TrustRadius
Mapr - a small review!
https://www.trustradius.com/hadoop-relatedMapRUnspecified8.615101
No photo available
December 02, 2015

Mapr - a small review!

Score 8 out of 101
Vetted Review
Verified User
Review Source

Overall Satisfaction with MapR

My team was maintaining multiple Hadoop clusters on a high UCS hardware configuration powered by MapR. We were also maintaining a big cluster in a production environment and other clusters for development, QA, disaster recovery and POC. All clusters were configured with high availability. Multiple internal teams used to run their application jobs on our cluster. My team was responsible for managing and maintaining these clusters. We evaluated and implemented new big data and related tools introduced by Mapr. The goal was to make sure application customers using our cluster stay happy. Day to day jobs on our cluster include traditional Java MapReduce, Streaming, Pig, Hive, Mapr-Tables and other in-memory application jobs like spark for analytics on our company internal data. Nearly a total of 50 different use cases using Hadoop were implemented in our various clusters. At times we required getting support from Mapr on complex issues which could not be resolved by my team.
  • Out of the box high availability on multiple Hadoop services, which will really bring enterprise standards. High availability of JobTracker, CLDB in Hadoop 1.x, HA for Impala services etc. Less headache for my team when it comes to service failure.
  • Performance enhancements when migrated from Hbase to Mapr Tables.
  • HDFS-NFS integration pioneer.
  • Volume concept of HDFS storage allocation which could be controlled from MCS GUI was great.
  • It takes time to get latest versions of Apache ecosystem tools released as it has to be adapted.
  • When you have issues related to Mapr-FS or Mapr Tables, its hard to figure them out by ourselves.
  • Sometime new ecosystem tools versions are released without proper QA.
  • Less manual intervention for maintaining a cluster.
When we were shopping, Mapr had the momentum, high availability even on Hadoop 1.x, an improved file system and better a central control system. Now it looks like the situation has changed a lot.
Choose according to your use case. In our situation, we never had to worry about service interruptions. When we kicked off a couple of years ago, Mapr was doing excellent work. Support is not great, [you'll] be lucky to get good support engineers on request. If you have issues with MaprFS or Mapr Tables, never waste time looking into it.