![]() | Only 14 pages are availabe for public view |
Abstract algorithms was proposed in [78][79][80][81][83]. A cost based vertical and horizontal fragmentation approach was proposed [85]. A vertical partitioning algorithm using intelligent agents was proposed [86]. A methodology for applying CRUD to the detailed phase of object oriented design was presented [7]. 2.4 Summary The primal motivations for the distributed databases are improve performance, expandability, access facility and increase the availability of data. To design an efficient traditional distributed and parallel database system and especially in the case of designing it over the cloud environment, several interacting design decisions need to be taken. The most important design decisions are: How to fragment database tables, which type of fragmentation will be used, when to replicate fragments, what is the optimal number of replicas that can be taken for each fragment to enhance the systems performance and increase availability, how to allocate fragments to sites where they are most frequently accessed, do we group distributed database sites into disjoint clusters and which type of clustering will be used. Fragmentation, replication and allocation are considered the most important design issues that lead to optimal solutions, particularly in a dynamic distributed environment. They also have a great impact on the Distributed Database Systems (DDBS) performance. In this chapter a survey of the previous works done to handle these decisions and their limitation and future work was presented. Summary Proper data fragmentation, allocation and replication are the key performance factors for an efficient functionality of distributed database systems (DDBSs). HoweHoweHoweHowever,ver, ver, many emerging applications of distributed database systems generate very dynamic workflows with frequent changes in data access. Consequently, in this realistic dynamic environment, where the access probabilities of nodes to fragments and its replicas change over time, the optimum data re-allocation and replication of those fragments is the only way to increase the performance, efficiency, reliability and availability of the distributed database. In this chapter a dynamic reallocation and replication algorithm was proposed, which allows the cluster heads to work in parallel to manage the nodes of its cluster. Each cluster head makes decisions to migrate and/or replicate independently of other cluster heads using the proposed OFRAR algorithm. This makes it possible to use the proposed algorithm without communication overheads or even using the proposed algorithm on all sites in the system. The OFRAR algorithm incorporates data access patterns, time and site constraints of the DDBS, as well as the volume of data transmitted from the site that holds the fragment or replica, overloads of the sites and threshold values of accesses to dynamically re-allocate and replicate fragments to sites at runtime This chapter firstly thoroughly discusses the experimental results of the proposed cluster based distributed and parallel database design over a cloud environment and its components Experimental results show that the clustered based distributed database design, results in a significant reduction of the execution time needed for the transactions to reach the data in an appropriate site and the time taken to notice an error when an invalid query or the data not found occurs. Secondly, it thoroughly discusses the proposed optimal fragment re-allocation and replication (OFRAR) algorithm. The experimental results showed that the OFRAR algorithm improves the overall performance and efficiency of the distributed database by imposing a more strict condition for fragment re-allocation and replication in the distributed database. Moreover, it results in a significant reduction in the frequency of unnecessary fragments and replica migrations from one site to another over the network and the overload of the sites. In addition, it substantially reduces the data transmission costs as compared to the Optimal algorithm and the Threshold algorithm. Thirdly, it thoroughly discusses the Vertical Fragmentation, Allocation and Replication scheme (VFAR) and also the Full Vertical Fragmentation, Allocation and Replication scheme (FVFAR) of distributed database. Experimental results of the FVFAR scheme can be summarized in these points: It results in a significant reduction of total communication costs to execute the DDBS queries than the previous vertical partitioning approaches. It has a significant reduction of the overheads of the complicated computations than the previous vertical partitioning approaches |