|
|||||||||||||||
 |
EventsGraduate SeminarFall 2003 ScheduleLocation: SAGE 4101
AbstractsHouda Lamehamedi - Proactive Replica Management in Data Grids Abstract: Grids offer an integrated architecture that enables and coordinates the sharing and aggregation of geographically distributed resources in order to solve large-scale problems. As part of the Grid Architecture, a Data Grid connects large collections of distributed computers and storage resources to facilitate sharing of data. Data Grids provide the infrastructure to support large-scale data-intensive applications that generate large data sets. However, ensuring efficient access to such huge and widely distributed data is hindered by the high latencies of the Internet. We address these challenges by employing intelligent replication and caching of objects at strategic locations. In our approach, replication decisions are based on a cost-estimation model and driven by the estimation of the data access gains and the replica's creation and maintenance costs. These costs are in turn based on factors such as runtime accumulated read/write statistics, network latency, bandwidth, and replica size. To support large numbers of users who continuously change their data and processing needs, we introduce scalable replica distribution topologies that adapt replica placement to meet these needs. Based on the Open Grid Service Architecture (OGSA) we introduce the Proactive Replica Management Services (PRMS). This design approach supports high-level services that can be implemented on top of the existing low-level Grid services. In this talk I will give an overview of the emerging Grid computing standards and architecture, and the Grid toolkit that our work is based on: the Globus toolkit. I will then present PRMS, describe our approach to deploying dynamic replication in Data Grids, and describe the analytical model of our system. I will conclude by presenting the design concepts and the results of simulated dynamic replication techniques. Amir Youssefi - Visual Web Mining Abstract: Researchers analyzing web sites face two challenges of the increasingly large amount of data published online as well as web sites with more complex structures. In our work, Data Mining and Visualization techniques are used in the web domain in order to benefit from both the power of computing and the perception of human vision toward Visual Web Mining. In response to the two aforementioned challenges, a generic framework is proposed. Under this framework, we apply Data Mining techniques to large datasets and use Information Visualization methods on the results. The goal is to compare the outcomes of mining Web Usage Logs and extracted Web Structure by visually superimposing results. We have designed and implemented interactive 3D visualizations to amplify User Access Patterns on websites. Latest results are visualization of web usage logs on top of a web graph with about 500,000 nodes. Web logs of http://www.cs.rpi.edu are used for analysis as well. For more information please see project homepage Co-authors of the ICDM paper are: Prof. Duke (Bath,UK), Prof. Zaki(RPI) and Prof. Glinert (NSF,RPI). Aytekin Vargun - Code-Carrying Proofs: A New Approach to Safety, Security, and Functional Correctness of Mobile Code Abstract: How can computer systems be protected from bad code, which through either malicious intent or programmer error, would subject them to harmful effects? Among several protection schemes in current use or under development, the ones that come with the strongest guarantees are those that insist on a proof-of-correctness policy: don't accept delivery of new code unless it is accompanied by a formal proof of its correctness that can be checked by the consumer. The main manifestation of this proof-based approach in current research is called Proof Carrying Code (PCC). An alternative approach we are working on is Code Carrying Proofs (CCP). In this talk I will describe both PCC and CCP, and discuss some running sample CCP code we implemented on Athena programming Language. Joel Branch - An Introduction to Steganography Abstract: Steganography is the art and science of hiding communication; information can be embedded in data such that the original data does not appear corrupted in any way, failing to attract an eavesdropper's attention. For example, hidden messages within a JPEG file. This talk will be based off of an article published in IEEE Security & Privacy, and perhaps some other sources. The presentation will cover some existing steganographic systems and detection via statistical means, and some methods of getting around selected detection algorithms. Mayuresh Kulkarni - Analog Computation Abstract: We describe analog devices to solve problems usually associated with digital computers. These devices are very different from traditional analog devices (like the differential analyser). It is interesting to look at such idealized physical devices and ask what common problems like prime number detection, sorting, testing for irrationality etc. can be solved by building such devices. Karlton Sequeira - Improving Spatial Locality of Programs via Data Mining Abstract: In most computer systems, page fault rate is currently minimized by generic page replacement algorithms which try to model the temporal locality in programs. In this presentation, we propose an architecture for program-specific code restructuring to increase spatial locality within a program. We propose two algorithms, one greedy and the other stochastic, for increasing spatial locality. Each algorithm effectively decreases average working set size and the page fault rate. We also present evidence that this method is more effective than the traditional approaches due to use of domain information. Also, the restructuring is done offline, so as to cause no overhead to runtime characteristics. Finally, we illustrate the efficacy of our algorithms on actual data mining algorithms. |
|
|||||||||||||
News
