Round robin is not efficient in our current design.
Another problem is that when the daemon thread finds out desired num is greater than actual num, it will put all data to queue at one time. A better solution would be: add at most N docs every time.
Round robin is not efficient in our current design. Another problem is that when the daemon thread finds out desired num is greater than actual num, it will put all data to queue at one time. A better solution would be: add at most N docs every time.