currentTimeMills() once per batch
added logging for fhr orphan debugging
pom.xml added additional dependencies
Processing messages with a larger batch size (increased from the default of 100 to 1000 for telemetry and 2000 for FHR) has a large positive effect on throughput. More than a 2x speedup for Telemetry and maybe 3-5x speedup for FHR. Increasing Telemetry batch size from 1000 to 2000 didn't cause a significant improvement in tests today (July 11, 2013).