I've just returned from a trip to Silicon Valley, where I visited Cisco, Google, and Yahoo -- all of whom have generously given me research money this year, and hence the CA visit. Besides thanking them in this blog, I thought I'd say a few things about the trip and what I saw on my brief stops at the various places. The purpose of these visits is mostly to see if I can get collaborations going, but I admit, some part of it is just giving face-time to the people and companies who have given me research money. They deserve some of my time, and I'd like to encourage them to keep providing funding!
The first stop on my trip was actually a visit to Microsoft Silicon Valley Research Lab (MSRSV). I still haven't figured out how to get research money from Microsoft, but MSRSV "started" when a lot of my colleagues at what had been DEC/Compaq/HP Systems Research Center moved en masse to Microsoft, so I have historical ties and recent collaborations with people there as well. Since my visit last year, MSRSV has moved into a very nice new building. Lots of open spaces and whiteboards everywhere. It seems wonderfully set up for group collaborations. (One very nice space for group-work, though, is a bit too close to a loud and frequently used coffee machine for my taste...) Besides catching up with everybody, Udi Wieder and others indulged me by talking about some of the many variations of trace reconstruction problems that are still open. Hopefully we'll get somewhere on some of them.
Cisco is a huge sprawling collection of buildings, and the visit there itself similarly felt chaotic. They asked me to give two talks, which caused me a bit of stress the week before the trip as I reworked some slides. I ended up talking about my work with Salil Vadhan on Why Simple Hash Functions Work (talk, paper, blog post), and gave a mini-survey covering my work with Adam Kirsch (and part with Udi Wieder) on how to use CAMs to improve cuckoo hashing (talk, various papers on my papers page). [Actually, I have a new survey article covering this stuff I'll put up shortly.] Cisco still seems very, very generally interested in hashing, and applications of hashing in network measurement and monitoring in particular. I had about 40-50 people show up for the first talk, and the second mini-survey talk was broadcast and recorded for Cisco -- about 50 people showed, and apparently more than that were also listening remotely. (Just like when I teach, and my class is taped...) They have a pretty elaborate setup for these recorded technical talks, with a room set up for guests like Steve Wozniak (who was there a couple of weeks ago) rather than me. Besides giving talks there were a lot of high-level discussions about things going on with Cisco where I might be able to collaborate usefully with them.
One thing I noticed at Cisco was a much larger number of women than usual at my talks. Perhaps EE is turning out more female graduates than CS recently, or it's somehow reflective of Cisco's hiring practices.
Visiting Cisco is always very exciting. They're a lot more short-term focused than research labs, but there is this wonderful sense that what you're talking about could become a part of the fundamental network architecture. They keep me away from details, but multiple-choice hash tables and Bloom filters seem to be standard tools in their arsenal now. I'm hoping some form of cuckoo hashing might be as well someday.