Demystfying nosql databases
Download as PPTX, PDF0 likes79 views
This document provides an overview of NoSQL databases, including what they are, common types, uses cases, and examples. It defines NoSQL as distributed, schema-less databases and discusses four main types: columnar, document, key-value, and graph databases. The document also examines specific NoSQL databases like HBase, MongoDB, Redis, and Neo4j, providing information on their architectures, features, and how they are used. General recommendations are made around reading documentation, design reviews, and isolating NoSQL clusters.
Demystfying nosql databases
- 1. Demystfying Nosql Databases Mike King & Matt Thomas Enterprise Technologists, Big Data
- 2. 2 Dell - Restricted - Confidential What are databases? • Tedd Codd & Chris Date – 13 rules – An Introduction to Database Systems • Wikia/Wikipedia • Mike – An organized collection of data offering varying levels of availability, scalability, performance, consistency, management, accessibility and quality. • Matt Databases defined
- 3. 3 Dell - Restricted - Confidential What types of databases exist? • Network – Adabas • Hierarchical – IMS • Relational – PostgreSQL • Object Oriented – Versant • Nosql – MongoDB, Hbase • Newsql – VoltDB, MemSQL • XML – MarkLogic, Xyleme
- 4. 4 Dell - Restricted - Confidential Nosql background, issues and considerations • History – Google Big Table, Amazon Dynamo • What does schema-less mean? – On read – Still structured – Embedded – Can vary between records • Languages & formats used – Java, Python – JSON, BSON, XML, CSV
- 5. 5 Dell - Restricted - Confidential Nosql background, issues and considerations continued • Eric Brewer’s CAP theorem – Can’t do all three. • What does NoSQL really mean? – Distributed, shared-nothing aggregate oriented database – “Not only SQL” versus “No” • What are the factors for the various choices? – Best fit – Use case(s) – KV – HA, Multi-site – Network – Kevin Bacon • Sharding – Partitioning
- 6. 6 Dell - Restricted - Confidential NewSQL • SQL as predominant access method • OLTP • Larger user populations than nosql • Better consistency than nosql • Still subject to Brewer’s CAP theorem • Examples – VoltDB, MemSQL, Clustrix, NuoDB
- 7. 7 Dell - Restricted - Confidential RDBMS or NOSQL?-tablify • RDBMS – Large user populations – Structured – Static schema – Strong typing – Access by PK, AK, indexes – Complex structures – Feature rich – Multi-purpose, shared by apps – OLTP – ACID – Complex queries – >3 way joins – Small to medium sized dbs – COTS pkgs – Datamarts • Nosql – Smaller user populations – Multi-structured – Schema evolution – Weak typing – Mostly random access by PK – Simple structures – Bare bones functionality – Single purpose/use case, not shared by apps – Not transactional – BASE – Simple queries – VLDB – Horizontal scalability
- 8. 8 Dell - Restricted - Confidential NoSQL Database Types • Four types – Columnar – Hbase, Cassandra – Document – MongoDB, Couchbase – KV – Riak, Redis – Graph – Neo4j, Titan • How many do you need? – By type – Within type • Who will manage them? – DBAs • How do you access them? – SQL, nosql – Sequential
- 9. 9 Dell - Restricted - Confidential Nosql Commonalities • Mostly open source • Weak typing • Multi-structured • Horizontal scale • No standardization • VLDB • Single purpose, per database
- 10. 10 Dell - Restricted - Confidential Nosql Differences • Access • Formats supported • Features • Management • Administration • VLDB • Performance & tuning • Resource consumption • Language bindings • APIs • Security • Persistence • Programmability • ?Schemas
- 11. 11 Dell - Restricted - Confidential How are nosql databases typically used? • As an adjunct to Hadoop • As a partial replacement for some RDBMS workloads • To scale linearly • As a data store for semi-structured and multi-structured data
- 12. 12 Dell - Restricted - Confidential What questions do our customers ask? • Why is my Hbase cluster so CPU hungry? • Do you have an RA for <Your favorite nosql db goes here>? • Can I replace all my Oracle databases w/ some nosql databases?
- 13. 13 Dell - Restricted - Confidential What are some common problems? • Cohabitation with Hadoop and other programs on a cluster. • Poor db design • Falling prey to vendor hype
- 14. 14 Dell - Restricted - Confidential How about some general recommendations? • Read a book or two on your target nosql db. • Search thru the blogosphere & twitterverse. • Don’t use more than one type, unless you’re an SI or large service provider. • If performance & service levels are important isolate the cluster. • Review your database design w/ DBAs & those that have done it already. – Presentations, conference proceedings, boutique consultancies
- 15. 15 Dell - Restricted - Confidential Nosql Examples, Diving Deeper • Hbase • MongoDB • Redis • Neo4j
- 16. 16 Dell - Restricted - Confidential Hbase • Columnar – Column families • Uses ZK • Has a master • WAL • Region servers • Memstore • Hfiles • HDFS • Uses jvm heap • Access – Row key – Get – Put – Scan – Bulk load • Design – Beware of skew – Tune for peaks • Perf – CPU intensive – Very fast for puts & gets by key
- 17. 17 Dell - Restricted - Confidential Neo4j • Property graph – Nodes, edges, relationship/arc, direction, data/properties(node & arc) – Edge labeled multi-digraph • REST API • ACID • Fast, scaleablable lookups • Lucene index for search
- 18. 18 Dell - Restricted - Confidential Our Contact Info • Mike_King2@dell.com • @MikeDataKing • 901-262-7918 • Matt_Thomas@Dell.com • ?twitter? • 904-429-6709
Editor's Notes
- #3: https://en.wikipedia.org/wiki/Codd's_12_rules
- #4: https://en.wikipedia.org/wiki/NewSQL https://en.wikipedia.org/wiki/XML_database
- #5: ? Aggregate orientation VS relational?
- #9: https://itsavant.wordpress.com/2013/04/23/can-you-get-by-with-just-one-nosql-database/ https://en.wikipedia.org/wiki/NoSQL