Chapter 11. Monitoring the Health of Your Caches

only for RuBoard - do not distribute or recompile

Chapter 11. Monitoring the Health of Your Caches

Active monitoring of your cache is one of the most important steps you can take to providing a first-rate service. Most caching products provide statistics in real time via SNMP, a web browser, or some other specific interface. This data is extremely valuable in quickly diagnosing problems and complaints. When a user calls to complain that the cache is making things slow, you can tell if it's a real problem or an isolated incident.

I also feel it is important to keep a long history of measurements. If one day you suddenly notice that your CPU utilization is 80%, you'll wonder if it's always been that way, or if it's a sudden problem you should investigate. Having a long-term view also makes capacity planning for the future much easier. By extrapolating bandwidth usage data, you can decide when to upgrade your Internet connection capacity. Similarly, a long- term analysis of cache request rates and response times may tell you when your existing solution reaches its performance limits.

In this chapter, I'll talk about the statistics you should monitor and why you should monitor them. In addition to the standard measurements, such as request rate and response time, I'll explain how tracking some of the less obvious parameters can help you diagnose problems quickly. If you're on a tight budget, you can build a very nice monitoring system with freely available Unix software. I like to use UCD-SNMP and RRDTool. I'll help you get started along those lines with some installation tips and quick examples.

only for RuBoard - do not distribute or recompile