Spark is a useful technology for processing of streams and running map-reduce within the memory of a cluster. Thus it is important to fold it into the streaming augmentation of OODT.
Currently, the wengine runner extension point seems to be the best place to put this. In order for the project to be fully backwards compatible, some processing point in OODT must allow multiplexing of jobs (some are run on mesos via resource manager, some are run in Spark, etc.).