Distributed Object Computing: Ideally, a distributed object-oriented application should be developed such that the interface to the object remains uniform regardless of the location of the object. This is ideal only on a functional and interface level since performance and efficiency reasons, as well as business policy reasons, impose restrictions on how the objects can be distributed. I am interested in object architectures for high-performance computing and in exploring the relationships between the various distributed object models that are prevalent today, including the CORBA model, the OLE/COM model and others.
High-performance transaction processing: My interests in this area have been in utilizing transaction processing monitors, such as BEA's Tuxedo or Transarc's Encina, to enable reliable, highly available three-tier client-server environments. I am very interested in the integration of TP monitors and Object Request Brokers and how one can develop seamless migration strategies between TP monitors.
Distributed/replicated databases: My interests in this area have been in increasing the performance of replicated databases by using the semantics of applications. My PhD dissertation dealt with increasing the concurrency of transactions using non-serializable but correct schedules, while also increasing the autonomy of sites in executing transactions. The dissertation introduced the notion of bounded ignorance that allows a system to increase concurrency at the expense of violating integrity constraints to a bounded extent. We have developed results and algorithms that outline how integrity constraints expressed in first-order logic can be relaxed, and how the "ignorance matrix" can be derived under certain assumptions. For instance, the integrity constraints of a relational database can be relaxed to provide increased concurrency between insert and delete operations acting on relations. Another area I have been interested in is the comparison of the several semantics-based conflict relations that exist in the literature.
Mobile computing and data management: Information services atop wireless are fast becoming a reality. Such services include personalized news and financial information, banking, sales inventory, travel information and the like. The work I have been doing deals broadly with providing Personal Information Services and Applications (PISA) to mobile users via a PCS network (which could also incorporate the wired PSTN infrastructure).
Workflow management: Much of my work in this area has been done in the context of the the Managing End-to-End Operations (METEOR) project at Bellcore. METEOR develops a framework for re-usable specification and execution of transactional workflows. Our approach is based on an integration and adaptation of several approaches in the literature. We provide a model based on task structures that allows the flexible integration of heterogeneous tasks into a workflow. Furthermore, by using the notion of user-defined compound tasks that can include simple tasks or other compound tasks, we allow workflows to be nested. A unique feature of METEOR (adopted due to the requirements of the telecommunications environment) is the support for dynamic workflows, where at run-time several tasks and dependencies can be created anew. We have developed two high-level languages for specifying a workflow at different levels of abstraction: the Workflow Specification Language (WFSL) is a declarative rule-based language used to express the application-level interactions between multiple tasks, while the Task Specification Language (TSL) focuses on the issues related to individual tasks. We have also addressed the problems of multi-database concurrency control and recovery in the telecommunications environment. By using the semantics of the applications, we have defined a simple variant of altruistic locking for ensuring global serializability.
Courses
CS46a: Operating Systems Concepts
Selected Publications