Current Projects
System Management of VM based Cloud Computing
Design and development of control and management system for heterogeneous and virtualized computing clusters. Research issues include VM dependency in resource allocation, VM process migration, power optimization, and dependable & fault-tolerant operation.
Scalable QoS Switch Architecture
Design and implementation of scalable switch architectures for large networks with varying quality of service. In addition to performance design metrics, implementation metrics such as power, space, and cost effectiveness are explored in the design of switch architectures.
Converged Network Architectures
Design of unifying network architectures based on GMPLS control plane for various existing and future services. Research issues include intelligent traffic management, secure and fault-tolerant control plane, management of multi-layer and cross-layer interactions, and optimization of network resources in multilayer networks.
Past Projects
Argus -Next Generation Network Security Management System
Design and development of distributed network security management system for heterogeneous networks including wireless infrastructure. Research issues include modeling and detection of network attacks, machine learning and data mining of network data, real-time network and host monitoring, and fast response, fault isolation, and prevention of network failures.
Minerals: Large Scale Network Monitoring and Management
Recent studies have shown that router misconfigurations are pervasive and can have dramatic consequences on the operations of networks. They can compromise the security of a single network, partition a network, violate import and export policies or even cause global disruptions in the Internet connectivity. Several solutions have been proposed to detect a number of problems in real configuration files. However, existing solutions share a common limitation, they are rule-based.
Rules are defined a priori, and violations of these rules are deemed as misconfigurations. As policies typically differ between networks, rule based approaches are limited in the scope of mistakes they can detect. We address the problem of network misconfigurations using data mining. We apply association rules mining techniques to configuration files from routers across an administrative domain to discover local, network-specific policies. Deviations from these local policies indicate potential misconfigurations. We have evaluated our scheme on configuration files from a large state-wide network provider, a large university campus and a research network. The results are promising. We discovered a number of errors that could have jeopardized the operations of the network had they not been discovered. These misconfigurations would have been difficultto detect with current approaches.
