The x86/cache updates for the in-development Linux 6.14 kernel are extending the Resource Control “resctrl” code for Intel Resource Director Technology / AMD Platform Quality of Service to have the capability of total memory bandwidth monitoring.
At present the Resource Control code allows the user to specify memory bandwidth limits for each domain of a control monitoring group. The code for Linux 6.14 allows now specifying this on a total memory bandwidth basis rather than being limited to each domain. The motivation here is that some systems may only support total memory bandwidth monitoring but nothing more fine-grained and for some large workloads across multiple NUMA domains you may be more interested in just the total memory bandwidth. Thus resctrl is now having a “total bandwidth event” added for Linux 6.14.
This resctrl work for total memory bandwidth monitoring has undergone a number of rounds of code review by Intel engineer Tony Luck.
More details for those interested via the x86/cache pull request.
