When setting up an APM, it is essential to define a small set with important key metrics. It can be tempting to start with a higher number of metrics without fully understanding what they mean, or what they are showing. Following on from our earlier article which looked at APMs in more detail, here we will take a look at standard key metrics.
TL; DR Summary
- Apdex Score – Industry-standard metric that measures user satisfaction on performance.
- Response time – The time taken to respond to a transaction or all transactions.
- Error rates – Number of errors produced by the system for a given time.
- Throughput – The rate at which requests are served by the system for a given time.
Key Metrics
APM metrics are key to understand the various aspects of the system. In general, any APM tool should support the following metrics,
- Apdex score: This is the APM-specific, industry-standard metric and it measures and tells what the performance level satisfaction of a user with an overall system or specific transaction is.
- Response time: This measures and tells the response time of the overall system, dependencies, and individual transactions for a specific period.
- Error rates: This measures and tells the current error rate of the system for a specific period.
- Throughput: This measures and tells at what rate requests are being served by the system for a specific period.
- CPU usage: This measures and tells the CPU usage of the system for a specific period.
- Memory usage: It measures and tells what the Memory usage of the system for a specific period is.
- Instance count: This measures and tells how many instances are currently running in a specific time range.
- Customer experience: This tracks and tells how users are using the system and their experience through various interactions.
Once all the key metrics of APM and how to infer them are understood, it is essential to define the key metrics and implement them with any APM tool.
Create a framework to iteratively optimise the existing metrics and introduce other necessary metrics. Ensure these metrics are producing more reliable results, if not improve them or remove them when no longer valid.
Author: Sreenivasan Madhaiyan