Browsing by Author "Eng Koon SZE"

Now showing 1 - 2 of 2
  • No Thumbnail Available
    Item
    Efficient View Maintenance Using Version Numbers
    (2000-02-01T00:00:00Z) Eng Koon SZE; Tok Wang LING
    Maintaining a materialized view in an environment of multiple, distributed, autonomous data sources is a challenging issue. Given that the view site does not control the transactions at the data sources, results of incremental computation is affected by interfering updates and compensation process is required to resolve such problem. The compensation algorithm in our previous work handles the resolving of interfering updates without making any assumption on the first-sent-first-received and non-lost delivery of messages over the communication network. In this paper, we improve the incremental computation process by cutting down unnecessary maintenance queries and thus their corresponding query results, through using the knowledge of referential integrity constraints and functional dependencies of the base relations. We reduce the number of times of sending sub-queries to a data source with multiple base relations, as well as avoiding the execution of cartesian products, by using the join graph of the view to determine the access path of querying these base relations. We also provide a compensation algorithm that makes the accessing of multiple base relations from the same data source within a single sub-query possible. Overall performance is improved by cutting down the sending of unnecessary view maintenance queries and results, sizes of results are reduced drastically as cartesian products are avoided, as well as the number of times of issuing sub-queries to each data source.
  • No Thumbnail Available
    Item
    A Model for Evaluating and Comparing Materialized View Maintenance Algorithms
    (1999-11-01T00:00:00Z) Eng Koon SZE; Tok Wang LING
    Providing integrated access to information from multiple, distributed, autonomous data sources has received much attention from both the research communities and industries in recent years. Having a materialized view is a straightforward solution to provide fast access to such information, but maintaining such a view to reflect the changes at the data sources is not an easy task. Numerous algorithms have been proposed to handle the materialized view maintenance. The problems faced in this view maintenance include the presence of interfering updates, misordering of messages and the loss of these messages. In this paper, we provide a framework for comparing the merits of each of these approaches. We classify our criteria under four main categories in terms of the environment that they function in, the correctness of the algorithms, their efficiency, and the application requirements that these algorithms provide. The environment criteria includes the number of data sources the algorithm supports, and the nature of the compensation process. The correctness criteria looks at how accurate is the identification of interfering updates, and the proper working of the view maintenance process when messages are misordered or lost during their transmission through the network. Efficiency consideration includes the number of base relations accessed per sub-query of the incremental computation, the sequential or parallel handling of incremental computation both within the same update and between different updates, the use of partial self-maintenance in the incremental computation and how is modification handled. Application requirement involves the flexibility of the view definition, degree of consistency provided, and whether is quiescent state of the system necessary before the view can be refreshed. Based on these criteria, we evaluate and compare the existing algorithms in detail.