Priority control based on website categories in edge computing

Abstract

Modern websites consist of many rich objects dynamically produced by servers and client terminals at diverse locations. To deliver dynamic objects efficiently, edge computing in which objects are dynamically generated and delivered to client terminals on edge servers located at edge nodes will be effective. It is anticipated that the effectiveness of edge computing depends on the geographical pattern of object deployment, and the tendency of geographical distribution of objects is different among website categories, e.g., Sports and News, so it seems effective to prioritize website categories for edge computing. In this paper, we first propose a platform for measuring the geographical tendency of object deployment in each website category. We then propose to differentiate the caching priority in edge computing among website categories based on the measured deployment pattern of objects. Through the experience of accessing about 1,000 of the most popular websites from 12 locations worldwide using PlanetLab, we clarify that we can improve the reduction ratio of web response time by about 20% by carefully differentiating caching priority among website categories.