Research Of Data Allocation Strategy In Distributed Database

Distributed database system is a product of the combination of database system and computer network system. Data allocation has a significant impact on improving the entire distributed database application system, data availability, and efficiency and reliability of distributed database. Domestic and foreign scholars have developed a variety of data allocation strategies. But these strategies have some shortcomings such as cost formula is too complex, efficiency is low or the result is much worse than the optimal allocation. In order to solve data allocation better, this paper presents a strategy based on genetic algorithm.Statistical information is basic information to solve data allocation. According to the importance of statistical information, cost of access, and the impact on the complexity of the cost formula, this paper mainly counts the database information, the application information and network communication cost, and then adopts a cost formula which mainly considers cost of transaction processing. In the allocation strategy of this paper, for each data fragment, algorithm first initializes population according to the ratio of updating and searching to the data fragment. Each individual in the population represents a allocation scheme for the data fragment. Then for each individual, the algorithm evaluates fitness of individuals according to the cost formula. During the evolutionary process, the algorithm adopts the strategy which combines fitness-proportionate selection and elitist selection to choose individuals. The algorithm adopts adaptive crossover operator to maintain the balance between searching speed of the algorithm and retaining superior genes, and adopts adaptive mutation operator to maintain the balance between population diversity and searching random of the algorithm.At the end of this paper, this strategy is validated through multiple sets of test data. The experimental results show that using allocation strategy in this paper can get a better result than using heuristic remove fragment copies allocation algorithm, heuristic add fragment copies allocation algorithm or allocation strategy based on the accessing characteristic of data fragment. Moreover, the genetic algorithm has high parallelism, which can significantly improve efficiency of the algorithm. And the operation mode and the implementation steps of the genetic algorithm are normal. So the genetic algorithm is easy to apply.