MapR offers Hadoop-as-a-service

The beta of MapR 2.0 has been announced by its developers, MapR Technologies. The new release includes multi-tenancy support within the same cluster and new monitoring services. MapR comes in two editions; M3 is the community-supported edition while M5 is provided on a subscription basis and includes mirroring, snapshots, NFS HA and other features.

The new Amazon EMR with MapR offers both the M3 and M5 editions at different prices per hour and allows customers to create their own custom MapReduce cloud that is tuned for their tasks. Amazon is offering the older MapR 1.2 versions, announced last year, rather than the new beta 2.0 though.

MapR 2.0's new job monitoring and management now graphically displays resource and time usage of jobs and tasks in the system. This lets users track cluster usage and detect where nodes are running slow. The MapR control system can produce line charts and histograms based on a variety of statistics gathered about tasks and jobs. The user also gets the ability to control where jobs and data reside on the cluster to better match with node performance, but this feature is only in the M5 edition.

Other improvements include support for central logging and configuration, selectable compression algorithms and improved security, and new versions of HBase, Hive, Pig and other open source components. SELinux support is among the security enhancements. MapR is now also supported on SUSE Linux Enterprise Server, in addition to Ubuntu and Red Hat Enterprise Linux.

MapR is available to download as pre-compiled distributions, but users are directed to the original Apache-licensed Apache Hadoop projects, HBase, Pig, Flume and so on, for the source code. This does make recreating the actual distribution as open source code somewhat difficult though. Subscription pricing for MapR M5 is available on request from the company.

(djwm)