Dynamic Sample Selection for Approximate Query Processing近似查询处理的动态样本选择.pptVIP

下载本文档

9
0
约5.93千字
约 23页
2017-02-13 发布于上海
举报
版权申诉

Dynamic Sample Selection for Approximate Query Processing近似查询处理的动态样本选择.ppt

1、本文档共23页，可阅读全部内容。
2、有哪些信誉好的足球投注网站（book118）网站文档一经付费（服务费），不意味着购买了该文档的版权，仅供个人/单位学习、研究之用，不得用于商业用途，未经授权，严禁复制、发行、汇编、翻译或者网络传播等，侵权必究。
3、本站所有内容均由合作方或网友上传，本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺！文档内容仅供研究参考，付费前请自行鉴别。如您付费，意味着您自己接受本站规则且自行承担风险，本站不退款、不进行额外附加服务；查看《如何避免下载的几个坑》。如果您已付费下载过本站文档，您可以点击这里二次下载。
4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等，请点击“版权申诉”（推荐），也可以打举报电话：400-050-0827(电话支持时间：9:00-18:30)。
5、该文档为VIP文档，如果想要下载，成为VIP会员后，下载免费。
6、成为VIP后，下载本文档将扣除1次下载权益。下载后，不支持退款、换文档。如有疑问请联系我们。
7、成为VIP后，您将拥有八大权益，权益包括：VIP文档下载权益、阅读免打扰、文档格式转换、高级专利检索、专属身份标志、高级客服、多端互通、版权登记。
8、VIP文档为合作方或网友上传，每下载1次，网站将根据用户上传文档的质量评分、类型等，对文档贡献者给予高额补贴、流量扶持。如果你也想贡献VIP文档。上传文档

Dynamic Sample Selection for Approximate Query Processing Brian Babcock Stanford University Why Approximation is Useful Large data warehouses Gigabytes to terabytes of data Data analysis applications Decision support Data Mining Query characteristics: Access large fraction of database Seek to identify general patterns / trends Absolute precision unnecessary $89,000 after 5 secs vs. $89,034.57 after 2 hrs Two Phases of Approximate Query Processing (AQP) Offline pre-processing of the database E.g. generate histograms or random samples OK to use considerable space and time (hours) Runtime query processing Query answers must be fast (seconds) Only time to access small amount of data E.g. extrapolate from random sample AQP Example Non-uniform Sampling “Biased” samples often more accurate than uniform samples All data records are not created equal Frequently queried values Extreme high and low values Uncommon values Optimal bias differs from query to query Past work: carefully select biased sample to give good answers for many queries Related Work Non-sampling-based approaches Online Aggregation Hellerstein, Haas, and Wang 97 Histograms Ioannidis and Poosala 99 Wavelets Chakrabarti, Garofalakis, Rastogi, and Shim 00 Sampling-based approaches AQUA project Acharya, Gibbons, and Poosala 99 Congressional Acharya, Gibbons, and Poosala 00 Self-Tuning Ganti, Lee, and Ramakrishnan 00 Outliers Chaudhuri, Das, Datar, Motwani, and Narasayya 01 Workload Chaudhuri, Das, and Narasayya 01 Dynamic Sample Selection Dynamic Sample Selection Small vs. Large Groups Consider group-by aggregation queries. E.g. Total sales of CPUs in each state E.g. Avg sale price for each product in each state Number of records per group may vary widely Problem: Rare values are under-represented in uniform sample “California” much more common than “Alaska” “Alaska” only appears a few times in the sample Approximate answer for “Alaska” likely to be bad In a group-by query, small groups are hard Small Gro