what stands for NUMA ?

what stands for NUMA ?

Non-uniform memory access (NUMA) is a new server CPU and memory design architecture. Under the traditional server architecture, the memory is put into a single storage pool, which works well for single-processor or single-core system. However, this traditional way of universal access will cause resource contention and performance issues when multi-cores access the memory space at the same time. After all, CPU should be able to access all of the server memory, but do not always remain occupied. In fact, CPU only needs to access the memory space required by workload when actual running.

Therefore, NUMA has changed the way memory presents CPU. This isachieved by the partition of each CPU memory of the server. Eachpartition (or memory block) is called NUMA node, and the processorassociated with this partition can access NUMA memory faster, anddoes not need to compete with other NUMA nodes for resources on theserver (other memory partitions are allocated to other processors).

The concept of NUMA is associated with cache. Processor speed ismuch faster than memory, so the data is always moved to a faster localcache, where the processor access speed is much faster than thecommon memory. In essence, NUMA configures a unique overallsystem cache for each processor, reducing contention and delay whenmulti-processors try to access a unified memory space.
More details here:
https://www.sangfor.com/Uploads/ ... r%202016%20June.pdf

Non Uniform memory access is called NUMA.