Performance monitoring

So far we have seen how to measure and improve both CPU and memory performance, but there is one part we have completely skipped over. Performance changes due to external factors such as growing amounts of data are very hard to predict. In real life applications, bottlenecks aren't constant. They change all the time and code that was once extremely fast might bog down as soon as more load is applied.

Because of that I recommend implementing a monitoring solution that tracks the performance of anything and everything over time. The big problem with performance monitoring is that you can't know what will slow down in the future and what the cause is going to be. I've even had websites slow down because of Memcached and Redis calls. These are memory only caching servers that respond well within a millisecond which makes slowdowns highly unlikely, until you do over a 100 cache calls and the latency towards the cache server increases from 0.1 milliseconds...