Self-regulating Message Throughput in Enterprise Messaging Servers – A Feedback Control Solution

Abstract

Abstract-EnterpriseMessaging is a very popular message exchange concept in asynchronous distributed computing environments. The Enterprise Messaging Servers are heavily used in building business critical Enterprise applications such as Internet based Order processing systems, pricing distribution of B2B, geographically dispersed enterprise applications. It is always desirable that Messaging Servers exhibit high performance to meet the Service Level Agreements (SLAs). There are investigations in this area of managing the performance of the distributed computing systems in different ways such as the IT administrators configuring and tuning the Messaging Servers parameters, implement complex conditional programming to handle the workload dynamics. But in practice it is extremely difficult to handle such dynamics of changing workloads in order to meet the performance requirements. Additionally it is challenging to cater to the future resource requirements based on the future workloads. Though there have been attempts to self-regulate the performance of Enterprise Messaging Servers, there is a limited investigation done in exploring feedback control systems theory in managing the Messaging Servers performance. We propose an adaptive control based solution to not only manage the performance of the servers to meet SLAs but also to pro-actively self-regulate the performance such that the Messaging Servers are capable to meet the current and future workloads. We implemented and evaluated our solution and observed that the control theory based solution will improve the performance of Enterprise Messaging Servers significantly.

Self-Managing Performance in Application Servers – Modelling and Data Architecture

“…The current solution is a subset of the bigger research problem that we are trying to address dealing with development of a generic end-to-end framework that enables in creating robust Application Servers that are more adaptive and self-managing in performance management. We have implemented various controllers for the JEE server components with encouraging results [10] [11]112] but most of the implementation is simulation based. We intend to extend our work in validating our theory by running in actual Application Server environments.…”

“…There are solutions applying control systems to self-manage the performance using adaptive controllers. Many of them are specific solutions to manage the performance of web servers [4][5], web services [6], web caching [7][8], EJB Servers [9], JMS Servers [10], JDBC drivers [11] [12]. Most of such solutions are specific to the components in the Enterprise Application Server stack.…”