Today, unlike other days, I plan to be a total and complete geek. See, LinuxWorld is in Boston, and I’ve got me a pass to the exhibition floor. I’m ‘a gonna walk around and see all the cool toys. Here’s what I’m interested in:
* Which update management scheme is going to step up to the plate left vacant by RedHat’s up2date. I’ve tried SuSE, SLES, Gentoo, BSD, Fedora, and RHLE (no I haven’t tried Mosix. Go away). I’ve seen people start with each of these and hack their way over to a different update management scheme. I’ve seen nasty, nasty hybrid clusters which kept an old os (RH 7.3) for the compute nodes, and switched over to SLES for the portals. It seems that there’s no consensus in the community anymore. It’s possible that we’re in the middle of growing ourselves a mature, robust way of dealing with application level updates which might wind up being distribution independent. On the other hand, we could just be in the middle of a painful code fork.
* Where are the 4 and 8-way SMP desktop and rack mount boxes?
* How are all those crazy specialized hardware companies doing? Who’s offering an accelerator board for gcc, for example? How about blade systems with 64GB of shared memory between 64 CPUs, and hyper-connect switching between the 64-ways?
* Who bought one of Cray’s linux based systems? Do they like it? How about Deep Blue? How are they doing? Is it just stunt computing, or can I buy one yet?
* What’s the big market push this year? Is it still bioinformatics and the grid, or have we moved on to bigger and better things (not that anything could ever be bigger or better than the grid…mind you).
Leave a Reply