Replies

There's actually a forth, *BSD (Free and Open and Net).

You are correct, and I can only guess as to why he would not mention that. Probably A) he considers their market share and likelihood to increase to be very small, no matter how correct that may or may not be or B) he only listed operating systems that were available on the retail market.

My guess would continue that it is probably B, and as *BSD is without a significant x-tier support structure from anywhere (unless you or someone knows of one) he doesn't actually consider them much of a competitor to the high end business or government customers that Sun currently supports.

And also, a possible C) since remember Sun is trying to walk a fine line of being pro-"open source" as well as offer proprietary solutions, so they probably are purposefully avoiding criticisms of all things open source. Interested in your perspective, if you have time.

I think the BSDs have zero or very limited multiprocessor support, and that has limited their adoption beyond web servers and firewalls. It is too bad, because BSD is absolutely superior for network intensive operations. Its IP stack is like a Ferrari.

As for Linux, with the upcoming 2.6 kernel it is clearly being architected for the mid-tier. It has a much improved threading model, which is perfect for J2EE appservers.

The thing that will hinder Linux on the database tier are its lack of strong fault logging compared to Solaris, HP-UX, and AIX, and Linux's "grab bag" of disk volume managers and filesystems. Until their is a shakeout of disk volume managers and filesystems, none will get the programming attention needed to make them as mature as Veritas or IBM's AIX products. I know IBM has provided an open source version of JFS to Linux, but it is not that strongly adopted. There are just too many different ones for Linux right now: IBM JFS, SGI XFS, EFS, Reiser, etc.

The "proprietary" implementations of UNIX (Solaris, HP-UX, and AIX) are continually being improved in the areas of availability technologies and resource management. As such, they are becoming the new MVSs of the datacenter. I do not see how an open source model will ever catch up with them, however if "proprietary" UNIX becomes a niche player, "proprietary" UNIX may fade away.

Perhaps the answer is a "proprietary" UNIX that looks and feels like Linux, runs on the same hardware, and runs Linux binaries, but has all of the high-end features needed to run mission critical databases.

My guess is if IBM, HP, or Sun do not come out with such a software-only play, Microsoft will continue to add availability features to Windows Datacenter and eventually take this market.