Fixing A High-Speed Server

The quote server wasn't keeping up.

Our client, a trader and software developer, had a quote server written by another consultant. It had to handle 2.4 million messages a second, distributing market updates and orders to various clients.

Though it wouldn't crash, its erratic latencies were a big problem.

We spent a half day walking through the code with him. We pointed out some potential culprits, and identified other places where he could improve efficiency in the most performance-critical sections.

Our suggestions did the trick, bringing his server's performance into the acceptable range with minimal impact to the existing code.

On modern hardware, most software runs fast enough. Making a user wait imperceptibly is fine.

A real-time data-feed won't wait. That makes it a different animal.