Nicholas Anastasi, A Decentralized Approach to Fast Allocation of Shared e-Services
Andrew R. Freed, A Theoretical Design for University Department Web Pages Based on a Task Analysis
Cassandra Gekas, Analysing Group Work
George Kroner, Interactive Information Seeking through the use of an Information Retrieval Software Agent
Mohammad Mirza, Trends in E-Commerce Searching
Chris Stubbs, Self and Team Evaluation Project
Darin Tranquillo, Modeling an Information Architecture
working with Dr. Jonghun Park and assistance from Ki-Seok Choi (Georgia Institute of Technology)
A Decentralized Approach to Fast Allocation of Shared e-Services
With all the emerging e-services, there are an enormous amounts of building blocks available for future e-business applications. These applications are able to dynamically discover and make use of existing e-services, and components thereof which results in more valuable, aggregate services. One of the main problems in such systems is the service co-allocation when the composite services begin use of (execution) or on-the-fly reconfiguring of remote components. As part of the fellowship, a simulation was developed which would be used to test multiple protocols for distributed allocation of resources. The protocols are as follows:
Protocol 1. A simplistic all-or-nothing approach, whereby a process will initially request all resources that it needs. If any respond with a negative response, cancel messages are multicast, and the process will sleep for a certain time, as determined by the backoff formula and will initiate the same request upon wakeup.
Protocol 2. Processes employing Protocol 2 works in a step-by-step manner, allocating resources one at a time until it satisfies it's randomly-picked goal state. If at any point, the process receives a negative response from the resource it is currently requesting, it will merely sleep and retry upon wakeup.
Protocol 3. This protocol, unlike the previous two, is lease-based. Lease time is generated based on a current snapshot taken by the resource manager, which controls the states of all it's children resources. Processes have a limited amount of time to allocate and commit to the usage of all necessary resources, otherwise they sleep and retry upon wakeup. Protocol 3 is designed to allow for the fastes possible allocation of resources while eliminating the possibility of deadlock, through the use of expired leases.
Protocol 4. While all the other protocols prevent deadlock through some means, protocol 4 is the only protocol that prevents both livelock and deadlock. Protocol 4 is also the only protocol which will guarantee a maximum arrival time. This protocol employs the same allocation algorithm as protocol 3, but is extended to use a "safety check" which is the mechanism by which livelock and deadlock are prevented.
The results will include thousands of trials with different parameters. It is expected that protocol 3, after fine tuning the backoff and leasing formulae, will outperform the three other protocols. The main consideration for this will be average waiting time, but percent resource usage and number of outgoing messages will also be taken into account in evaluating performance. Also, it is expected that protocol 4, as well as protocol 1 should outperform protocol 2.
Software specification was the output.
Andrew R. Freed arf132@psu.edu working with Frank Ritter
A Theoretical Design for University Department Web Pages Based on a Task Analysis
University department's web pages are the focal point for a wide variety of users, including prospective students, parents, alumni, and current students. We provide a task analysis of what users look for on university department websites. Our task analysis starts to specify what each user group will be looking for on these sites. Later work can use this task analysis to consider what pages of the site the user will use to accomplish their goals and if this is even possible. Checking a site to ensure that common user's tasks can be performed will help ensure completeness of a site. Further reflection on these tasks can enhance the clarity and consistency of the procedures used when providing navigational aids. Good task analyses are often reusable as a design aid. This analysis of university web sites can be viewed more generally for other web sites, such as a corporation, non-profit, an e-store, and so on.
This project resulted in a tech report and shortly, a journal article submission, and an ongoing research relationship.
Cassandra Gekas ceg147@psu.edu, working with Michael McNeese
Analysing Group Work
The work that I did concentrated on two distinct areas. The first was a cognitive analysis of the problem solving modules Jasper and Repsaj. We input and analyzed raw data from student groups as they progressed through the problem solving process. The goal was to search for patterns and methods in the group approach as they discovered the deep structure of the problem. In particular, we were interested in how working in distributed teams, using groupware and chat rooms, effected overall communication and the problem solving process. Funded by a grant from the Navy, this research has eventual implications to the formation of multi-national peacekeeping teams. They second project, run by the undergraduate team was to deconstruct the Jasper and Repsaj modules using abstraction hierarchy and concept mapping. Working both as a team and on an individual level, we collected information useful in developing hidden profiles, which will facilitate the next level of experimentation, running in the fall.
George Kroner gik103@psu.edu working with Jim Jansen
Interactive Information Seeking through the use of an Information Retrieval Software Agent
[also available as Jansen, B. J. and Kroner, G. (2003). The Impact of Automated Assistance on the Information Retrieval Process. The ACM SIGCHI 2003 Conference on Human Factors in Computing Systems. p. 1004 - 1006. Fort Lauderdale, Florida, 5 - 10 April 2003.]
Locating information pertinent to one's needs is becoming a progressively daunting task. With more and more searchable data becoming available every day, the use of increasingly sophisticated information retrieval systems becomes necessary. Of primary interest to Dr. Jansen is the design, functionality, usability, and robustness of information retrieval systems (such as Internet search engines). For his doctoral dissertation, Dr. Jansen designed a software agent for the performance enhancement of an information retrieval system known as the Agent Improved Information Retrieval System (or AI2RS).
AI2RS works by capturing a user query, making structural and grammatical corrections to the query, sending the corrected query to a search engine, retrieving the results from the search engine, and then making recommendations on how the search could be improved (see Figure 1). It was originally programmed in the Tcl/Tk programming language for use on a UNIX-based system. The main objective of this project was to successfully port the AI2RS algorithm into the JavaScript programming language (and accompanying functionality as necessary) for implementation in a Microsoft Windows environment using the Internet Explorer web browser.
Work first began with the coding of a special application known as a "wrapper." The purpose of the wrapper is to record user interaction with the web browser such as the viewing, saving, bookmarking, or printing of a web page. This application was successfully coded in Microsoft's Visual BASIC programming language. Work then progressed with the porting of the AI2RS algorithm from Tcl/Tk into JavaScript.The spell checking functionality of AI2RS such as has been suitably altered to use the Microsoft Word spell checking engine. It is hoped that the Microsoft Word thesaurus engine can be used to provide words related to those of a user's query.
Future work on this project will continue with the implementation of AI2RS on a Microsoft Windows based computer running Microsoft's Internet Information Services web server. Planned testing of AI2RS will be done utilizing the Text Research Collection (TREC) document collection.
This has resulted in a conference paper (CHI 2003: "The Impact of Automated Assistance on the Information Retrieval Process") and an ongoing research relationship.
Mohammad Mirza mam1003@psu.edu working with Amanda Spink
Trends in E-Commerce Searching
Web queries are an important mean to analyze trends in users' Web searching behavior. The analysis of Web users' queries is useful for the design of more effective search engine interfaces and systems, and the development of niche business Web search engines. Several data sets of user queries from the Excite (U.S. based) and FAST (European based) search engines were examined to isolate e-commerce characteristics and compared with previous results from U.S.-based Web search engine users. E-commerce queries were classified into various categories to compare among the data sets and various statistical calculations were done to find out the major differences among the European and US based queries. An initial conference paper is in progress based on the findings and analysis of these data sets. It will be followed by a journal article that will detail the findings of the research.
This has resulted in a conference paper (submitted) and a journal article in preparation.
Chris Stubbs Stubbers26@aol.com working with Keith Bailey
Self and Team Evaluation Project
Our SURF 2002 project was to overhaul the self and team evaluation process currently employed in IST courses. Our first task was to design and develop the database to store the self and team evaluations. The remainder of our project was broken into two phases. Phase 1 saw the development of a student side website. The student site will be used to submit evaluations.
Phase 2 of the project was to design to administrative site that would support the self and team evaluation process. This site will be used by instructors to create teams, enroll students, add students to team, create projects, and view student scores, among other things.
The system is scheduled for a pilot test during the end of the Governorís School program (end of July, 2002), with a complete system rollout expected for the start of the fall 2002 semester.
Though version one of the project is nearing completion as the abstract is written, we have several enhancement ideas for the future. We plan to develop a customizable evaluation form, to allow instructors to create their own personal evaluations. We hope that such as addition will expand the use of our tool beyond the IST department. We also may proceed to a phase 3 of our project, and develop a ìsynchableî version of the evaluation, so that students can synch their evaluations into the database, rather than submit them over the internet.
We hope that this project will aid both instructors and students in the evaluation process.
This has resulted in software.
Darin Tranquillo dtt112@psu.edu working with Steve Haynes
Modeling an Information Architecture
The School of Information Sciences and Technology's Information Architecture Project is currently working on modeling its Information Architecture (IA). An Information Architecture, as defined in the Project Mission, is "three separate layers of information management: the technology layer comprising computing and networking hardware and software, the content layer comprising logical and physical data models, and the administration layer comprising defined processes for managing information as it flows into, out of, and within the School".
The IA Project Committee developed a set of UML models that portray both the School's information and the methods used for accessing that information. The models were first broken down into specific packages. Each package then contained a unique set of class, use-case and activity diagrams. The models were maintained and updated with any changes suggested during the IA Project meetings.
Our main research for the IA Project has been to develop an online requirements gathering mechanism. By this we mean that the requirements for the Information Architecture will be gathered through the online viewing of the IA models. The models will be accessible via the School's intranet as well as the Internet. In addition, users will have the ability to post comments on the models which will create a dynamic environment for the IA models.
Currently, the website and the online requirements gathering mechanism is being developed. The overall design for the website has been completed and work on implementing the mechanism has begun. Once the gathering mechanism is fully implemented, the original IA models developed by the IA Project Committee will become an evolving set of requirements for the School of Information Sciences and Technology's Information Architecture.
This has resulted in some software, and an ongoing research relationship.