Metrics for a web application
October 13, 2010
Routinely monitoring metrics for a web application is critical. Here is a brief introduction to metrics and some points on which are the critical ones.
Introduction
Web applications are fortunate that it is possible to observe what the users are doing with the application, accurately and in real time. Running a web application without the metrics would be like flying blind.
Types of metrics
There are mainly two types of metrics:
Operational metrics - these include Number of application servers live, server CPU, memory and disk consumption, background queue length, number of emails sent per day etc. Basically, these are the things Engineering team would be interested in to make sure the application is healthy and fast.
Business metrics - these include thing like Daily active users (DAU), number of new registrations, account cancellations, revenue etc. Basically, the metrics that the Business team (Sales and Product management) would be interested in.
There are a lot of opinion about which metrics are absolutely critical to track. Some are mentioned below in the references.
With each metric, it is also important to assign an SLA. An SLA is a high or low watermark signalling that something is wrong and the team needs to take action to improve it. For example, you may set an SLA on page load time or number of errors thrown per day.
Operational metrics - these include Number of application servers live, server CPU, memory and disk consumption, background queue length, number of emails sent per day etc. Basically, these are the things Engineering team would be interested in to make sure the application is healthy and fast.
Business metrics - these include thing like Daily active users (DAU), number of new registrations, account cancellations, revenue etc. Basically, the metrics that the Business team (Sales and Product management) would be interested in.
There are a lot of opinion about which metrics are absolutely critical to track. Some are mentioned below in the references.
With each metric, it is also important to assign an SLA. An SLA is a high or low watermark signalling that something is wrong and the team needs to take action to improve it. For example, you may set an SLA on page load time or number of errors thrown per day.
How often to track
Although the metrics should be tracked very closely. However, some critical metrics should be tracked daily by key team members and the rest should be tracked weekly. A weekly meeting with discussion around metrics is often the practice.
Tools
For operational metrics, you can start with a log analyzer to spot potential improvements in your app. AWS has very good server metrics and if you need something more powerful, New Relic is very good in monitoring app performance as well as servers. If you are starting from scratch, you can start tracking Business metrics using Google Analytics or Clicky.