[jrockway]

Yup. When I was a junior engineer, I was working for a bank and the various teams were all blaming each other about high latency in the stack, which involved Tibco Rendezvous. Basically, a way to distribute market data over multicast. They blamed Rendezvous, they blamed the network hardware, they blamed everything except their spreadsheets. (Yes, the tech stack at the time was spreadsheet <-> spreadsheet over multicast.) Anyway, I wrote a small Perl application that measured the end-to-end latency (when market data changed, vs when the published output changed) in addition to network induced latency (basically echo packets; write something out, see when it came back over the link). What the data showed was that the data processing spreadsheet introduced tons of random latency, and the network was totally fine. I sent that up the chain and the bug in the spreadsheet was found almost immediately, solving a multi-month blame game in hours.

A little off topic but good memories.