Scalable high performance streaming processing application instrumentation framework via improving dynamic concurrency

Abstract

The world has already moved to a highly technological stage and internet-based services plays a vital part of day to day life. Performance of those internet based services is a key factor of quality of the service and developers are forced to develop the best possible performant system. Usually gaining the best possible performance is hard due to low visibility and flexibility of the system in performance improvement phase. This research is focusing on developing the framework ‘concor: A framework for high performance streaming applications, instrumentation in-built’ by combining the pre-placing instrumentation probes and data flow based architectures. The framework provides an API to form data flows, while providing in-built performance monitoring capabilities. Furthermore, the possibility of implementing a dynamic thread reconfiguration mechanism is also researched and included in the framework. Dynamic thread reconfiguration mechanism is used in simplifying the bottleneck isolation. Apart from this, dynamic thread configuration mechanism effectively lifts the initial concurrency design overhead from the developers and provides a new dimension of runtime performance tuning. Keywords: Instrumentation, Concurrency framework, Dynamic concurrency, Runtime performance tuning, dynamically assigned thread pools. Bottleneck identification, data-flow architecture, event streaming.