Sushant Goel

Replica synchronisation in grid databases

Grid architecture is heterogeneous in nature. Replica synchronisation protocols, normally used in homogeneous systems, are not applicable in heterogeneous architectures. In this paper, a replica synchronisation protocol for heterogeneous Grid architecture is proposed. The protocol uses the Grid middlewares metadata service and timestamps to synchronise the replicas among distributed sites. The protocol is then extended to handle multiple network partitioning. Grid sites are prone to multiple network partitioning. Most of the replica synchronisation protocols deal with simple network partitioning. The protocol and use contingency quorums are extended to deal with multiple networks partitioning in Grid databases.

Concurrency control issues in grid databases

Grid architecture is a fast evolving distributed computing architecture. The working of databases in the Grid architecture is not well understood. In view of changing distributed architecture we strongly feel that concurrency control issues should be revisited and reassessed for this new and evolving architecture. Implementing global lock table and global log records may not be practically possible in the Grid architecture due to the scalability issues. In this paper, we propose a correctness criterion and the Grid concurrency control protocol, which has the capability to deal with heterogeneity, autonomy, distribution and high volume of data in Grids. We then prove the correctness of the protocol followed by performance evaluation of the protocol.

Preserving Data Consistency in Grid Databases with Multiple Transactions

High performance Grid computing provides an infrastructure for access and processing of large volume, terabyte or even petabytes, of distributed data. Research in data grid has focused on security issues, resource ownership, infrastructure development and replication issues assuming presence of single transaction in the system. In this paper we highlight that grid infrastructure comes with new set of problems in maintaining the consistency of databases in presence of multiple transactions. Traditional distributed data management techniques may not meet the requirements of databases in grid environment. We first show the circumstances where grid infrastructure may produce incorrect results and then propose a correctness condition – Grid Serializability Criterion that preserves the consistency of data in data-grids.

Message-Oriented-Middleware in a Distributed Environment

Middleware technologies have been facilitating the communication between the distributed applications. Traditional messaging system’s are synchronous and have inherent weaknesses like – limited client connections, poor performance due to lack of resource pooling, no store-and-forward mechanism or load balancing, lack of guaranteed messaging and security as well as static client and server’s location dependent code. These weaknesses and increasing e-business requirements for the distributed systems motivated us to undertake this research. This paper proposes an asynchronous communication architecture – Transfer of Messages in Distributed Systems. The advantage of the proposed architecture is that the sender of the message can continue processing after sending the message and need not wait for the reply from other application.

Atomic commitment and resilience in grid database systems

Atomic Commitment Protocol (ACP) is an important part of any distributed transaction. ACPs have been proposed for homogeneous and heterogeneous distributed database management systems (DBMS). ACPs designed for these DBMS do not meet the requirement of Grid databases. Homogeneous DBMS are synchronous and tightly coupled while heterogeneous DBMS, like multi-database systems require a top layer of multi-database management system to manage distributed transactions. These ACPs either become too restrictive or need some changes in participating DBMS, which may not be acceptable in Grid Environment. In this paper, we identify requirements for Grid database systems and then propose an ACP for grid databases, Grid-Atomic Commitment Protocol (Grid-ACP). We then extend Grid-ACP to handle failure of database sites.

Atomic Commitment in Grid Database Systems

Atomic Commitment Protocol (ACP) is an important part for any distributed transaction. ACPs have been proposed for homogeneous and heterogeneous distributed database management systems (DBMS). ACPs designed for these DBMS do not meet the requirement of Grid databases. Homogeneous DBMS are synchronous and tightly coupled while heterogeneous DBMS, like multidatabase systems, requires a top layer of multidatabase management system to manage distributed transactions. These ACPs either become too restrictive or need some changes in participating DBMS, which may not be acceptable in Grid Environment. In this paper we identify requirements for Grid database systems and then propose an ACP for grid databases, Grid-Atomic Commitment Protocol (Grid-ACP).

Advances in high performance database technology

This tutorial will be based on the recently published book, High-Performance Parallel Database Processing and Grid Databases (John Wiley & Sons, 2008). The sizes of databases have seen exponential growth in the past and such growth is expected to accelerate in the future, with the steady drop in storage cost accompanied by a rapid increase in storage capacity. To effectively manage such volumes of data, it is necessary to allocate multiple resources to it, very often massively so. The processing of databases of such astronomical proportions requires an understanding of how high performance systems and parallelism work. Besides the massive volume of data in the database to be processed, some data has been distributed across the globe in a Grid environment. This important new book provides readers with a fundamental understanding of parallelism in data-intensive applications, and demonstrates how to develop faster capabilities to support them. It features not only the algorithms for database operations, but also quantitative analytical models, so that performance can be analyzed and evaluated more effectively.

Failure recovery in grid database systems

Failure is unavoidable in any computing environment and hence any computing architecture must address recovery issues. Recovery becomes more complicated when sites are distributed, autonomous and heterogeneous. Grid architecture is such an evolving distributed architecture. Databases operating in Grid architecture have different recovery issues than their other distributed counterparts – distributed and multidatabase. In this paper we focus on maintaining correctness of data in case of site failure in Grid database.

Transaction Management in Distributed Scheduling Environment for High Performance Database Applications

Synchronising the access of data has always been an issue in any data-centric application. The problem of synchronisation increases many folds, as the nature of application becomes distributed or volume of data approaches to terabyte sizes. Though high performance database systems like distributed and parallel database systems distribute data to different sites, most of the systems tend to nominate a single node to manage all relevant information about a resource and its lock. Thus transaction management becomes a daunting task for large databases in centralized scheduler environment. In this paper we propose a distributed scheduling strategy that uses a distributed lock table and compares the performance with centralized scheduler strategy. Performance evaluation clearly shows that multi-scheduler approach outperforms global lock table concept under heavy workload conditions.

Multi-scheduler Concurrency Control for Parallel Database Systems

Increase in amount of data stored and requirement of fast response time has motivated the research in Parallel Database Systems (PDS). Requirement for correctness of data still remains one of the major issues. Concurrency control algorithms used by PDS uses single scheduler approach. Single scheduler approach has some inherent weaknesses such as - very big lock tables, overloaded centralized scheduler and more number of messages in the system. In this paper we investigate the possibility of multiple schedulers and conclude that single scheduler algorithms cannot be migrated in the present form to multi-scheduler environment. Next, we propose a Multi-Scheduler Concurrency Control algorithm for PDS that distributes the scheduling responsibilities to the respective Processing Elements. Correctness of the proposed algorithm is then discussed using a different serializability criterion - Parallel Database Quasi-Serializability.