Glossary · Data, timing and communication
Backoff
Also known as: backoff strategy
In systems engineering, backoff is a controlled delay strategy between retries that prevents overload of a system or repeated conflict between senders. Common forms are fixed, linear and exponential backoff, often with a random component.
- Data and interfaces
- Integrated systems
In one sentence
Backoff spaces out retries so a failed link or overloaded service is not flooded; its limits must be specified for every integrated system.
Example
After the MES server stops responding, the PLC gateway of a packaging line retries after 1, 2, 4 and 8 seconds and then raises an alarm instead of retrying indefinitely.
How it applies
- Integrated systems: Specify the backoff pattern in the interface specification: initial delay, growth factor, maximum delay, maximum number of attempts, and whether a random component is added so that many clients do not retry in lockstep.
- Functional safety: Backoff adds delay. Do not let a function whose timing matters depend on a communication path whose worst-case recovery time, including backoff, has not been counted against the protective response time or other time budgets.
- Fault handling: Backoff must not hide a fault. After the last attempt, a defined fault reaction has to follow.
- Technical documentation: Integration and service manuals should state what users see while a system backs off, for example “Connection lost, retrying,” and when they need to intervene.
Backoff vs. retry
A retry is the repeated attempt; backoff is the waiting rule between attempts. Retries without backoff can turn a short outage into an overload of the system that is trying to recover.