An adaptive partition-based caching approach for efficient range queries on key-value data

Abstract

Range queries are real demands in big data scenarios, such as analytic and time-traveling queries over web archives. Here we design AdaSI, an adaptive partition-based caching approach for efficient range queries on key-value data. AdaSI partitions data into a number of data slices (consecutive data items). Then the AdaSI Hotscore Algorithm is designed to maximize the cache-hit probability under the limitation of cache space. By measuring Dutyrate and Hotscore of data slice, the partitioning precision and adjustment sensitivity are pursued by finer partitioning on hot data, whereas the cold data are partitioned with relatively larger granularity to reduce storage overhead and search cost of queries. Our results show that the AdaSI Hotscore Algorithm could obtain a cache hit rate nearly as high as the record-based cache policies, as well as a significant speedup and space reduction, far outperforming record-based policies.