Student Abstracts: Computer Science at LBNL

Developing a High Speed Subnet for Testing Windows 2000 Server and LINUX Red Hat 7.1 operating systems. MARK CARPENAY (Queens College, Flushing, NY 11367) EVERETT HARVEY (Ernest Orlando Lawrence Berkley National Laboratory, Berkley, CA 94720) .
The goal of our project was to test the Windows 2000 server and the Linux Red Hat 7.1 operating systems for stability and reliability, using an isolated subnet. The subnet was configured using four PCs. The first machine was configured as a server to boot the Windows 2000 Server and Linux Red Hat 7.1 operating systems. The second PC was configured to be a LINUX server and the other two machines were configured to be windows 98 clients. The clients were used to test the various features that were installed on the on the servers. Some of the features installed and tested on the servers were Domain Name Services and Routing. The Windows 2000 server was configured to be a Domain Controller and Active Directory, enabled to test the latest securities and scalability features of Windows 2000. This machine was also configured to be a DNS server and a Router. DNS servers are used to maintain records for domains they host and respond to queries for a given host name with the IP address stored in the DNS database for that host. This machine was also configured to be a router to link our subnet to the Lawrence Berkeley Laboratory Network. The design and building of subnets are important for maintaining large efficient networks, of which the current largest network being the internet. In conclusion we found the LINUX operating system to be more stable and reliable than Windows 2000 Server and thus better suited for high end networks with greater traffic.

Using Java to Process Streams of Performance Event Data. JENNY YJ CHUU (Las Positas, Livermore, CA 94550) DAN GUNTER (Ernest Orlando Lawrence Berkley National Laboratory, Berkley, CA 94720) .
I have participated the summer project at the Data Intensive Distributed Computing (DIDC) Group, Distributed Systems Department at NERSC, LBNL. The DIDC group has begun to develop an archive for performance monitoring data, such as application logs from NetLogger or ping and netstat data from JAMM (Java Agents for Monitoring and Management) or Enable. The DIDC group is also working on a language-independent publish/subscribe protocol to access this data that can perform real-time filtering on high-throughput streams. The main focus of my summer project is to learn Java I/O coding related to the monitoring network performance. Examples of using java.io package and the File, OutputStream, InputStream, Writer, Reader classes for stream input/output operations and processing files are described. Also demonstrate how to use java exceptions when try and catch blocks signal some particularly unusual events in the program that deserves special attention. In addition, the potential interests of this technology to the biological applications are discussed.

Analysis, and Implementation of an Online Research Document Management System. GIRISH GHATIKAR (California State University, Hayward, Hayward, CA 94542) DR. JAMES MCMAHON (Ernest Orlando Lawrence Berkley National Laboratory, Berkley, CA 94720) .
The research project involves coming up with a solution for a web-based (online) document management system that lets you easily store, access, and search the Energy Efficiency Standards Groups' Research documents collected for twenty years in a secure work environment, both at micro and macro level. This also allows a designated user on any system to post and retrieve information in certain commonly used format(s). Text, scanned images, Microsoft Office documents, web links, etc. can be managed over the Web. Users have the power of the information access, ease of use, and electronic storage on the Web at their disposal. The most suitable implementation was assessed after a thorough investigation and analysis pertaining to vendors. The existing infrastructure of the division formed the crux of the problem, and thus to come up with an answer that is viable and cost effective as well. With this solution, sharing information is as easy as storing it on a hard drive of a computer, and finding it is as easy as browsing the Web, where one has the access from virtually anywhere. This becomes very important, considering the future energy analysis and comparisons based on the former research and analysis.

Adapting a Nonlinear Equation Solver to Scale the Residual Functions Dynamically. MEGAN MCCLEAN (University of California, Berkeley, Berkeley, CA 94709) DR. DAVID LORENZETTI (Ernest Orlando Lawrence Berkley National Laboratory, Berkley, CA 94720) .
Newton-Raphson is an iterative method for solving nonlinear problems. It begins with an initial guess at the solution, and then generates a sequence of points that step increasingly close to the real solution. When the initial guess is far from the solution, the Newton-Raphson method may diverge. Descent methods are used to control divergence. A descent method requires that each step reduce some measure of residual magnitude, commonly the sum of the squared residuals. Preliminary numerical tests showed that the global convergence of the Newton-Raphson method can be improved by weighting the residuals with weights that incorporate sufficient information from the residual models at each iteration. The goal of our research was to reproduce these results by modifying an existing nonlinear equation solver to dynamically scale the residuals using the five weighting formulas employed in the previous study. We modified TENSOLVE to update the weights for the residual functions at each iteration. The algorithm was tested on 14 problems, with and without the weighting. Weighting the residuals neither improved the number of problems solved, nor decreased the number of Jacobian evaluations required to solve problems. Our results are inconsistent with those of the previous study. The previous study showed a reduction in the number of Jacobian evaluations for all five weighting formulas, and an increase in the number of problems solved successfully for four of the five weighing formulas. Future research will investigate why results from the previous study, and our study using the modified TENSOLVE algorithm, do not agree.

AN INTEGRATED MODELING SYSTEM TO STUDY THE IMPACTS OF CLIMATE VARIABILITY ON WATER RESOURCES IN THE SAN JOAQUIN BASIN, CALIFORNIA. PALLAVI RAMARAJU (Contra Costa College, San Pablo, CA 94806) DR. NIGEL W.T. QUINN (Ernest Orlando Lawrence Berkley National Laboratory, Berkley, CA 94720) .
Water resource planners need to develop contingency plans to deal with the potential impacts of climate variability and changes in the frequency and magnitude of extreme weather events in the San Joaquin Basin. Studies suggest that warmer winter storms, earlier runoff from the Sierra snowpack, and reduced summertime flow in the tributary streams could adversely affect the water supply, water quality and agriculture production. Planning studies involving suites of complex mathematical models are often compromised owing to the inordinate amount of time devoted to data processing as the output from model is manipulated to become the input to the next in sequence. Hence the objective of this research is to develop an integrated modeling system using Object User Interface (OUI), a software package developed by the U.S. Geological Survey. OUI, a map based interface, provides an environment for efficient database/model integration, aids in map-based communication with databases, and offers controls for model execution. OUI also has tools that provide for the graphical and statistical analysis of the results. Successful linkages of various water resource management models, newly developed within State and Federal water agencies can assist in the management of water quality, water supply and agriculture production in the San Joaquin Basin and Bay-Delta. The model system will also aid analysts in performing vulnerability analysis and suggesting management strategies for mitigating the impacts of increased climate variability and more frequent extreme weather conditions hence reducing the vulnerability of the existing system to permanent damage.

Creating a High Speed Subnet Behind Firewalls. CHAKAMEH ZAHEDKARGARAN (Contra Costa College, San Pablo, CA 94806) DR. EVERETT HARVEY (Ernest Orlando Lawrence Berkley National Laboratory, Berkley, CA 94720) .
The recent surge of viruses and hacker attacks has increased the necessity for making the computers at Lawrence Berkeley lab more secure. The goal of my group was to design and implement a firewall scheme that improves the system security. For this purpose we chose to use both Linux and Windows 2000 server operating systems. Since our experiment had to be isolated from the LBL domain, we created our own private subnet. We configured both a Linux and a Windows 2000 machine to serve as a DNS (Domain Name System) server and a router and later we implemented our firewall scheme on them. However, the Linux operating system was preferred for this project because of its flexibility and the unlimited authority it provides for the administrator. The Domain Name Services (DNS) is a distributed Internet directory service. DNS is used mostly to translate between domain names and IP addresses, and to control Internet email delivery. If DNS fails, web sites cannot be located and email delivery stalls. A routing configuration allows the packages from the Local Area Network (LAN) to be forwarded to the Internet depending on the firewall permissions. To enable routing and firewalls on Linux machine, I updated and recompiled the kernel with the packet filtering and routing options enabled. Then I wrote a script of appropriate IPCHAINS commands that contains the permission and denial of package forwarding between different domains and ports, and create barriers in order to prevent unauthorized access to our network.