After three developer previews, six release candidates and over 1500 closed tickets the Apache foundation has announced version 1.0 of Apache HBase, a NoSQL database in the Hadoop ecosystem. After more than 7 years of active development, the team behind HBase felt that the project had matured and stabilized enough to warrant a 1.0 version.
The 1.0 release focused on three major themes, the first of which was laying a stable foundation for the future of the 1.x line. To that end the 1.0 version introduced new APIs and deprecated some of the commonly-used client side APIs. This work will allow the HBase project to guarantee API compatibility for all public classes for the duration of the 1.x line.
The second focus of the release was on stabilizing the HBase cluster and its clients. This work took shape over two major initiatives. The first was an initial, experimental release of high availability for reads. This feature enables read only region replicas which make it possible to provide consistent, low latency reads even in the face of region failures lasting 10s of seconds. The second was a forward port of the work Facebook has done to enable online config changes for a subset of the server configuration.
The final focus of the 1.0 release was on making versioning and compatibility explicit. To this end HBase has adopted semantic versioning starting with the 1.0 release and continuing forward. This will allow HBase to user their version numbers to be explicit about the compatibility guarantees.
HBase, started as a contrib project inside of Hadoop in 2007 and in May of 2010 it graduated to its own top-level Apache project. HBase was originally derived from Google's Bigtable paper and now underlies a number of other open source projects such as OpenTSDB, a open source time series database, and Apache Phoenix, an SQL layer on top of HBase.