dualaffinity

Dualaffinity is a concept in computer systems design describing a binding policy in which a resource, such as a thread, process, or service, is associated with two simultaneously active affinity domains. The idea is to balance locality and redundancy by allowing operation within two resource groups, which can help preserve data locality while providing failover or load-balancing capabilities.

In practice, dual affinity is most often discussed in the context of CPU affinity and network path selection. For CPU affinity, a thread might be allowed to execute on two cores or two CPU clusters, enabling migration within the two domains to avoid contention while minimizing cache misses. In networking, dual affinity can refer to bindings across two network interfaces or paths to improve throughput and resilience.

Implementation approaches vary. Some systems use explicit affinity masks or policy hints that include two domains; others rely on scheduler heuristics or policy-based scheduling to approximate dual-domain binding. Effective use typically requires consideration of data locality, cache coherence, and the potential for contention between the two domains.

Limitations include increased scheduling complexity, potential disruption to cache locality, and added overhead from tracking activity across two domains. When properly tuned, dual affinity can offer gains in throughput and fault tolerance, particularly on heterogeneous or NUMA-aware architectures.

See also: processor affinity, CPU affinity, affinity masking, NUMA, load balancing, failover.