The Apache Foundation has promoted a fast data-processing tool out of the Apache Incubator in a further sign of the maturity of the Hadoop family.
Apache Spark is a fast processing layer for computing data stored within the open-source Hadoop file system or other shared file systems such as NFS. It supports Scala, Java, and Python. In some tests it has demonstrated a speedup of 100 times over Hadoop when dealing with in-memory sets, and 10 times for hard-disk-held data.
On Sunday, Spark was unanimously voted to graduate from the Incubator, and some of those voting included Hadoop luminaries such as the technology's creator Doug Cutting.
Now that Spark has been promoted, a project management committee will be established for the software, and Databricks co-founder and former AMP Lab PHD student Matei Zaharia will be appointed to the role of 'Vice President, Apache Spark".
Like Hadoop, Spark has become the foundation for other data-processing engines as well, such as Shark for SQL-on-Hadoop queries, MLib for machine learning, Spark Streaming for dealing with streaming data, and GraphX for graph processing.
Some of the technology's users include Baidu, Databricks, IBM's Almaden research group, TrendMicro, Yahoo! and Alibaba.
The graduation of Apache Spark caps off a vertiginous rise for the data-processing system, which was created at the University of California at Berkeley's AMPLab in 2009 and was published as open source in 2010.
Since then, the system has gained a vigorous developer community, and more than 120 developers from 25 companies contribute source code. There seems to be enough activity around the software for businesses to smell money – as last week Hadoop hothouse Cloudera announced commercial support for the tool. ®