Hello everyone,
I decided to make this its own thread as opposed to a discussion off a TP3
Hadoop question from Don Dra(p)e(r).
TinkerPop is not interested in maintaining numerous connectors to various
vendors (and their various versions). After TP2, we decided that vendors will
maintain their own connectors to TP3. TP3 would only provide 3 "reference
implementations" -- TinkerGraph (POJO in-memory OLTP and OLAP), Neo4j (OLTP),
and Hadoop (OLAP -- Giraph/Spark).
Currently we only support Hadoop1 in TP3. However, there is good reason to
support Hadoop2.
1. SparkServer works inside YARN.
2. All the major Hadoop distributions (Cloudera, Hortonworks, etc.) are
using Hadoop2.
3. Hadoop2 is newer than Hadoop1.
I am more than happy to bump TP3 support to Hadoop2, but am not willing to
support both Hadoop1 and Hadoop2 (as that means we will then have
Giraph1/Giraph2, Spark1/Spark2, etc.). With that, can people reply to this
email with:
+1 (please have TP 3.1.0 support Hadoop2)
-1 (please have TP 3.1.0 support Hadoop1) [keep it the same]
If you wish to provide a comment with your +1/-1, please do.
Thanks everyone,
Marko.
http://markorodriguez.com
