Date of Award
Spring 2004
Document Type
Thesis - Restricted
Degree Name
Master of Science (MS)
Department
Electrical and Computer Engineering
First Advisor
Corliss, George
Second Advisor
Harris, Douglas
Third Advisor
Feng, Xin
Abstract
The goal of the project is to develop a Peer-to-Peer (P2P) Distributed Computing environment, named Octopus, such that we can capitalize unused computer resources to solve computing intensive problems as well as simple problems. The main features of the architecture are 1. Solve large scale problems 2. Capitalized unused computer resources (wired and wireless) 3. Be independent of problem type 4. Scalable 5. Be fault tolerant and available The Octopus is designed to solve a problem by pulling idle resources from the network of computers that are either wired or wireless (mobile). The Octopus architecture is independent of the type of problem. It can solve a large-scale problem that requires a lot of computing resources as well as small requests. For example, when users are not using the desktops in an office, the processing time of those computers can be used to solve a problem that benefits the company. A processing intensive problem is a prototypical problem for this type of computing. For example, a portfolio optimization problem is computing intensive. With the Octopus, we can optimize buys and sells from a universe of stocks that maximizes the return of the portfolio. We solve the problem with idle resources in the enterprise rather than with a dedicated large expensive server. In essence, in a P2P Distributed Computing environment, the data for the problem is sliced into small units of work (work unit) and sent to different computers (volunteers). These volunteers do not know the requestors of work unit or their peers. A publisher, a volunteer, publishes the work unit into a space. Volunteers with idle resources collect a work unit from the space to solve the problem. These volunteers process according to the instructions they receive in the work unit. Once their unit of work is complete, the results are published into the space. The publisher acquires solutions from the space. This environment is like a group of computers working to solve a problem with little knowledge about their peers. We take couple of problems to demonstrate that the prototype of the Octopus architecture. The problems that we chose are finding a stock quote and a Monte Carlo simulation of value of a simple stock portfolio. We choose two different types of problem to demonstrate that the Octopus architecture is truly independent of the type of problems. We have designed, developed, and demonstrated that it is a formidable solution for common business problems by using wired or wireless resources. Our simple demonstration has shown that the Octopus architecture is a viable framework for P2P Distributed Computing. Though the current demonstration version contains a minimal set of function, we made a case that this is a viable product for solving business problem and managing corporate information.
Recommended Citation
Ranjit, Shirish, "Peer-to-Peer Distributed Computing Applied to Prototypical Problems" (2004). Master's Theses (1922-2009) Access restricted to Marquette Campus. 3898.
https://epublications.marquette.edu/theses/3898