[Note: shameless name-dropping follows.]
The grand-daddy of all operating systems conferences is in town this week: USENIX. 20 years ago this was a forum for trumpeting the importance of Unix in a marketplace that was still dominated by vendor-specific operating systems (VMS, Domain, MVS, and even this insignificant upstart called MS-DOS…). I remember demonstrating PC-NFS at USENIX (and the related show for “suits” called UniForum), and everybody was amazed that these toy systems could actually play with the big guys. O tempora, o mores… Today USENIX is about operating systems in general, and this week’s symposium is OSDI’06, on operating systems design and implementation. (The other big USENIX event is LISA, where sysadmins for Really Big Systems get together.)
I didn’t actually sign up to attend USENIX (Amazon.com is much more frugal about these things than Sun used to be), but several of my friends are involved in the event, and I arranged to have breakfast this morning with Jim Waldo. We were joined by Margo Seltzer, and had an interesting discussion about varieties of systemic errors in large-scale distributed systems. Jim and I had planned to meet for dinner, but during the day he emailed me to suggest that I join him for the poster session that evening. So I did. I had a great time, met a lot of old friends, and made some new acquaintances including Jim Thornton of PARC, who used to work with my Amazon.com colleague Marvin Theimer, and Liuba Shrira from Brandeis – it turned out that she was an ex-neighbour from Brookline!
I’m actually not very good at poster sessions. I find that I want to actually read the interesting ones, which usually conflicts with the expectation of the poster presenters who want to talk. And sometimes (rather too often, unfortunately), when I finish reading the poster, I realize that most of the ideas have already been incorporated in some other piece of research, or perhaps even a commercial product. It’s really hard to tell a bright-eyed grad student that they need to go back and redo the literature search phase of their project. (Marvin is better at it than I.)
Having said that, there were two initiatives that I definitely want to follow up in the cold, clear light of day: Shirako from Duke, and Plush from U.C. San Diego. The problem statements look exactly right; I’ll be interested to see how much progress they’ve made.