Towards an architecture for real-time decision support systems: challenges and solutions

Abstract

In large enterprises, huge volumes of data are generated and consumed, and substantial fractions of the data change rapidly. Business managers need up-to-date information to make timely and sound business decisions. Unfortunately, conventional decision support systems do not provide the low latencies needed for decision making in this rapidly changing environment. The paper introduces the notion of real time decision support systems. It distills the requirements of such systems from two real-life IT outsourcing examples drawn from our extensive experience in developing and deploying such systems. We argue that real time decision support systems are complex because they must combine elements of several different types of technologies: enterprise integration real time systems, workflow systems, knowledge management, and data warehousing and data mining. We then describe an approach to addressing these challenges. The approach is based on the message brokering paradigm for enterprise integration, and combines this paradigm with workflow management, knowledge management, and dynamic data warehousing and analysis. We conclude with lessons learnt from building systems based on this architectural approach, and discuss some hard research problems that arise.