Cassandra简介8.pptVIP

Scalability[2] Scale Out Vs Scale Up Scale Up-在同一个逻辑单元内增加资源,例如增加CPU/内存/网卡数量等. Scale Out-增加多个逻辑单元的资源,并使它们如同一个集中的资源那样提供服务(集群/分布式/负载均衡等) Scale Up较为简单,但是规模有限,代价越来越大 Scale Out需要从架构层面设计,规模没有限制,代价由架构决定. 基本的存储模型 行存储 Vs 列存储 Vs 混合存储 行存储适合查找整行的存储,不过需要配合索引 列存储适合查找少量列,适合做基于列的统计/查询 混合列存储. 将需要经常组合查询的列组合在一起. 将其他列(列的组合)单独存储. 基本的存储模型[2] CAP Theorem Consistency the system provides a view of the distributed state which is consistent between observers 所有的用户都可以看到一致的系统状态 Availability The system as a whole should continue functioning , even if servers should fail or be unreachable due to network failures 无论何时,哪怕出现硬件故障,数据中心故障,系统也可提供服务,哪怕是降级的服务 Partition Tolerance The system as a whole should continue to function, potentially with degradations in service, even if the network can fail in arbitrary ways. 哪怕在网络出现分割的情况下,各个独立的子系统都可以继续提供服务 Can Only Choose Two From Above Three CAP Theorem[2] BASE Basic Availability Soft state Eventual Consistency ACID Atomic Consistency Isolation Durability CAP Theorem[3] /visual-guide-to-nosql-systems Cassandra使用案例 Cassandra使用案例[2] user site application others evaluated Jonathan Ellis-3 Rackspace stats collection (testing, almost production),Mail Apps division (early testing) HBase, Hypertable, dynomite,and Voldemort Ryan King Twitter storage for all tweets a custom mysql impl, voldemort, hbase, mongodb, memcachdb, hypertable, and others Edmond Lau Ooyala store and serve our near real-time video analytics data HBase, Cassandra, Voldemort, and some others Joe Stump SimpleGeo real-time location infrastructure scott w Onespot a subset of our data store Tokyo, Voldemort and Riak and Cassandra Vitaly Kushner Astrails a project for one of our clients(early development stage) ? Dan Di Spaltro Cloudkick store monitoring statistics and running analytics over the data ? Eric Lubow ShermansTravel mailing system,social network usage Cassandra and Tokyo Cabinet/Tokyo Tyrant Richard grossman bee.tv smart TV and movies recommendations , index each day all the TV shows in eve