Amazon.com, Inc.
Managed scaling of a processing service

Abstract:

A system implements managed scaling of a processing service in response to message traffic. Producers produce messages or other data and the messages are stored in a queue or message system. On behalf of consumers of the messages, workers of a client of the queue poll the queue or message service to obtain the messages. For example, a primary worker of the client polls the queue for messages and upon receiving a message, activates a secondary worker from a pool of secondary workers to start polling the queue for message. Now both workers are obtaining messages from the queue, and both workers may activate other secondary workers, exponentially scaling the message processing service in embodiments. When a secondary worker receives an empty polling response, the secondary deactivates back to the pool. The primary thread does not deactivate, even when empty polling responses are received.