We'll need to use three streams simultaneously (we'll need to use Stream.iterator).

1. A stream of existing systems (ordered by hypervisor id, then by inventory id)
2. A stream of existing hypervisors (ordered by hypervisor id)
3. A stream of HBI records (ordered by hypervisor id, then by inventory id)

Iterate through the HBI records, and reference the two existing record streams:

• update the existing system record (or create if needed)
• update the existing hypervisor record (or create if needed)
• delete any existing records that don't have a corresponding record in HBI.
• once we've reach a configurable batch size, flush and clear the entitymanager to reclaim memory, default to 128. 

Done criteria: 

• Given 1 million HBI records in an acct, the tally process should be able to handle that without consuming more than 4GB of memory.