Adapting eFinanced web server farms to changing market demands

Abstract

The evolution of the eFinance Market presents new and changing requirements on web server farm architecture. Web server farms built during the boom years of the web were designed to provide service to a very large number of users, where each request however placed relatively little load on the system. Further, the requests displayed a large amount of statistical similarity, so that caching mechanisms could be successfully applied. However, as professional eFi- nance tools migrate to web-based ASP technology, and as different companies cooperate to build these tools using the Web Services paradigm, these basic assumptions no longer hold: The number of users decreases, while the demands placed by each user increases. Further, each user places highly specialized demands on the application, decreasing the similarities between different requests. The success of conventional caching techniques drops even further if the web server farm is now called upon to provide XML-based Web Services instead of HTML web pages. This paper analyzes these new requirements on the basis of a case study: We present the design process behind the addition of an On The Fly Calculation Server to IS Innovative Software's Web Server Farm architecture. This new server has been developed to provide advanced financial (MPT) calculations to professional users. The resulting design represents a form of intelligent memory, optimized to minimize the movement of data and thus request latency. The stages in the development of this new server, and the integration of this server into the existing web server farm, are presented.