On Friday, we had a classic consulting gig. We were brought in to advise on accelerating and parallelizing the processing of gene chips. On walking into the room it was totally clear that we were also catalyzing a conversation that needed to happen between the group who runs the machine, and the groups who use the data. Turned out that, so far as we could discern, the gene chip process involves a 16 hour hybridization (rate limited by biochemistry), a 10 minute scan (rate limited by the scanner…and by the fact that a human has to move the chip into the scanner), and 2 or 3 minutes of processing to update a database. Clearly, we weren’t going to do anything for throughput by optimizing that 2 or 3 minute process. Anybody (capital_l) know a case where parallelizing the GCOS server for an Affy chip setup was smart and necessary?
Still, I think they got value from the “wooden indian” effect. Wheel the wooden indian into the room, explain your problems to him, and suddenly the answer seems clear. Thank you, wooden indian.
I’m in PA this weekend helping my brother-in-law (Doug) and his wife (Michelle) move into the house they just bought. It’s one side of a duplex, and I’m very impressed. The neighbors seem to be young professionals, and several of them came over and introduced themselves while we were hefting desks and sofas. We return to RI tonight.
Tomorrow, I fly to St. Louis and help the Danforth Center install and configure their new cluster for two days. Should be a fun gig. The people involved seem cool and competent, all the hardware is already there, and their stated goal for the trip is to “learn how it’s set up and configured and to become self sufficient with maintenance.” Who could ask for better? My life is good.
Leave a Reply