@JIELOGAN, is your question about which version of Hadoop to use, v1 or v2?
If so, that's not an ES-Hadoop question as the answer depends on which version of Hadoop you are running, 1.x or 2.x. If you're starting new with Hadoop, I'd definitely recommend going with the most current version, 2.8 (assuming you're not limited to a given version via some other requirement)
With regards to write performance from Hadoop into Elasticsearch, see: https://www.elastic.co/guide/en/elasticsearch/hadoop/current/arch.html. In short, the level of parallelism for the writing is based on the number of splits and mappers in your dataset/cluster: more splits == more mappers == more parallelism