Research On Data Allocation Algorithms In Distributed Database Systems

With the increased demands of databases and the rapid development of the information tech--nology,centralized database does not meet the current demand for data storage,which lead to theappearance of distributed databases.Data allocation is a key aspect in the design of distributed data--bases.If the problem of data allocation is solved properly,the availiability and reliability of the data--base systems will be improved greatly.In this paper,in order to reduce the communication cost and improve overall perfor--mance,the genetic algorithm based on clustering is proposed.The statistical information isthe important data sources to solve the problem of data allocation,which includes the info--rmation of database, application,site and network. First of all, Clustering-genetic algo--rithm selects the statistical information used in the cost function, which is based onthree major principles.At the same time,Cost function is also an important basis to eva--luate the algorithm’s results,which includes the cost of local retrivals,local updates,remote update and remote communication.The purpose of the algorithm is to minimize thesum of this four costs.At first,the algorithm groups all the sites into different clusters bythe communication cost between the sites.and assess the total cost when the data fragm--ents are allocated in each cluster.secondly,the improved genetic algorithm is used to getthe distribution of data fragments in the clusters,which is based on the fitness,namely thesum of all the costs.finally, the data fragments are allocated to the sites in each cluster,which is based on the principle of optimization,the calculation of the total costs still takethe former cost function.In the end of paper,with the comparison of experimental results of the clustering--reallocated algorithm,the results of this paper show a less data fragments and less totalcost and average cost. Meanwhile,reducing the overall average cost of communication.