[image: pasted1] (The figure shows S2Graph as lambda architecture view. As mentioned in [S2GRAPH-15] <https://issues.apache.org/jira/browse/S2GRAPH-15>, S2Graph provides a great real-time view with serving layer on HBase.)
Hi folks. I think we'd better implement "the common Spark environment" for batch and speed(streaming) layer. This environment is valuable for loader and s2counter which are currently parts of S2Graph as well as other jobs in the project roadmap(draft: http://markmail.org/message/uigl6pbt6urelgma?q=list:org%2Eapache%2Es2graph). In my opinion, the environment includes three important features: 1. Spark launcher with HA and fault tolerance scheduler using such as Marathon or Chronos. 2. Resumable Kafka stream (at least once or exact once) 3. Configurable Engines like predictionio Thanks Minseok
