Developing a pieces of data allocation method in distributed databases using Bayesian networks

Determining the location of the storage of data on different nodes of a distributed system is called allocation of pieces of data and is considered as the most important factor influencing the cost of the implementation of distributed database. Allocation of piece of data is a NP-hard problem that needs heuristic solutions. The main purpose of the process is to allocate parts with the lowest cost to the node that has the most access to the data segment. In this article, in terms of a structural oriented learning method, using analysis of the factors influencing the process of allocating a piece of data to a node, (i.e. distance from apiece of data, cost, node loading, availability of data, etc), the dependence relations between adjacent nodes to target nodes was determined and the structure of Bayesian network was obtained. In the second stage, the obtained Bayesian network is trained using gradient method and tested using the data collected from the data collection x. The proposed model efficiency is compared with the results of Huang and Chen's method which have used a heuristic approach using the imperialist competitive algorithm. The results indicate that this model is an efficient tool in the optimization process of allocating Pieces of Data in the distributed systems.