NoSQL databáze

Definition:

NoSQL databases = Next generation databases mostly adressing some of the points: being non‐relational, distributed, open‐source and horizontally scalable. The original intention has been modern web‐scale databases. Often more characteristics apply as: schema‐free, easy replication support, simple API, eventually consistent, a huge data amount, and more.

Features

• goal of NoSQL databases is to respect the real-world nature of data
  • move from traditional relation databases to NoSQL
• scaling-out instead of scaling-up (this is the traditional approach, vertical scaling)
  • ⇒ distributed systems across many hosts
• sharding, replication (Distribuce a Replikace)
• a lot of processes are automated, there is no need for highly skilled relational database administrators
  • automatic recovery, automatic distribution, tuning…
• BASE model (more relaxed consistency)
  • ACID represents traditional and strong consistency
• schemalessness
  • in relational databases the schema is stricly enforced
  • in NoSQL the schema is relaxed or completely missing
    • but there is always some implicit schema (we have to maintain some schema to match with the applications that are using the database)
    • design changes are easier
• they are often open-source (community and enterprise versions)
• the API is also really simple, often on the base of HTTP or similar (stateless interface)

Challenges

• a lot of NoSQL databases are still pre-mature with key features missing
• they are open-source, no strong support
• sometimes difficult to install and maintain
• limited support for business intelligence tools for analytics and ad-hoc queries
• no enough big data experts on the market

Typy NoSQL databází

Key-Value databases

• Redis, Valkey, Scalaris, Riak KV Dokumentově orientovaná databáze / document stores
• MongoDB, CouchDB Sloupcové databáze / wide-column stores
• Apache Cassandra, Apache HBase Grafové databáze
• Neo4j

Další druhy mohou být:

• Native XML databáze (XPath, XQuery, XSLT)
  • dokumenty jsou v kolekcích (podobně jako v Dokumentově orientovaná databáze)
  • stromová struktura (XML), nested elements, attributes, text values
  • Sedna, Tamino, MarkLogic
  • XML
• RDF databáze (případ grafových databází)
  • dotazovací jazyk SPARQL
  • triples (subject, predicate, object) can be viewed as graphs
  • Apache Jena, MarkLogic, rdf4j

Aggregates

• aggregate je datová jednotka s komplexní strukturou, je to kolekce dat, které chceme z nějakého důvodu držet při sobě
• příklad: value v key-value stores, dokument v MongoDB, column family v sloupcových databázích
• většinou se databáze pak designují, aby splnily atomicity ve smyslu “one aggregate at a time”
• jsou dva typy NoSQL databáze:
  • aggregate-ignorant: relational, graphs
  • aggregate-oriented: key-value, document stores, wide column stores