Distributed deadlocks in distributed databases

Abstract

The developments of database management system (DBMS) technology has run in parallel with the developments in computer networks and distributed computing technologies.

Consequently there is the birth of distributed database management systems. These systems cannot be ignored any longer because they have now dominated data management

applications. Many DBMS have incorporated some degree of distribution into their products, thus it is very important that the issue of deadlocks in distributed databases has to be solved.

A deadlock is a potential problem in any system having multiple transactions which share resources. Transactions involved in a deadlock will indefinitely wait for the resources to be released, unless the deadlock is detected and recovery measures are taken. In distributed systems, lack of information about the complete system states compounds the deadlock detection problem.

There has been a wide range of research with focus on deadlocks in distributed systems in general, but distributed deadlocks in distributed databases has been neglected to some extend. This thesis describes in detail the distributed databases in terms of the architectures and types distributed databases. There are a number of algorithms that try to resolve and manage distributed deadlocks and some of them have not yet been proved to be correct. In this thesis

the researcher analysed some of the algorithms that have been implemented by some researchers. The algorithms analysed used the prevention method which requires that a transaction should have all the required resources before it proceeds which is virtual impossible in a busy distributed environment.

The researcher proposed the ISKE algorithm that shows that it is possible for any requesting transaction to proceed processing without wasting unreasonable amounts of time. The

algorithm showed that employing a method of priority as well as queuing (FIFO) can greatly manage the deadlocks effectively well.