Couchbase 2.0 released; implements JSON document store

Version 2.0 of the Couchbase Server NoSQL database reaches GA, adds document store capabilities to the product's key-value store roots.
Written by Andrew Brust, Contributor

Couchbase is a prominent NoSQL database that, along with MongoDB, explicitly targets Big Data applications.  Although based on technology from the Apache CouchDB document store project, Couchbase itself had been implemented instead as a key-value store (read on for an explanation of both terms).  But with version 2.0, Couchbase itself adopts the document store model, just like its NoSQL, Big Data competitor, MongoDB.

All NoSQL databases allow for variability in data structure from record to record.  Key-value stores organize data into data couplets, with the first member of each couplet storing an identifier and the other storing a value associated with it (e.g. {CUSTOMERNAME, "BRUST"}).  Document stores host more complex structures using JavaScript's Object Notation (JSON).

With the change in architecture, Couchbase gains a few significant features, including:

Secondary indexes and views: wherein data can be indexed on more than just its key, and where pre-written "views" can serve up data, rather than forcing developers to query and process the raw data directly.

Incremental Map Reduce: Couchbase's Map Reduce isn't quite the same as Hadoop's, but it serves some of the same purposes.  Map functions pre-process data, allowing specific structure to be imposed at query time.  Reduce functions allow for summarization, or aggregation, by key of the data emitted form Map functions.  This provides for analytics to be served up right form the database.

Cross-data center replication (XDCR): while not derived from the document store architecture per se, this feature is nevertheless a significant one.  Couchbase server now replicates data not just between nodes in cluster that are in the same data center, but across data center boundaries as well.  This enhances availability, redundancy and, much as a content data network (CDN) would, it also enhances locality of data to users.

What it isn't good for
NoSQL data stores work especially well for non-transactional, high-scale Web and mobile applications, rather than transactional line-of-business (LOB) apps.  Relational databases are still far superior for LOB applications, but for applications which track data like user profiles, catalog items and other data of varying structure, NoSQL databases work well, and the prevalence of clustered architectures among NoSQL databases is especially well-suited to global, consumer-facing sites and back-end infrastructure.

Real deal
Couchbase has customers like McGraw Hill Education and Orbitz on deck with testimonials about the 2.0 version of the product, which has been in Beta for some time now, making clear this technology is not just for startups.  

NoSQL is more than just a fad, and its importance in the Big Data space in undeniable.

Editorial standards