Presentation of a new method based on modern multivariate approaches for big data replication in distributed...

Area of Science:

• Computer Science
• Data Storage and Processing
• Distributed Systems

Background:

• Increasing data volumes and distributed systems necessitate efficient data replication strategies.
• Reducing data replication is crucial for lowering traffic load in grid environments.
• Existing methods for replication site selection often lack comprehensive consideration of multiple factors.

Purpose of the Study:

• To propose an algorithm for reducing data replication in big data transfer.
• To decrease the overall traffic load on grid systems.
• To enhance the efficiency of replication site selection using multivariate decision-making.

Main Methods:

• Data classification based on sent data types for optimal distribution.
• Utilizing the VIKOR method for multivariate decision-making in ranking replication sites.
• Considering parameters like time, number of replications, and replication site for improved access.

Main Results:

• The proposed algorithm achieved an average improvement of approximately 30% over LRU, LFU, BHR, and Without Rep. algorithms.
• Demonstrated a 30% improvement compared to existing multivariate replication methods.
• Effective reduction in data replication leading to lower grid traffic.

Conclusions:

• The proposed algorithm offers a significant improvement in reducing data replication and network traffic.
• The VIKOR method effectively integrates multiple parameters for optimal replication site selection.
• This approach enhances data access efficiency by replicating data only when beneficial.