Avoiding insurmountable queue backlogs

Queues offer a promise of reliability for integrating applications, but can cause more subtle operational risks.

Queues exhibit ‘bi-modal behavior’ — low latency when everything is working well, but can rapidly form large backlogs when a failure occurs. This can dramatically increase the recovery time from outages.

Amazon offer an extremely interesting article about the reliability of queuing, with strategies to limit and manage backlogging.