Stability and reliability issues are related to co-processor use in MapR-DB. In HBase, co-processors become part of the critical path processing for every record. This includes all inserts, updates, and deletes. The API is defined in the HBase code, and it's exposed for use. As such, any problems with the co-processors can cause major instabilities with the whole HBase instance. All region servers in the installation would be affected by any coding issues with the co-processor.
Since MapR-DB is a part of the MapR-FS daemon, any introduction of stability issues would directly affect the stability of the whole cluster. In addition, the HBase co-processor API is a Java API, and it requires an inordinate amount of engineering to 1) expose it from MapR-FS and 2) guarantee that any misbehaving code wouldn't crash the MapR-FS daemon.
For very select applications that require co-processors, we recommend Apache HBase running on a MapR cluster. We think HBase on MapR is better than HBase on stock Apache HDFS.
You forgot security. ;-)
Retrieving data ...