Is there a calculation to determine how many daemons are needed based on batch jobs and concurrent users?
The number of daemons you’ll need will depend on the number of addresses you need to process concurrently. An individual address is processed in sub-second response time, and addresses are processed in a FIFO queue fashion (First-In First-Out). So even if a batch job is running for 300K addresses, it will only have one address in the queue at a time since it would be a synchronous transaction, and a user request would still get processed. >>> Some addresses for certain cities may take longer to process than others, based on the data supplied by the user and the number of USPS® addresses in a postal code. If more algorithms are required to correct an invalid address for these larger cities, it may take up to a second to process. (i.e. Some New York City blocks have a larger percentage of addresses per postal code) >>> Overall, we’ve seen throughput as high as 700K addresses per hour for the daemon. >>> For these reasons, we recommend at least 2 daemons per instance for concurrency.
