Dependable Virtual Networks

 

Abstract

The design of reconfigurable high-speed computer networks based on COTS represents a considerable technical challenge; these networks should be inexpensive, scalable and dependable and are currently assembled by utilizing large number of complex hardware and software components that are heterogeneous and subject to continuous upgrade, replacement, and scaling. These networks are too complex to be formally modeled; moreover, the reduction of the failure rates of the individual components may not substantially reduce the failure rate of the overall system due to unexpected interactions between the components. One may consider that failures are inevitable. The goal of this research is to specify, implement and validate a novel method (NetRec) for reconfiguring arbitrary network topologies in a user-transparent way, when anomalies (cyber attacks, performance failures, viruses, node and link failures) occur. Properties (such as correctness, termination and liveness) of NetRec will be presented together with its complexity analysis (in terms of messages exchanged .) A Virtual Testbed, based on COTS, for validating virtual networks has been implemented and seamless task execution for a real-life application (data compression) will be demonstrated for different virtual networks on different platforms - such as System Area Network (ServerNet 2) and Ethernet.

Bio of the Speaker

In total, Dr. Dimiter Avresky has published 89  papers in the area of: scalable high-speed networks (performance analysis, network control, routing, fault tolerance), protocol testing and verification, parallel computers, parallel programming, computer architecture, software verification, testing and fault tolerance and functional programming. Dr. Avresky is a guest editor of  four  IEEE  journals ( IEEE Transactions on Computers 02/2002, IEEE Transactions on Parallel and Distributed Systems 02/2001, IEEE Micro 2001 and  IEEE Micro 1998)  and  The Journal of  Supercomputing, 2000,Kluwer Academic Publishers in USA.  In addition, seven books and five book  chapters have been  published.  Dr. Dimiter Avresky has supervised eight  Ph.D. and five MS students. He is presently the major advisor of six   Ph.D. students.  Dr. Avresky has been funded by NSF, NASA, Hewlett Packard research Labs, Motorola Research Labs, Bell Labs  and others institutions. He served as General and Program Chair  of   twelve   IEEE workshops in the field of dependable computing and the Founder and a program Chair of two IEEE International Symposiums in the field of Network Computing  (NCA_01 and NCA_03), Cambridge, MA. Also served as a reviewer  for The IEEE Transactions Computers, IEEE Transactions on Parallel and Distributed systems and other refereed journals and a member of the Program Committee and a reviewer for numerous IEEE conferences.

BACK