+ Follow

NoSQL databases

by Harri Kauhanen

26,133 views

NoSQL database paradigms, how to query NoSQL, about relationships, comparison with RDBMS

Accessibility

View text version

Upload Details

Uploaded via SlideShare as Apple Keynote

Usage Rights

File a copyright complaint

25 Embeds 2,876

http://blog.nosqlfan.com	1678
http://nosql.mypopescu.com	304
http://spf13.com	214
http://www.slideshare.net	195
http://www.cxybase.com	111
http://www.kellblog.com	105
http://www.lifeyun.com	90
http://kellblog.com	79
http://www.v1r3n.com	56
http://zmh-12.blogspot.com	10
http://bbs.bingyan.net	7
http://www.linkedin.com	6
http://irr.posterous.com	6
http://translate.googleusercontent.com	3
https://si0.twimg.com	2
http://cache.baidu.com	1
https://socialtext.allmerica.com	1
http://twitter.com	1
http://feeds.feedburner.com	1
http://localhost	1
http://webcache.googleusercontent.com	1
https://server16.kproxy.com	1
http://www.hanrss.com	1
http://www.taaza.com	1
http://coderwall.com	1

More...

Statistics

Likes: 74
Downloads: 1,302
Comments: 6
Embed Views: 2,876
Views on SlideShare: 23,257
Total Views: 26,133

NoSQL databases Presentation Transcript

Harri Kauhanen 2010-03-09 NoSQL databases

Database paradigms • Relational (RDBMS) • NoSQL • Key-value stores • Document databases • Wide column stores (BigTable and clones) • Graph databases • Others

Relational databases • ACID (Atomicity, Consistency, Isolation and Durability) • SQL • MySQL, PostreSQL, Oracle, ...

dog bark id: integer id: integer name: varchar dog_id: integer mood: varchar bark: text birthdate: date color: enum comment id: integer bark_id: integer dog_id: integer comment: text id name mood birth_date color 12 Stella Happy 2007-04-01 NULL 13 Wimma Hungry NULL black 9 Ninja NULL NULL NULL

Key-value stores • “One key, one value, no duplicates, and crazy fast.” • It is a Hash! • The value is binary object aka. “blob” – the DB does not understand it and does not want to understand it. • Amazon Dynamo, MemcacheDB, ...

“value” “key” ... name_€#_Stella^^^ mood_€#_Happy^^^ dog_12 birthdate%/// 135465645) ...

Document databases • Key-value store, but the value is (usually) structured and “understood” by the DB. • Querying data is possible (by other means than just a key). • Amazon SimpleDB, CouchDB, MongoDB, Riak, ...

“document” “key” { type: “Dog”, name: “Stella”, dog_12 mood: “Happy”, birthdate: 2007-04-01 } vs. id name mood birth_date color 12 Stella Happy 2007-04-01 NULL 13 Wimma Hungry NULL black 9 Ninja NULL NULL NULL

{ type: “Dog”, name: “Stella”, mood: “Happy”, birthdate: 2007-04-01, barks: [ { bark: “I had to wear stupid..” comments: [ { dog_12 dog_id: “dog_4”, comment: “You look so cute!” }, { dog_id: “dog_14”, comment: “I hate it, too!” } ] } ] }

Wide column stores • Often referred as “BigTable clones” • "a sparse, distributed multi-dimensional sorted map" • Google BigTable, Cassandra (Facebook), HBase, ...

“column” “row-id” “column family” “title” “time” “value” dog_12 dog birthdate 15 2007-04-01 dog mood 11 Angry dog mood 45 Happy dog name 25 Stella dog color 34 Black bark text 11 I had to wear...

Graph databases • “Relation database is a collection loosely connected tables” whereas “Graph database is a multi-relational graph”. • Neo4j, InfoGrid, ...

Dog Stella type name mood dog_12 barks bark_59 I had to wear stupid... Happy birth_date comment_to 2007-04-01 comment_83 You look so Cute comments dog_4

• Relationships in RDBMS are “weak”. • You may “define” one by using constraints, documenting a relationship, writing code, using naming conventions etc. • Relationships in graph databases are first class citizens. • There are no relationships in key-value stores, document databases and wide column stores. • You may “define” one by using validations, documenting a relationship, writing code, using naming conventions etc.

• Relational databases have almost limitless indexing, and a very strong language for dynamic, cross-table, queries (SQL) • That’s why they handle all kinds of relationships well and dynamically. • NoSQL databases... • ...might have limited support for dynamic queries and indexing • ...don’t support JOIN like operations of SQL • ...but you can store some relationships into document itself

How to query NoSQL? • Key-Value • Row-id/column-family:title[/time] • “stella_12”/”dogs”:”name” → Stella • Graph traversal • API • Query-language • Integration to indexing and search engines • Map-Reduce

Map-Reduce • “MapReduce is a programming model and an associated implementation for processing and generating large data sets.” • Often JavaScript (NoSQL implementations)

Map-function function map(doc) { if (doc['type'] == 'Dog') { emit(doc['mood'], doc['birthdate']); } } • Generates “indexed view” of data/ documents • This view is just another hash, but both key and value can be “anything”

Reduce-function • Aggregate results for a “view” (after the Map-function) function reduce(mood, listOfBirthdates) { return averageBirthDate(listOfBirthdates); } • Map-phase is easy to distribute, but you is also easy to write poor Reduce-functions

Theorems • CAP • Consistency, Availability, Partition tolerance • “Pick two” • N/R/W (adjusting CAP)

No consistency?

Eventual consistency

Why NoSQL? • Schema-free • Massive data stores • Scalability • Some services simpler to implement than using RDBMS • Great ﬁt for many “Web 2.0” services

Why NOT NoSQL • DRBMS databases and tools are mature • NoSQL implementations often “alpha” • Data consistency, transactions • “Don’t scale until you need it”

RDBMS vs. NoSQL • Strong consistency vs. Eventual consistency • Big dataset vs. HUGE datasets • Scaling is possible vs. Scaling is easy • SQL vs. Map-Reduce • Good availability vs.Very high availability

Questions?

zjj3143 good ! 1 year ago

Are you sure you want to

Sachin Sachdeva, Solution Architect at Nokia Siemens Networks Really good and crisp, 1 year ago

Arnaud ZIEBA, consultant at nomad simple comparison between NoSQL and RDBMS paradigm - nice intro 2 years ago

Harri Kauhanen, Web Developer, Freelancer at Humaneq Thanks, Arto, for mentioning that as a comment here.

My knowledge of graph databases is actually quite weak, and thanks to you, I learned something I perhaps should have known while preparing the presentation :) 3 years ago

Arto Bendiken, Founder at Datagraph Harri, as far as graph databases go, the SPARQL query language would have deserved a mention. It is a W3C-standardized web-scale graph query language that is distinctly more powerful than SQL and is now becoming the lingua franca for the emerging Linked Data web. 3 years ago

Anders Nawroth, Community Manager at Neo Technology Really nice to see the descriptions on slide #14. A widespread misconception is that you actually define relationships in a RDBMS. In reality, you set up constraints to check if a given value matches an existing key. In each and every select statement you'll have to tell the RDBMS how the entities are connected, as the DB doesn't actually handle this for you. In SQL you deal with keys, not relationships. - Enough RDBMS ranting for now :-) 3 years ago