I can sum up a number of reasons why I don’t think it is a Linux killer, but two major reasons are pretty much all you need. #1 Drivers. I cannot get Solaris drivers for my collection of computing gear. #2 Applications. I cannot just run my Linux apps on Solaris in a container. Janus was supposed to get us there. Now Sun is pushing Xen for this.
There may be others. The *BSD crowd would love to take over another niche. Possibly some RT/OS folk or embedded OS folk would make the arguments that they will dominate.
The mantra is mindshare. Applications follow mindshare. Device support follows applications, marketshare, and mindshare. Users want something that works, and they want choices. Plug it in, turn it on, instant gratification when they can run. Which of these OSes enables the widest regions in hardware choice, and application choice?