We propose and analyze a distributed cooperative\udcaching strategy based on the Evolutive Summary Counters\ud(ESC), a new data structure that stores an approximated record\udof the data accesses in each computing node of a search engine.\udThe ESC capture the frequency of accesses to the elements\udof a data collection, and the evolution of the access patterns\udfor each node in a network of computers. The ESC can be\udefficiently summarized into what we call ESC-summaries to\udobtain approximate statistics of the document entries accessed\udby each computing node.\udWe use the ESC-summaries to introduce two algorithms that\udmanage our distributed caching strategy, one for the distribution\udof the cache contents, ESC-placement, and another one for the\udsearch of documents in the distributed cache, ESC-search. While\udthe former improves the hit rate of the system and keeps a large\udratio of data accesses local, the latter reduces the network traffic\udby restricting the number of nodes queried to find a document.\udWe show that our cooperative caching approach outperforms\udstate of the art models in both hit rate, throughput, and location\udrecall for multiple scenarios, i.e., different query distributions\udand systems with varying degrees of complexity.
展开▼
