Computational & Technology Resources
an online resource for computational,
engineering & technology publications
Civil-Comp Proceedings
ISSN 1759-3433
CCP: 101
Edited by:
Paper 35

A Dynamic Hybrid Data-Management Mechanism for Data Grids

K.K. Pattanaik and U.B. Reddy

Atal Bihari Vajpayee-Indian Institute of Information Technology and Management, Gwalior, Madhya Pradesh, India

Full Bibliographic Reference for this paper
K.K. Pattanaik, U.B. Reddy, "A Dynamic Hybrid Data-Management Mechanism for Data Grids", in , (Editors), "Proceedings of the Third International Conference on Parallel, Distributed, Grid and Cloud Computing for Engineering", Civil-Comp Press, Stirlingshire, UK, Paper 35, 2013. doi:10.4203/ccp.101.35
Keywords: data grid, data management, scheduling, OptorSim, access cost.

The static replication strategy replicates a file whenever there is a miss without considering file popularity at the source site. As a result of replication of even non-popular files there is an increase in the replica management cost and the overall job execution cost. The existing strategies for dynamic replication, replicates a file when a file is popular at the source site. Request for access to non-popular files is done through remote access causing increased file access latency and increased average job execution time. Relocating such files to the point of consumption would result in decreased file access latency and hence decreased overall job execution time. Consideration of the replication approach alone does not optimize the scenario. Hence this necessitates a mechanism that incorporates relocation along with replication dynamically. The proposed dynamic hybrid data-management (DHD) mechanism is implemented based on the access behaviour of files to decide on creating a replica or relocating. The mechanism is implemented in the replica optimizer component of OptorSim. The simulation was carried for different sets of jobs for different job schedulers to evaluate the implication of pure replication, pure relocation and our PDIR strategy on the overall disk space utilization and overall job execution time. Our scheme DHD has been evaluated with different schedulers such as the random scheduler, the queue length scheduler, the access cost scheduler and the queue access cost scheduler against mechanisms such as complete relocation, complete replication using the optimizers LRU, LFU and the economic model Zipf distribution. Average job execution times and the percentage of storage element filled graphs were drawn for the mechanisms considered. Also with the number of jobs being increased the variation of average job execution time and percentage of storage element is shown. It is concluded that the enhanced dynamic replication strategy results in a performance-efficient system in terms of overall job execution time and overall percentage of the storage element filled.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £40 +P&P)