Research Meeting Jaeseok Myung
Copyright 2009 by CEBT Summary TA DB: 중간고사 채점 – 평균 : 66.04, 표준편차 : – 지난학기 평균 : 59.33, 표준편자 : WEC: classroom( ), 강사 (SKT, 정준용 매니저 ) 서울대 멘토링 진행중 관리자 업무플로우, 금주 마감 예정 (3 주 ) 디비 스키마, 매칭 모듈 설계 진행 예정 (1~2 주 ) Research SPARQL BGP Processing with Iterative MR – Using finer keys for map tasks => Scailability – Using advanced storage for selection task => Performance – Using selectivity for BGP analysis – Using MR pipelining Center for E-Business Technology
Copyright 2009 by CEBT Using Advanced Storage for Selection Task Center for E-Business Technology SELECT ?A ?B ?C ?D … WHERE { ?A memberOf ?B. ?B type Univ. … } 1212 SPO a1memberOfb1 typeUniv ……… TP_NOAB… 1a1b1… 2 NULL… ……… Selection BigTablePP SOO SOO SOO SOO
Copyright 2009 by CEBT Optimization for Selection Task Triple Indexing SP-O SO-P PS-O PO-S OS-P OP-S S-PO P-SO O-SP Center for E-Business Technology Jaeseok MemberOf ?X Jaeseok ?x IDS Jaeseok ?x ?y BigTablePP SOO SOO SOO SOO OO PSS PSS PSS PSS SS OPP OPP OPP OPP Up to 3 times bigger disk space is needed for eliminating full table scan Reducing the use of disk by using dictionary encoding
Copyright 2009 by CEBT Optimization for Selection Task Implementation Using Hadoop – HBase Adding Data Loader Component – N-Triple => HBase Implementation of Selection Tasks using HBase Comparison between N-Triple and Hbase Dictionary Encoding Center for E-Business Technology