Earlier it was (re)initializing earthengine client upon every ingestion call, which in turn was calling ee highvolume api.
In beam dataflow, we essentially need to initialize earthengine only once in a device, not even a dataflow worker. These changes initializes ee only if necessary. Now it makes 50% less EE calls in filter and ingestion step.
Effect: Increases ingestion speed by roughly 100%.
Earlier it was (re)initializing earthengine client upon every ingestion call, which in turn was calling ee highvolume api.
In beam dataflow, we essentially need to initialize earthengine only once in a device, not even a dataflow worker. These changes initializes ee only if necessary. Now it makes 50% less EE calls in filter and ingestion step.
Effect: Increases ingestion speed by roughly 100%.