concept: noSQL

RDBMS are hard to do below functions efficiently in a distributed system

• Transactions
• Referential Integrity: Joins

Polymorphism

• aka. schemaless
• This eliminates the need for maintaining a central system (data) catalog or updating an object-relational mapper (ORM)
• downstream techniques, e.g. SCD types and design, will not be available, because every dimension is already joined to the fact data in extremely denormalized ways.

Key-value stores

based on Dynamo paper published by Amazon at the ACM Symposium on Operating Systems Principles

Vendor

Riak

• Riak implements the principles from Amazon's Dynamo paper
• Written in Erlang
• pluggable storage backend for its core storage
  • Bitcask is default
  • LevelDB
• open source in 2017

Redis

Aerospike

Column-family stores

A very wide, sparsely populated table structure

• It includes a number of families of columns that specify the keys for this particular table structure.
• Journey is started from BigTable paper in 2006 at the Operating Systems Design and Implementation (OSDI)

Use cases

• In many cases, these systems are combined with batch-oriented systems that use Map/Reduce as the processing model for advanced querying
• Like Document store with table schema

Vendor

• GCP BigTable
• Apache Cassandra
• HBase

Document stores

• Document can be an XML, JSON, YAML, LibreOffice/MS 365
• Document is stored and retrieved in a schema-less fashion

Vendor

• MongoDB
• Apache CouchDB
• MarkLogic
• Virtuoso

Graph databases