Microsoft PowerPoint - shudo-NoSQL-data-model ppt

Size: px

Start display at page:

Download "Microsoft PowerPoint - shudo-NoSQL-data-model ppt"

せぴあよどぎみ
4 years ago
Views:

1 NoSQL 1

2 NoSQL, key-value store, documentoriented DB, graph DB, memcached, Bigtable, Dynamo, Amazon SimpleDB, Cassandra, Voldemort, Ringo, VPork, MongoDB, CouchDB, Tokyo Cabinet/Tokyo Tyrant, Flare, ROMA, kumofs, Kai, Redis, Hadoop HBase, Hypertable, PNUTS, Scalaris, Dynomite, ThruDB, Neo4j, IBM ObjectGrid, Oracle Coherence, Velocity, 2

3 NoSQL NoSQL is an umbrella term for a loosely defined class of non-relational data stores that break with a long history of relational databases and ACID guarantees. Data stores that fall under this term may not require fixed table schemas, and usually avoid join operations. The term was first popularised in early (Wikipedia ) RDB ACID ACID: DB atomicity/, consistency/, isolation/, durability/ 1970 CAP eventual consistency 3

4 / GREE: MySQL 12 /load average 2.0 Flare 6 /load average mixi: DB, 1 queries/sec, 1 Tokyo Tyrant/Tokyo Cabinet 2 IBM: RFP msec RDB ObjectGrid / / Google:, TB (2006 ) Bigtable Facebook: RDB memcached, 800, 28 TB (2008/12) Amazon: Dynamo: 4

5 NoSQL OSS memcached map (key-value) Amazon Dynamo Google Bigtable Amazon SimpleDB map (key-value) table table Windows Azure Table table 5

6 atomicity 1 key, 1 key-value (row, item, entity) key-value, key, availability atomic key-value 6

7 memcached in-memory key-value store: map queries/sec () set akey avalue get akey RDB, livedoor, mixi, Facebook, YouTube, Digg, Twitter, Wikipedia, Facebook: TB (2008/12) こういうアレ : amap.put(akey, avalue) value = amap.get(akey) 7

8 memcached: memcached consistent hashing / 8

9 Dynamo key-value store in-memory Berkeley DB, MySQL Amazon SOSP 07 Amazon, : : consistent hashing availability: always-on experience : eventual consistency vector clocks [Lamport1978] Service Level Agreements (SLA) SLA (e.g. ) : 500 req/s 99.9% 300 msec 9

10 Dynamo: ( ) 10

11 Bigtable table atomic :, not sparse, distributed, persistent multi-dimensional sorted map Google OSDI 06 Google File System, Chubby Google Google App Engine (PaaS) : Hadoop HBase, Hypertable : : URL, :,, Google Analytics, Earth, 11

12 Bigtable: row key com.cnn.www com.example.www 2009/12/ /12/ /12/3 contents: <html> <html> <html> <html> anchor:cnnsi.com anchor:my.look.ca anchor: CNN CNN.com Example.com C++ Figure 2 // Open the table Table *T = OpenOrDie( /bigtable/web/webtable ); // Write a new anchor and delete an old anchor RowMutation r1(t, com.cnn.www ); r1.set( anchor: CNN ); r1.delete( anchor: ); Operation op; Apply(&op, &r1); row key Megastore :,, row key, 12 atomic

13 Amazon SimpleDB table atomic :, not Amazon Web Services (AWS) :, REST SOAP over HTTP Dynamo / Amazon SimpleDB automatically indexes your data, creates georedundant replicas of the data to ensure high availability, and performs database tuning on customers behalf. Amazon SimpleDB also provides no-touch scaling. There is no need to anticipate and respond to changes in request load or database utilization; the service simply responds to traffic as it comes and goes 13

14 Amazon SimpleDB: API domain: /table item: /row attribute: /column item attribute void PutAttributes(domain, item, ) REST, SOAP GetAttributes(domain, item, attribute ) : select * from mydomain where Year > 1975 and Year < 2008 item SQL attribute atomic 14

15 Windows Azure Table table atomic :, not :,, Windows Azure Storage Windows AzureTable, Blob, Queue.NET API, LINQ REST : SQL Azure: RDB SQL Azure: SQL Server (RDB) Amazon RDS: MySQL 15

16 Windows Azure Table table entity: /row property: /column property PartitionKey partition RowKey partition entity ID Partition 1 PartitionKey RowKey Date Example Doc V /8/2 Example Doc V /9/28 Description Committed version Alice s version Partition 2 FAQ Doc FAQ Doc FAQ Doc V1.0 V1.0.1 V /5/2 2007/7/6 2007/8/1 Committed version Alice s version Sally s version 16

17 timestamp をキーとする map: Bigtable com.cnn.www com.example.www 2009/12/ /12/ /12/3 contents: anchor:cnnsi.com anchor:my.look.ca <html> <html> <html> <html> CNN com.cnn.www+contents:+2009/12/17 com.cnn.www+anchor:cnnsi.com+2009/ com.cnn.www+anchor:my.look.ca+2009/ <html> CNN CNN.com CNN.com Example.com anchor: com.example.www+contents:+2009/12/17 <html> com.example.www+contents:+2009/12/10 <html> com.example.www+contents:+2009/12/13 <html> Row key + column key + 17

18 : Azure Table Partition 1 Partition 2 PartitionKey RowKey Date Example Doc Example Doc FAQ Doc FAQ Doc FAQ Doc V1.0 V2.0.1 V1.0 V1.0.1 V /8/2 2007/9/ /5/2 2007/7/6 2007/8/1 Description Committed version Alice s version Committed version Alice s version Sally s version V1.0 Example Doc V2.0.1 Date Description Date Description 2007/8/2 Committed 2007/9/28 Alice s version 18

19 Amazon SimpleDB GetAttributes() item, select Windows Azure Table PartitionKey : Bigtable 19

20 RDB, NoSQL SQL 他それぞれ map relational data model Megastore library table (multi-dimensional sorted map) (sorted) records + indices (sorted) map (sorted) map 20

21 NoSQL NoSQL ( ) : message queue 21

22 22

スライド 1

スライド 1 丸山先生レクチャーシリーズ 2010 第 3 回 2010 年 4 月 22 日 ( 木 ) 楽天株式会社本社 key-value ストアの基礎知識首藤一幸 1 内容 key-value ストア NoSQL データストア特徴とその根源さらなるスケーラビリティをデモスライド 30 枚 2 key-value ストアとはキーと値の組を書き込みキーを指定することで値を読み出せるデータベース管理ソフトウェア