Overview
NUMA is used in large-scale multi-processor systems where a single shared bus (as in SMP) would become a bottleneck. In NUMA, each processor has its own local memory but can still access the memory of other processors over a high-speed interconnect.
Performance Impact
Accessing local memory is much faster than accessing 'remote' memory (memory attached to another processor). Software must be 'NUMA-aware' to ensure data is placed close to the processor that will use it.
Common Use
High-end servers with multiple CPU sockets (e.g., dual-socket EPYC or Xeon systems).