Research On Adaptive Web Documents Replication Mechanism

Replication mechanism is widely used in Internet to reduce access time of web documents, increase web sites' availability, overall network traffic and average server loads. Designing a large-scale web documents replication mechanism, we need to address two most important issues: replica placement and consistency enforcement. Unfortunately, there is no single replication strategy that can efficiently manage all documents, because of heterogeneity of documents, various access models, changing system and network environments.This paper proposes an adaptive web documents replication mechanism. It uses differentiating replication strategies for each web document. By monitoring some key metrics to trigger adaptation of strategies, and dynamically selecting optimal strategies for documents, this adaptive replication mechanism provides its clients with the best available performance while consuming as little resources as possible.The paper discusses four issues of web documents replication briefly. They are adaptation triggering, replica placement, consistency enforcement and request redirection. We propose the adaptive web documents replication mechanism, which uses distributed web documents to support differentiating replication strategies for each document. Based on Autonomous Systems, it clusters clients and monitors system metrics. It sets threshold to trigger adaptation, and selects the subset of replication strategies by transition heuristic. It does trace-driven simulations to test replication strategies and selects an optimal arrangement by cost function arrangements. We use real-time trace-driven simulations to show that differentiating replication strategies clearly outperform any global strategy, test effect of x and effect of the length of traces in transition heuristic. According to the results of our simulations, it is obviously that the extra costs incurred by those operations is small compared to the performance improvements that per-document replication strategies with dynamically selecting replication strategies.For implementing the adaptive replication mechanism in large-scale distributed systems and data grid, we need integrate it with current web server. This is the workin the future.