For whatever reason, the decision to implement a proper web application monitoring is usually an afterthought. It’s possibly because the entire development of the application is lead by a development team, not always very interested in the operational aspect of the application.
As a result of great development efforts and marketing, an application might take off and attract new customers quickly and gain in popularity. The demands on the application may change dynamically as new features are brought in and as updated versions occur on a weekly, if not daily, basis. Everyone is happy with the growth and the development team is recognized for its efforts.
Then comes the day that customers call in because the application is not responding quickly as it did four weeks ago, with some basic functionality hangs at mid-session, or even worse.. That’s when the system administrator gets a call and it’s panic. The blaming game starts… and he will sifting through the logs trying to get a clue of what happened.
Web application monitoring at Syloé
We believe in preventative measures and implementing web application monitoring as the applications are being rolled out. We develop application performance tests with our customer to identify problems in the various steps of an application request: authentication (basic or token-based), simple front-end query, simple back-end query, complex query, logout. For all these various stages we record response times/codes and present them in a graph. Based on customer-designated targets (or performance test results) we configure alerts when response times are not as expected : when certain steps could not be performed in time or we receive an error code from the application (50x, 40x etc).
In addition to this, we will monitor the uptime of the service against an SLA target ( % availability <x> nines). We even graph out the application version such that we can see how an application develops in time and how changes in response times line up against version changes. With these screens, our customers are able to understand how their applications are performing. The alerts and graphs help us prevent major problems by early identification of problem areas, allowing us to scale resources as needed and determine whether a version update has unexpectedly introduced a slowdown in performance.