Electronic Commerce Networks (E2)
Project PI: Andrew Whinston
PROJECT DESCRIPTION
Creating and operating a digital economy is fast becoming a reality. It is expected that several hundred billion dollars per year of electronic transactions will take place by the start of the next century. The project will develop mechanisms for management of the network infrastructure that will enable effective commerce to take place and facilitate innovative applications exploiting the infrastructure. The Center for Research in Electronic Commerce has conducted pioneering activities in this area including the development of resource management mechanisms using priority pricing of network resources and estimation of response times from available data. While the ideas are based on concepts from economic theory the computational methods exploit an understanding of the transient behavior of public networks. The project seeks to extend the mechanisms previously demonstrated in simulation testing to real prototype networks. These mechanisms are characterized by their incentive compatibility and the decentralized resource management in real time. Techniques have been developed to estimate performance parameters over the network, and to update prices at each node of the network in real time based on such available information.
TECHNICAL CHALLENGE
The project addresses the larger issues of operationalizing any proposed mechanism. The decentralized real time price formulation rules we employed has been designed to be computationally manageable. The challenge in a prototype network is to enable running complex applications over the network in an efficient manner, and with reliable performance guarantees. The enhanced computing power, the predictable performance in real time, and the distributed nature of applications will be crucial determinants in realizing electronic commerce applications, for instance, in solving a supply chain management problem. Visions of electronic-commerce-transacted networked organizations have been hindered by the excessive congestion over networks and their unreliability. By charging prices that reflect the traffic at each node our mechanism manages congestion and provides predictable performance guarantees. Since a real time network does not support theoretical equilibria proposed by economists, our mechanisms are developed to handle non-equilibrium behavior.
IMPACT
The work will demonstrate the feasibility of managing resources over a computing network in real time, in an efficient manner, to achieve reliable performance levels. Further, the mechanisms used will be shown to be scaleable to larger systems, with no computational overheads. This will directly demonstrate that distributed computing models for complex tasks and networked execution of critical business processes can both be feasibly achieved over public networks. A computationally complex application (such as a Supply Chain Management solution) can be executed at much lesser cost over such a managed network of Intel systems.
EQUIPMENT
A network of Pentium processor based computers to execute distributed application will be displayed, while the usage is monitored and priced by software implementations of the pricing formulate. As a first prototype network, a cluster of 12 workstations and 2 Quad servers are to be used.
RESOURCES
This project is to be carried out at the Center for Electronic Commerce. The research will be an interdisciplinary initiative, with faculty from the departments of Management Science and Information Systems, Economics and Computer Science collaborating. Previous and ongoing work on developing the theoretical framework and testing it out by simulations has been supported by grants from NSF and the Advanced Technology Program of the State of Texas.
BENEFITS TO INTEL
An enterprise-wide application such as the Supply Chain Management problem is computationally complex, distributed in nature, and requires real time solutions, typically handled by high-end compute servers. Deployment of efficient network resource management mechanism over the connected PCs to solve such a problem is proposed. This will demonstrate that properly managed, a simple network of Intel PCs can pack the computing power and reliability of high end servers and will be capable of enabling Electronic Commerce with a broad reach.
RELATED WEB SITES
Follow this link to the Texas Business School Enterprise Systems and Management Web sites.
Also, visit the Austin American Statesman for a related story entitled: Virtual reality check.