综合论文：BlinkDB.QuerieswithBoundedErrorsandBoundedResponseTimesonVeryLargeData（英文）.pdfVIP

BlinkDB: Queries with Bounded Errors and Bounded Response Times on Very Large Data † ○ † † ○ ∗† Sameer Agarwal , Barzan Mozafari , Aurojit Panda , Henry Milner , Samuel Madden , Ion Stoica † ○ ∗ Abstract cessing of large amounts of data by trading result accuracy In this paper, we present , a massively parallel, ap- for response time and space. ese techniques include sam- proximate query engine for running interactive SQL queries pling [10, 14], sketches [12], and on-line aggregation [15]. To on large volumes of data. allows users to trade- illustrate the utility of such techniques, consider the following o query accuracy for response time, enabling interactive simple query that computes the average over queries over massive data by running queries on data samples all users originating in : and presenting results annotated with meaningful error bars. To achieve this, uses two key ideas: (1) an adaptive optimization framework that builds and maintains a set of multi-dimensional stratied samples from original data over time, and (2) a dynamic sample selection strategy that selects Suppose the table contains 100 million tuples for an appropriately sized sample based on a query’s accuracy or , and cannot t in memory. In that case, the above response time requirements. We evaluate against the query may take a long time to execut