Reuse of monitoring data
As hinted at in various other articles, there is a huge overlap between the data needed for monitoring, usage tracking & capacity planning, and (cost) charging. Assuming you set up the comprehensive set of probes required for Total Monitoring, you will be getting a heavy stream of information about response times to test requests, and also a set of platform data - CPU percentage busy, how much disk space is used, etc.
Although it is probably not required (or practical) to save every measurement in that stream, a summarised set of data (averaged or summed over longer periods) is exactly what is needed for tracking and forecasting usage trends, and for calculating a usage-based bill to send to your customers.
The diagram below illustrates this. The red elements are components of traditional real-time monitoring for immediate problems. The blue elements show how a subset of that monitoring data can be copied into longer-term storage for later use, in usage and charging tools. The thickness of the arrows roughly indicates how the quantity of data reduces as it passes through the various stages.
Obviously, your usage data could be populated by a separate set of data-gathering tools, but why invent a whole separate framework for them? If there are any metrics that are required for charging or tracking that have no use in real-time monitoring, you can still use the same probe framework to acquire the data, and merely ensure it doesn’t trigger any action in the monitoring tools.
