LMS Performance Overhaul

PrashantAdvait Foundation runs an education app that had grown to two million downloads faster than its backend could keep up — one core screen took over nine seconds to load. On a slow connection, for people with little patience to spare, nine seconds might as well be forever; they just leave. The database was doing more work than it needed to: queries with no index to speed them up, and screens pulling far more data than they actually used. I added the indexes and trimmed the queries down to what the screen actually needed. That took response time from 9.2 seconds to 141 milliseconds and cut total database load by 93 percent.

It's a nonprofit, not a startup with infrastructure budget to throw at the problem, and the app was already being used by tens of thousands of people a day — so the fix had to live in the query, not in more hardware, and it had to not break anything already live.