Research And Design On Replacement Strategy For Caching XML Query

XML(eXtensible Markup Language) with a good scalability has been widely welcomed and supported by the industry since the day it was released. It has emerged as a standard for representing and exchanging data on the World Wide Web. With the rapid growth of the application scale and complexity of XML, issues related to efficient processing of XML queries have received considerable attentions.Along with the continuous deepening of the XML query optimization, XML query for caching is gradually gaining more and more attention. The processing efficiency and response ability can be improved effectively through caching hot data (frequent query patterns). As the cache space is limited resource, appropriate replacement strategy should be designed to decide which frequent query patterns cache chooses while keeping the cache performance. So replacement strategy for caching needs to be focused on in solving the issues of XML query for caching.The paper uses semantic caching mechanism based on XML query as the effective means to improve the performance of XML query system. It designs an XML Document Query System introduced Semantic Caching on the basis of researching and analyzing the systems at present. After discussing the function of each module, the paper focuses on researching the replacement strategy in XML query cache and proposes an efficient approach for it. The approach will be carried out by three steps as follows:Firstly, the paper clusters user queries so that queries about similar information can be grouped together, analyzes the algorithm of mining frequent rooted subtrees and discovers frequent rooted subtrees from XML queries using the algorithm of FastXminer. And then the paper designs a novel approach to cluster XML queries based on their semantics.Secondly, the paper introduces and analyzes the concept of positive and negative association rules. FastXMiner mined frequent XML query patterns from the user queries. Once the cache is full, query regions of infrequent query patterns will be purged first, however, the frequent query pattern-based technique in FastXMiner ignores the temporal feature of user queries. The paper proposes an efficient algorithm of mining association rules from users' queries in XML cache after clustering XML queries. Since these association rules address the temporal sequence between user queries, it is more reliable in predicting future queries than the approaches that address the frequency or recency only. The paper considers the sequence between user queries to discover association rules. Both positive and negative association rules will be discovered after executing the algorithm.Thirdly, the paper gives the analysis of the current replacement strategies and designs an appropriate one for caching XML query on the basis of the knowledge of association rules mined before.At last, the paper and conducted various experiments towards the approach. Experimental results showed that the replacement strategy incorporated with discovered association rules had better performance than existing approaches.