The database replication mechanism ensures the duplication of data from one geographical location to another. In recent past, PDDRA (pre-fetching-based dynamic data replication algorithm) is investigated heavily under various classes of request arrivals. The request arriving at a particular server for replication depends on many factors such as bandwidth supported by link, its geographical location and routing algorithms. The number of served requests depends on the processing speed and buffering capacity of a particular server. If a request cannot be served at the server, it is dropped. If a bunch of requests arrive at a server, then traffic modelled as bursty traffic. The objective of this paper is to investigate the performance of dynamic replication algorithm under bursty request arrival. In a similar context, a mathematical model is detailed for the estimation of request burst length, such that both buffering capacity and transmission delay can be kept minimal. Finally, simulation results are present to evaluate the fraction of successful request under diverse conditions.

The database replication keeps up contents of database such as tables, in various geographical locations that fabricate an arrangement of distributed database. The necessities of database replication are expanding with the time as the use of internet is also increasing. With the purpose of meeting these prerequisites, priorities among requests can be included. In this work two types of priorities i.e., low and high priority is considered in this research article. It is noticeable that the high priority requests are more vital in comparison to low priority requests and therefore low priority requests can be delayed or dropped. The rate of request loss can be lessened utilizing the conditions of load balancing where a portion of the contending requests are transferred to other nodes. In this paper, internet is modelled as small world network, and performance assessment of A-PDDRA (Advanced Prebringing Based Dynamic Data Replication Algorithm) is carried out on small world network while considering both priorities and load balancing using computer simulation