大数据厂商联盟--为您的客户在30天内部署大数据应用精要.ppt

arc Inc. / arc Information Services GmbH * (c) arc 2014 * * * * * * arc Inc. / arc Information Services GmbH * (c) arc 2014 * * Notes: DMX-h actually ran in 2 nodes while Pig + Java ran in 3 nodes. Elapsed time was: DMX-h in 2 nodes: 26 mins Pig + Java in 3 nodes: 57 minutes Testing Environment: Pig: 3 Nodes Syncsort DMX-h: 2 Nodes * * NEO - CLUSTER - LOAD BALANCER - SPEC. HW - CLOUD - MONITORING - QUEUE - CACHE - END USER - VIZ - REPORTING OTHER DB - ENT. APP. INTEGR. - ETL - INDEXING - BULK - ETL RUNTIME * Velocity will sneak up on you – push downward through legacy database tiers and continue to affect more and more applications: BI, analytics, etc. * * * * * * * * * 客户 PoC: ParAccel MPP 对比 Greenplum * Confidential ? 2013 Actian Corporation Group Number ParAccel Greenplum PA times better ? DB1 1 14.0 132 9.4 5 24.0 220 9.2 4 9.7 110 11.3 Count distinct queries on highly distinct column 9 8.0 63 13.0 correlated sub queries 3 35.0 290 15.2 Outer join queries 8 12.0 102 17.4 12 Joins in a query with 2 large tables 11 2.7 35.1 22.7 19 5.0 66 28.9 12 5.0 110.3 35.4 10 3.9 102.5 44.6 GEOMEAN 1.75 3.29 AVERAGE 11.93 123.09 COMPOSITE 4.57 20.14 客户 PoC使用 ParAccel MPP 的好处：对比现有的 Teradata * Confidential ? 2013 Actian Corporation 延长现有数据仓库的使用年限 2年以上 提高数据分析性能 达到18倍 提高性价比 10倍对比 1/10 成本 缩减数据库管理 达到75% 减少基础设施成本，对比升级 1/10 提高数据分析量 达到10倍 提高分析人员利用率 超过50% 第三代超高速处理PB级大数据分析平台 SpaceCurve 适合空间地理气象数据处理 历史静态数据+动态数据+实时数据集成分析 快速部署大数据的可视化Datameer 海量数据的超高速查询 XQuery XML extension Powered by 海量数据的超高速查询 28.io Unified Information Access Platform Content sources storing documents, emails, books, articles, web pages, etc. Relational Database Filesystem Third-Party Content Feeds (XML, HTML, ZIP, etc.) NoSQL Database Query across disparate content repositories Editors, Content Curators Search, reuse, repurpose content Web Sites, Portals, Web Apps Mobile Apps Dynamically render and deliver content in multiple formats … … Custom Content Feeds Separate Task- Specific Repository (e.g., M