Today at work, I/we have reach a milestone. I’m in charge of the render-farm, that is a bunch of machines that all they do is process/crunch numbers. When I came the company, we had on average about 6% of the whole thing down, the number they wanted it to be was either 3% or 3.5%. Problem is that numbers vary on points of view, there is nodes/machines vs. Procs/CPUs.
For the Procs there are 5444, for nodes there are 2623. The discrepancy is based on each node having from 1 to 4 procs (Dual CPUs, Dual CPUs with dual core, etc.) We keep track not only of the hosts vs. CPUs, but also vs. the ones that are out for repairs and the ones that are about to be retire.
The average of procs down was about 6% when I came in, managed to shift it to 4-5% by jerry-rigging and fixing slowly, but now that it has been a slow time for us, I jump on it, got a vendor/support guy to help and have been able to bring it to 3-4%. Hopefully at this pace, it will be closer to 3% in the next few days, once I get the next shipment of fixed nodes.
About time I felt a bit accomplished.
Entries (RSS)