whatever works out to be simplest key points are • multiple jobs can be queued, but • only one map/reduce deployment allowed