ویراساینس.
مقاله

An Efficient Data Replication Strategy in Large-Scale Data Grid Environments Based on Availability and Popularity

N. Mansouri, M.M. Javidi

چکیده

The data grid technology, which uses the scale of the Internet to solve storage limitation for the huge amount of data, has become one of the hot research topics. Recently, data replication strategies have been widely employed in a distributed environment to copy frequently accessed data in suitable sites. The primary purposes are shortening distances of the file transmission and achieving files from nearby locations to requested sites so as to minimize retrieval time and bandwidth usage. In this paper, we propose a new replica selection strategy which is based on response time and security. However, replication should be used wisely because the storage size of each Data Grid site is limited. In addition, we propose a new replica replacement strategy that considers file availability, time of access, access frequency and size of file. The simulation results report that the proposed strategy can effectively improve mean job time, bandwidth consumption for data delivery, and data availability compared with those of the tested algorithms.

شکل‌ها (۱۵)

  • Fig. 1. Structure of the Data Grid architecture. of data file that can be copied by replication strategy. Thus, replication strategies are needed to decide when replication 1- 1- ISSUES RELATED TO DATA GRID must take place, what data files must replicate and where
    Fig. 1. Structure of the Data Grid architecture. of data file that can be copied by replication strategy. Thus, replication strategies are needed to decide when replication 1- 1- ISSUES RELATED TO DATA GRID must take place, what data files must replicate and whereصفحه‌ی ۱
  • شکل، صفحه‌ی ۱
    صفحه‌ی ۱
  • Fig. 2. A Replica Management Architecture. European Data Grid Project is provided. The final section Generally, replication algorithms are either static or dynamic. contains conclusion and future work. In static approaches, the created replica will exist in the
    Fig. 2. A Replica Management Architecture. European Data Grid Project is provided. The final section Generally, replication algorithms are either static or dynamic. contains conclusion and future work. In static approaches, the created replica will exist in theصفحه‌ی ۲
  • شکل، صفحه‌ی ۴
    صفحه‌ی ۴
  • Fig. 4. Pseudo code for replica replacement strategy. catalog, the replica manager determines the locations of the hil filfthdd fil physical files for the needed file.
    Fig. 4. Pseudo code for replica replacement strategy. catalog, the replica manager determines the locations of the hil filfthdd fil physical files for the needed file.صفحه‌ی ۶
  • Fig. 4. Pseudo code for replica replacement strategy. catalog, the replica manager determines the locations of the hil filfthdd fil physical files for the needed file.
    Fig. 4. Pseudo code for replica replacement strategy. catalog, the replica manager determines the locations of the hil filfthdd fil physical files for the needed file.صفحه‌ی ۶
  • Fig. 6. Architecture of OptorSim toolkit [40]. A resource broker manages job assignment on different sites like meta-scheduler. Each job needs different files that may be
    Fig. 6. Architecture of OptorSim toolkit [40]. A resource broker manages job assignment on different sites like meta-scheduler. Each job needs different files that may beصفحه‌ی ۶
  • Fig. 7. Mean job time with different access pattern generators.
    Fig. 7. Mean job time with different access pattern generators.صفحه‌ی ۷
  • Fig. 7. Mean job time with different access pattern generators.
    Fig. 7. Mean job time with different access pattern generators.صفحه‌ی ۷
  • Fig. 9. Mean job time based on the varying size of files. on the inter- region link. It is necessary to note that using As shown in Fig. 8, the bandwidth consumption of large scale the frequency of replica requests and the last access time of liiiifhlf lik gpg file even if the presence of replication is so high. Therefore, replica in EDRS as criterion for the placement of replica takes id k bdidhhld bli
    Fig. 9. Mean job time based on the varying size of files. on the inter- region link. It is necessary to note that using As shown in Fig. 8, the bandwidth consumption of large scale the frequency of replica requests and the last access time of liiiifhlf lik gpg file even if the presence of replication is so high. Therefore, replica in EDRS as criterion for the placement of replica takes id k bdidhhld bliصفحه‌ی ۸
  • Fig. 10. Mean job time with varying bandwidth.
    Fig. 10. Mean job time with varying bandwidth.صفحه‌ی ۸
  • Fig. 11. Effective network usage. gpgy this case the data is stored only in one location where the files are produced initially. But EDRS algorithm creates replicas dynamically in advance. Instead of storing files in
    Fig. 11. Effective network usage. gpgy this case the data is stored only in one location where the files are produced initially. But EDRS algorithm creates replicas dynamically in advance. Instead of storing files inصفحه‌ی ۹

شکل‌ها از متن کامل مقاله برداشته شده‌اند و حقوقشان از آنِ نویسندگان و ناشر است. منبع: صفحه‌ی اصلی مقاله

اسناد مشابه