Memory affinity management

A computer-implemented method for memory affinity management in a shared pool of configurable computing resources that utilizes non-uniform memory access (NUMA), the method comprising:

• monitoring an access relationship between a set of hardware memory components and a set of software assets;
• storing a set of memory affinity data which indicates the access relationship between the set of software assets and the set of hardware memory components;
• determining, using the set of memory affinity data, a NUMA utilization configuration with respect to the set of software assets;
• executing, based on the NUMA utilization configuration, a set of accesses pertaining to the set of software assets and the set of hardware memory components;
• monitoring the access relationship between the set of hardware memory components and a set of containers which utilizes an operating-system-level virtualization technique;
• storing the set of memory affinity data which indicates the access relationship between the set of containers and the set of hardware memory components;
• determining, using the set of memory affinity data, the NUMA utilization configuration with respect to the set of containers which utilizes the operating-system-level virtualization technique;
• executing, based on the NUMA utilization configuration, the set of accesses pertaining to the set of software assets and the set of hardware memory components;
• tracking, using a set of identifiers related to the set of software assets, a set of events corresponding to the access relationship between the set of software assets and the set of hardware memory components;
• detecting a set of cache misses;
• ascertaining, based on and in response to detecting the set of cache misses, a respective identifier of the set of identifiers which relates to a respective software asset of the set of software assets, and
• corresponds to the set of cache misses;
• collecting, using the set of cache misses, the set of memory affinity data;
• monitoring, using a hardware unit which is coupled with a hardware processor component, the access relationship between the set of hardware memory components and the set of software assets;
• storing, in a special purpose register by the hardware unit, the set of memory affinity data which indicates the access relationship between the set of software assets and the set of hardware memory components;
• structuring the hardware unit to include a hardware processor performance monitoring unit (PMU);
• configuring the hardware processor PMU to monitor a set of events corresponding to the access relationship between the set of software assets and the set of hardware memory components;
• collecting, based on the set of events, the set of memory affinity data;
• receiving a set of access requests from a set of hardware processors;
• capturing, pertaining to the set of memory affinity data and in response to receiving the set of access requests from the set of hardware processors, a set of identifiers of the set of hardware processors;
• carrying-out, by a hardware processor memory management unit (MMU) logic, a set of translations of a set of effective addresses to a set of real addresses;
• caching, in a set of effective-to-real address translation (ERAT) caches, the set of translations which pertain to the set of memory affinity data;
• receiving a new access request from an individual hardware processor of the set of hardware processors;
• identifying, using the set of effective-to-real address translation (ERAT) caches, the individual hardware processor of the set of hardware processors;
• logging, pertaining to the set of memory affinity data, the new access request;
• normalizing the set of memory affinity data to determine the NUMA utilization configuration with respect to the set of software assets;
• weighting the set of memory affinity data to determine the NUMA utilization configuration with respect to the set of software assets;
• and
• determining, based on and in response to both normalizing the set of memory affinity data and weighting the set of memory affinity data, the NUMA utilization configuration with respect to the set of software assets.