You are cordially invited to attend Computer Science Seminar Series
2:00 PM, Wednesday, December 4, 2002
3rd Floor Tower Room

Speaker:
Prof. Sten F. Andler
“Distributed Simulation via an Active Real-Time Database; Self-Stabilization and Diskless Real-Time Database Recovery”
University of Skovde, Sweden

ABSTRACT: We first present a communication architecture that uses a distributed active real-time database system as its communication medium. The proposed architecture incorporates a database as the shared whiteboard through which every participating node may communicate by writing and reading data. This approach is useful in complex sharing applications such as distributed real-time simulations. As a filtering mechanism on the shared simulation activities, active functionality is used to control complexity. This whiteboard communication architecture is suitable for distributed real-time simulations, and a single-node prototype implementation of such a database in conjunction with the distributed simulation standard High Level Architecture (HLA) has shown encouraging results. The prototype is being extended to multi-node simulation to gain additional experience before the communication architecture is deployed in actual systems.

Second, we discuss consistency in distributed real-time databases. Such databases generally have strict consistency requirements on data replicas, meaning that they are not allowed to diverge outside of transaction boundaries. For certain applications, this is too pessimistic, and it is often better to trade off immediate consistency for higher availability, performance, or predictability. In this presentation, we describe the replication protocol of DeeDS, a distributed real-time database which is eventually consistent, i.e., it allows replicas to diverge if the system can be guaranteed to eventually converge to a consistent state. DeeDS allows local transaction commit at the expense of relaxed consistency. The "super transaction" (SAGA) that includes the replication effort guarantees eventual consistency but may suffer from relaxed isolation.

Finally, we discuss real-time database recovery in distributed systems. A mechanism is introduced for diskless database recovery in fully replicated distributed real-time database systems. Database recovery relying on disk-resident redundant data cannot always be used in real-time or embedded systems for environmental, space, or cost reasons. The diskless recovery mechanism enables a recovering node to retrieve a copy of the database from any suitable remote node. This real-time recovery mechanism does not violate timeliness during normal processing and, during recovery, all nodes except for the recovering node are able to guarantee the timeliness of critical transactions. The mechanism uses a fuzzy checkpointing method to restore the database to the recovering node. This allows copying of the database without regard to concurrency control and, thus, does not increase data contention in the database.

The work reported here is joint work with Marcus Brohede, Sanny Gustavsson, AEgir Örn Leifsson, Jonas Mellin, and other members of the Distributed Real-Time Systems research group.

SPEAKER: Prof. Sten F. Andler received a Ph. D. in Computer Science in 1979 from Carnegie Mellon University, Pittsburgh, PA, and a Ph.D. in Computer Science from Chalmers University of Technology, Gothenburg, Sweden, also in 1979. He is a Professor of Computer Science at the University of Skovde, specializing in the areas of distributed systems, real-time systems, and operating systems. He was previously affiliated with IBM Almaden Research Center and IBM Software Solutions, San Jose, CA. Prof. Andler leads a research group in Distributed Real-Time Systems (DRTS), with interests in distributed real-time systems, active real-time databases, and testing of event-triggered real-time systems.