Hacker News new | past | comments | ask | show | jobs | submit login

Just to point out, MapR isn't a slam dunk for a couple reasons. Their HBase implementation is very fast, but it involves using their whole Hadoop distribution. In my experience:

- the management UI isn't as good as Cloudera Manager, or even Ambari in HDP2.0

- 3rd party support isn't there. If you want to use anything you find on GitHub, get ready to add support for MapRFS, and revert the API version back to 0.21

- there are bugs. Weird, specific bugs in the filesystem implementation which will bite you once every few weeks. This is one area where they lag the ASF project: many eyes do make these kind of things shallow. Their smaller install base and fewer developers means these things take a long time to notice and fix.




Their HBase implementation is very fast, but it involves using their whole Hadoop distribution...3rd party support isn't there. If you want to use anything you find on GitHub, get ready to add support for MapRFS, and revert the API version back to 0.21

The MapR team is sharp and responsive. It's founded by ex-Googlers from the search-infrastructure team (http://www.wired.com/wiredenterprise/2011/12/ex-google-man/).

M7 Tables are designed to be a drop in for HBase, with better performance and without the HBase complexity (it's much easier to run M7 Tables as the primary datastore for online apps).

M7 Tables support and certification for several key projects is coming along...

Spark 0.7 already works with MapR M7 Tables with the Hive images that were released with the Eco-1310 (http://doc.mapr.com/display/components/Hive+Release+Notes, http://answers.mapr.com/questions/7803/shark-spark-over-mfs).

MapR is in the process of certifying Titan (https://github.com/thinkaurelius/titan/wiki), and it's going is going through final QA right now (https://groups.google.com/d/msg/aureliusgraphs/RTeFVssIvoI/m...).




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: