Optimized Admission Policies for Intersection Caches using a Data Mining Approach

ESTEBAN FEUERSTEIN; TOLOSA GABRIEL

Deauville, Normandie

Simposio; Second International Symposium on Web AlGorithms iSWAG2016; 2016

iSWAG Consortium

Web Search Engines process huge amounts of databy using sophisticated structures and many optimizationtechniques that efficiently support search. One of themost important technique is caching. Basically, a searchengine implements a results cache on the broker side anda posting list cache on the search node side. Complemen-tarily, an intersection cache may be also implementedwhich keeps in the memory of the search node the resultsof intersecting the inverted lists of some selected terms.In this work we propose and evaluate an admissionpolicy for intersection caches based on a data mining ap-proach and we also integrate it into the query resolutionprocess. Our evaluation on a simulation framework andusing real datasets shows that it is possible to reduce thetotal processing cost close to 12%.