Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
ECE Department Seminar Fall 2008 Declarative Infrastructure Configuration Synthesis and Debugging Speaker: Dr. Sanjai Narain Telcordia Technologies, Inc. Joint work with Dr. Gary Levin, Telcordia Professor Daniel Jackson, MIT Professor Sharad Malik, Princeton Time: 11/12/2008 3PM-4PM Location: Babbio 319 Abstract: There is a large conceptual gap between end-to-end infrastructure requirements and detailed component configuration implementing those requirements. Today, this gap is manually bridged causing large numbers of configuration errors. Their adverse effects on infrastructure security, availability, and cost of ownership are well documented. This talk presents the ConfigAssure system for automatically bridging this gap. It proposes solutions to four fundamental problems: specification, configuration synthesis, configuration error diagnosis, and configuration error repair. Central to ConfigAssure is a Requirement Solver. It takes as input a configuration database containing variables, and a requirement as a first-order logic constraint in finite domains. The Solver tries to compute as output, values for variables that make the requirement true of the database when instantiated with these values. If unable to do so, it computes a proof of unsolvability. The Requirement Solver exploits the power of Kodkod, a SAT-based model finder for first-order logic. In turn, Kodkod exploits the power of modern SAT solvers, such as ZChaff, that can solve millions of constraints in millions of variables in seconds. However, a straightforward encoding of requirements in Kodkod does not scale to infrastructure of realistic size. ConfigAssure achieves scalability via two methods. The first is the use of a new quantifier-free form language called QFF consisting of Boolean combinations of simple arithmetic constraints. QFFs are efficiently implemented in Kodkod. The second is the use of a partial evaluator that transforms a first-order logic requirement into an equivalent QFF that truly requires Kodkod. ConfigAssure has been successfully trialed on a real enterprise network. Time permitting, a complementary diagnosis-only Telcordia product called ConfigAssure, deployed in a real enterprise, will also be outlined. Biography: Sanjai Narain is a Senior Research Scientist in the Information Assurance and Security Department at Telcordia Technologies in Piscataway, NJ. His current research is on systematic planning of secure and reliable infrastructure. This is based on his experience designing, building, testing and analyzing such infrastructure for large enterprises. To support his research, he has obtained funding from major government agencies such as DARPA, DISA, DHS and IARPA. He has organized and led several university-industry teams with partners such as MIT, Princeton, Cornell, Johns Hopkins, Boeing and Sparta. He is the lead editor of a new IEEE JSAC issue on network infrastructure configuration. He has served on program committees for ACM, IEEE and USENIX conferences or workshops. He is coauthoring a chapter on network configuration validation in Building and Running Robust Networks and Applications: A Handbook. He joined Telcordia in 1990 when it was called Bellcore. His earlier research at Telcordia was on network management tools for SONET, ATM and DSL networks. From 1981 to 1990 he worked at RAND Corporation where he developed technologies to reason about discrete-event simulation models. He has one issued patent on low-cost DSL loop qualification and three filed patents on configuration validation and synthesis. He has over twenty publications in journals, conferences and workshops. His formal training is in mathematical logic, programming languages, and electrical engineering. He studied logic with Professor Alonzo Church at UCLA. He obtained a Ph.D. in Computer Science from UCLA in 1988 where his thesis was nominated for the ACM Distinguished Dissertation Award. He obtained an M.S. in Computer Science from Syracuse University in 1981. He obtained a B.Tech. in Electrical Engineering from Indian Institute of Technology, New Delhi, in 1979 and graduated First Class with Honors. Contact: Yingying Chen < [email protected] >