Current Projects
ECEB Faculty have accumulated significant experience in e-business and areas related to the analysis, design and operation of e-business and e-government sites. This section provides a summary of several projects relevant to E-business.
Knowledge Sifter | Autonomic Computing | Agent Based E-Business | Analytic Modeling of E-Business Site Performance and Scalability | Business-oriented Resource Allocation Policies |Characterization of E-commerce Workloads | Dynamic Quality of Service (QoS) Control of E-Business Sites |Intelligent Ontology-based Search for Web Mining | Methodologies for Analysis, Design and Implementation of E-Business Systems | Performance Measurement of E-Business Sites | Scalable Architectures for the Web and Distributed Secure Systems |Secure Objects in E-Business Environments |
Knowledge Sifter: Agent-Based Search over Heterogeneous Sources Using Semantic Web Services.
Knowledge Sifter is a scaleable agent-based system that supports access to heterogeneous information resources such as the Web, open-source repositories, XML-databases and the emerging Semantic Web. The Knowledge Sifter agent architecture is depicted in the figure below. The architecture has three layers: User Layer, Knowledge Management Layer, and Data Layer. The User Agent interacts with the user to elicit user preferences that are managed by the Preferences Agent. These preferences include the relative importance attributed to terms used to pose queries, the perceived authoritativeness of Web search engine results, and other preferences to be used by the Integration Agent. The user indicates an initial query to the Query Formulation Agent. This agent, in turn, consults the Ontology Agent to refine or generalize the query based on the semantic mediation provided by the ontology services available. The Ontology Agent uses a conceptual model for the domain by means of the Web Ontology Language (OWL) schema specification of the Imagery Domain Model (based on ISO 19115 and 19139); we plan to use model link the complementary ontologies such as the temporal, spatial, feature, etc. The Knowledge Sifter architecture is general and modular so that new ontologies and new information resources can be easily incorporated. The Web Services Agent uses domain knowledge regarding the data sources, such as QoS attributes, source authoritativeness, and image sizes, to optimize the execution of subqueries. The Integration Agent is responsible for compiling the sub-query results from the various sources, ranking them according to user preferences supplied by the Preferences Agent.
Knowledge Sifter Three-Tiered Architecture
Knowledge Sifter project is sponsored by a NURI from the National Geospatial-Intelligence Agency (NGA) through their Innovision Program.
Drs. Kerschberg and Menascé are Co-PIs on this project. Mr. Hanjo Jeong is the Chief Programmer for Knowledge Sifter with assistance provided by Scott Mitchell.
For more information about Knowledge Sifter, please visit the Publications Page.
Autonomic Computing
Dr. Daniel Menascé and his students have been designing techniques to build autonomic computing systems since 1999. Their methods are based on combining analytic queuing models with combinatorial search techniques. They have employed their techniques to self-optimizing and self-tuning e-commerce systems, web servers, and to dynamic resource allocation in Internet Data Centers. Their work was the first along these lines with the first publications dating from 2001.
The picture below shows the basic architecture of the QoS controller designed and built at the ECEB and the following diagram shows that the QoS controller is capable of maintaining the QoS of an e-commerce site even in the presence of an increase in the workload intensity.
Quality of Service (QoS) Controller in Feedback Control System
Performance Improvement of Controled versus Uncontrolled QoS
Agent Based E-Business.
Kerschberg and
colleagues perform research in the use of agent-based architectures
and cooperative information agents to support e-business processes, to manage
workflow and to access and integrate information from multiple sources,
including the World Wide Web. Recently,
we developed an agency-based approach to the optimal negotiation and
sharing of resources for systems having to satisfy both local and global
constraints. We are presently
implementing a Distributed Agent Resource Protocol and associated communication
architecture that would be applicable to the optimal sharing of resources in
business-to-business marketplaces. 
In a DARPA-funded project on the Intelligent Integration of Information, Kerschberg and Gomaa developed a Knowledge Rover architecture consisting of a configurable family of agents that proactively search for information on the World Wide Web, retrieve the information, and integrate it into a resource repository. We have also studied the use of agents to maintain approximate consistency of active views defined over multiple databases.
Analytic Modeling of E-Business Site Performance and Scalability.
A large number of e-business sites suffer from problems of scalability and slowness. E-commerce sites in the US lost $48 billion in sales 1998 due to slowness of e-commerce sites. Menascé has been working for several years in the development of analytic models of performance of e-commerce sites. His work can be found in many publications and book. He has successfully applied these techniques to many large e-commerce sites.
Business-oriented Resource Allocation Policies.
Resources (e.g., CPU, disk, and network) of an
e-commerce site should be allocated in a way that optimizes business-oriented
metrics such as revenue throughput, measured in $/sec. Menascé et al developed a family of policies that can be used to
increase the revenue throughput of e-commerce sites. Studies showed increases in
revenue throughput of up to 43% (see figure).
Characterization of E-Commerce Workloads.
Menascé et al developed a graph model to characterize e-business workloads and capture user behavior as they navigate through an e-business site (see example in figure). This model, called Customer Behavior Model Graph (CBMG), provides interesting metrics such as buy to visit ratio, average session length, and average number of visits per session. By analyzing web logs and clustering sessions into groups of similar CBMGs, one can discover groups of customers that exhibit similar behavior (e.g., heavy buyers, occasional buyers).
Dynamic Quality of Service (QoS) Control of E-Commerce Sites.
The Quality of Service (QoS) of e-commerce sites plays a crucial role in attracting and retaining customers. Frustrated customers leave these sites and do not return, causing revenue to be lost. The performance of an e-commerce site is a function of a large number of parameters such as the number of threads at each level, the maximum number of connections accepted, the maximum number of requests served by each thread, cache sizes, cache replacement policies and parameters, as well as load balancing policies and parameters. The workload experienced by these sites tends to vary in a very dynamic way.
The complexity of e-commerce combined with the large short-term variations of the workload calls for automated methods for site configuration. This project deals with methods for dynamically tuning e-commerce sites so that desired QoS levels are attained. Our approach uses heuristic optimization techniques (e.g., hill- climbing and tabu search) with analytic queuing models to guide the search for the best combination of parameters. We validated our approach in an experimental setting by comparing the QoS levels of a TPC-W compliant e-commerce site with and without control. We showed that under increasing loads, the controlled system meets its QoS goals, while the uncontrolled fails to do so.
Intelligent Ontology-based Search for Web Mining.
This project addresses the problem of specifying Web searches and retrieving, filtering, and rating Web pages so as to improve the relevance and quality of hits, based on the user’s search intent and preferences. We present a methodology and architecture for an agent-based system, called WebSifter II, that captures the semantics of a user’s decision-oriented search intent, transforms the semantic query into target queries for existing search engines, and then ranks the resulting page hits according to a user-specified weighted-rating scheme. Users create personalized search taxonomies via our Weighted Semantic-Taxonomy Tree. Consulting a Web taxonomy agent such as Wordnet helps refine the terms in the tree. The concepts represented in the tree are then transformed into a collection of queries processed by existing search engines. Each returned page is rated according to user-specified preferences such as semantic relevance, syntactic relevance, categorical match, page popularity and authority/hub rating.
The WebSifter II prototype has been implemented in the Java Programming Language. Papers dealing with WebSifter can be found in the Publications page.
Methodologies for Analysis, Design and Implementation of E-Business Systems.
Kerschberg, Menascé and Gomaa have extensive experience in domain engineering, having designed the data and information architecture for GMU Independent Architecture Study for NASA’s Earth Observing System Data and Information System (EOSDIS). They recommended a federated client-server architecture for EOSDIS; this is currently being implemented by NASA. We will apply domain modeling and engineering techniques to the application domain of e-business. In particular, we plan to develop an e-business domain architecture from which architectural templates would be generated for specific e-business applications.
TOP
Performance Measurement of E-Business Sites.
One of the challenges in measuring and analyzing the performance of transactions submitted to e-business sites is that many different servers are involved in processing a single transaction. Performance monitors at each individual server provide local measurements only but do not explain the breakdown of the response time of a single transaction across multiple servers. Menascé participated in the design of PROFIT, a prototype tool for measuring the performance of large complex e-business sites.
Scalable Architectures for the Web and Distributed Secure Systems.
Menascé has been involved in the design of new architectures to help scale Web sites through the use of caching and prefetching and e-representatives. He is also involved with designing scalable authentication protocols for very large-scale distributed systems.
Secure Objects in E-Business Environments.
This project is investigating how to implement e-business security at the granularity of business objects. The approach has direct bearing on Digital Rights Management strategies. Our approach consists of a collection of agent-based brokers that implement business rules associated with their particular role within the enterprise. The brokers interact to support the business processes of the enterprise.
The figure below shows a conceptual Imagery Marketplace where customers may browse catalogs, place orders, and accept delivery of image products. Additionally, image suppliers may accept orders for special products, deliver products and metadata to the marketplace, and receive payment for services rendered. The goal is to perform these activities with the assurance that intellectual property is protected and information security is enforced.
Updated: Saturday, February 4, 2006