Gaining visibility to application performance is key. Application Performance Management (APM) solutions are not new and provide insight to tiers within an application stack. With the entry of cloud based computing in the past couple of years, the APM world got a bit more complex.
APM is mature enough to consider cloud-based providers in the application stack. In the classic model, an application has three layers in the stack: 1) Database layer, 2), Application layer and 3) Web layer. Depending on the complexity of the application, it may have 5 or more layers in the mix. Today, a cloud service provider may serve one or more of these layers.
Several solutions exist that support cloud-based APM. New Relic, OPNET, and CA are just a few examples. At the Under the Radar conference, Tracelytics presented their approach to APM. Tracelytics started two years ago by a small team of three to address a growing problem they observed in research from Brown University. I met with Spiros Eliopoulos, Co-Founder and CTO to discuss how Tracelytic’s approach differs from the competition.
So, what’s different? Bottom line: It has to do with the flexibility of the solution. As the application stack gets increasingly more complex, so does the management. The number of providers and shared resources is growing exponentially. According to Spiros, their solution “looks at each layer individually, then ties together the different layers to provide a complete view.” Tracelytics allows APM visibility through “drilldown performance across layers.” Their clever approach uses heat maps to visually find problem spots. Managing APM within layers and up/down the entire stack is key to providing clear visibility to correct problem areas quickly.
Many providers struggle with pricing strategies in today’s cloud and virtualized world. In the traditional computing world, it was easy to license solutions. Tracelytic’s approach continues to provide flexibility by focusing on the tracing volume rather than hosts or layers. The entire stack of an application is considered one application. So, whether you engage one application to report 10x per hour or 10 applications once per hour, the cost is the same. This is true regardless of the number of layers within the application stack. Nice!