Ensuring application performance is a never ending task that involves multiple products, features and best practices. There is no one process, feature, or product that does everything. A good place to start is pre-production and production monitoring with both an Application Performance Management (APM) tool and a Unified Monitoring tool.
The APM tool will trace/instrument your application and application server activity and often the end user experience via synthetic transactions. The development team and DevOps folks need this.
The Unified Monitoring tool will monitor the supporting infrastructure. The IT Ops team needs this. DevOps likes it too because it helps make IT Ops more effective, which in turn helps assure application delivery.
More Cost Effective
APM tools do not specialize in infrastructure monitoring like unified monitoring solutions do, and unified monitoring solutions do not provide application monitoring depth and diagnostics like the APM tools do. And on top of that, the different audiences need different information.
The best approach is to buy APM for the most critical applications. Most organizations use APM for 10% - 15% of their applications. It is too expensive to buy it for everything. Then for the second tier applications that need some monitoring, they use the unified monitoring solution. It is much less expensive and if you select one with synthetic transaction capability you can get "good enough" end user experience monitoring to know whether or not the application is performing well or not.
Service-Centric is Key
When it comes to unified monitoring, it is important to understand that most unified monitoring vendors provide endpoint monitoring. With endpoint monitoring alone, it is impossible to provide highly accurate root-cause isolation. And they don't identify which service, or application, is impacted. And they can't tell you the extent of the impact. Is it just at risk without impacting application delivery yet OR is it down OR is it somewhere in between?
Be sure the unified monitoring vendor is service-centric and models relationships between components, and that it identifies root-cause; the service or application impacted; and the extent of the impact. This can save hours when there is an outage.
Better yet, by identifying when services are at risk, this can help you to proactively identify and address issues before services/application delivery is impacted.
Scott Hollis is Director of Product Marketing for Zenoss.