Summary

This chapter considered the technologies to use and best practices to follow when you add logging features to your distributed system. In the real world, these designs are neglected far too often, leading to countless headaches when problems occur in an application that lacks a sufficient way to record them.

You should always consider these three final guidelines:

• Test your exception handling and logging code.

  Some reports suggest that logging code typically contains the highest concentration of errors relative to any other part of an application because it is seldom properly tested.

• Quantify your theory with stress testing.

  Unless you put your system under a heavy load and watch the performance counters, you won't know what unique bottlenecks are holding it back. Without proper profiling, all other considerations are just theory. Well-documented accounts show applications that have increased their performance by several orders of magnitude just by addressing a single bottleneck.

• Remember the difference between scalability and perfor- mance.

  When a bottleneck presents itself, ask yourself whether it is a scalability problem that appears only under heavy client loads or whether it is a performance problem that slows down every user. Answering this question will help you determine the most likely course of improvement and prevent you from wasting time with optimizations that won't solve the problem at hand.