Overheard at the April 2007 Monitoring SIG - II

I’m sure there are lots of examples of computing environments with “bursty” demand. For example, I once attened a talk by someone from the Stanford Linear Accelerator Center (SLAC) about the unique demands of capturing and logging data from atomic research experiments: the event of interest (typically, a particle collision of some sort) occurs in a fraction of a second, but gigabytes of data are generated instantaneously. How do you buffer it while you write it out to slow(er) storage? (And then, as an aside: how do you share it with collaborators at remotely located facilities like Fermilab and CERN?)

A more pedestrian example came up at April’s BayLISA Monitoring SIG, hosted here at GroundWork. Imagine a school’s class registration application: most of the time it sits largely idle, but at the start of each semester there’s enormous demand as students try to set up their schedules. How does your monitoring deal with a scenario like that? How can you use your monitoring to make the case to the school administration that more resources are needed to meet a narrow but critical peak demand?

The interesting thing about this scenario was that it was presented as an example by someone who doesn’t work at a school, and as he was talking I saw some heads up front nod in agreement. When we got to the introductions, it turned out that those nodding in agreement were four IT staffers from a local state college: obviously, they had intimate experience with these issues.