A lightweight method for analysing performance dependencies between services

Abstract

For many applications, performance is paramount. For example, to improve conversion rates in e-commerce applications or to comply with service level agreements. Current trends in enterprise level architecture focus on designing and orchestrating services. These services are typically designed to be functionally isolated from each other up to a certain degree. During the design phase as well as when the application is deployed, choices have to be made how services interact and where they need to be deployed. These choices have a profound impact on the responsiveness of an application as well as on which performance can be made. In this paper we propose a methodology to describe and analyse performance dependencies between services. The resulting model can then be used to assist in designing a service oriented architecture and improving existing solutions by pointing out performance dependencies of services.