A PARADIGM FOR MAINTAINING MUTUAL CONSISTENCY OF REPLICATED DATABASES IN PARTITIONED NETWORK

Abstract

Several replica control algorithms for managing replicated databases in the face of network partitioning due to site or communication link failures has been reviewed. Pessimistic algorithms ensure consistency at the price of reduced availability: they permit at most one (distinguished) partition to process updates at any given time. In this work a pessimistic algorithm, dyr~am.ic voting with Zinearty ordered copies(abbreviated as dynamic- linear)under normal operation (that is, with the assumption that no failures occur during the execution of the protocol) has been successfully implemented. This algorithm permit updates in a partition provided it contains more than half of up-to-date copies of the replicated files. Each site consist of two modules: coordinator and subordinate. A site wishes to process update on local copy of the file, act as coordinator or it follows the request from coordinator and processes them accordingly. Activity of coordinator and subordinate ensure the atomic execution of transactions (i.e., either perform update on copy of the file, or abort the update). All communication between coordinator and its subordinate, lies on the different sites, is by message passing using message queues facilities available in UNIX. A restart protocol ensure a failed site,after recovery can regain its voting status.