I was reading a discussion on a social media site about Application Performance Management, and realized that there is a lot of confusion about what is Application Performance Monitoring, Application Performance Management (APM) and IT Operational Analytics (ITOA).
Just looking at the words used, you would believe that Application Performance Monitoring is focused on watching data and monitoring it for a particular condition or state. Application Performance Management would lead you to believe that this is a wider field which includes a range of techniques to certainly monitor the application, but also to manage other aspects of the IT estate. The degree to which complex analytics are used is unclear, but potentially IT Operational Analytics could be seen as a subset of Application Performance Management, although the focus on application might make it more limited in its scope than ITOA.
To help clarify this rather muddy set of terms, we use two models which we find are much clearer and logical, and have less ambiguity than the APM and ITOA definitions.
The Monitoring Maturity Model
The first model we call the Monitoring Maturity Model, because it is a layered model where generally the higher levels are based on data collected from the lower levels. The model is:
1. Infrastructure Monitoring: Collection data on the servers, operating systems, network and storage and setting rule based alerts to catch potential problems.
2. Basic Application Monitoring: From interrogating the Operating System, capture and alert on data about the processes running on the servers. This would include CPU & memory utilization, disk I/O, network I/O etc.
3. Advanced Application Monitoring: Installing a tailored agent on the server which is capturing data specific to the application it is monitoring. This can be "inside the app" data or "outside the app" which is useful for Off the Shelf software products and middleware.
4. Flow Monitoring: This is capturing data about the information passing between applications and monitoring/reporting on data flows. This would include volumes/second, volumes per counterparty, latency etc.
5. Business and IT Analysis: This is the analysis of both business data and "machine" data from levels 1 and 2 to understand the business activity and the behavior of the IT estate.
Monitoring vs Analytics
The second model is separating monitoring from analytics. There is no hard definition which separates them so we break the types of analysis into three types:
1. Detect: This is a rule based detection of an alert condition. This is generally what people mean when they talk about Monitoring.
2. Analyze: This is the collection of lots of data, even data which did not trigger a rule in Detect, and analyzing it to discover more insight. This may be as simple as trends, or as complex as Machine Learning and time series pattern based Anomaly Detection. This would also include techniques like Bayesian Network Causal Analysis.
3. Predict: This uses current and historic data to try and predict future or “what if” scenarios. Again, this can be as simple as extrapolation, or as complex as comparison of current state to empirically derived behavioral data, the likes of which you might have gathered in a performance lab when stress testing an application.
Whichever way you model your IT estate and the behavior of your applications, it is necessary to have a clear language so that people are talking about the same thing.
Guy Warren is CEO of ITRS Group.