)]}'
{"specs/victoria/approved/numa-topology-with-rps.rst":[{"author":{"_account_id":9542,"name":"Pavlo Shchelokovskyy","email":"pshchelokovskyy@mirantis.com","username":"pshchelo"},"change_message_id":"d4ad26148d9cc54363aa45422dc66d1e3783107f","unresolved":true,"context_lines":[{"line_number":139,"context_line":"                                   +-------------------------------+"},{"line_number":140,"context_line":"                                   |  (no specific traits)         |"},{"line_number":141,"context_line":"                                   +--+---------------------------++"},{"line_number":142,"context_line":"                                      |                           |"},{"line_number":143,"context_line":"                                      |                           |"},{"line_number":144,"context_line":"               +-------------------------+                   +--------------------------+"},{"line_number":145,"context_line":"               | \u003cNUMA_NODE_O\u003e           |                   | \u003cNUMA_NODE_1\u003e            |"}],"source_content_type":"text/x-rst","patch_set":1,"id":"d3f3d1ba_d3663e90","line":142,"updated":"2024-12-17 18:15:07.000000000","message":"I\u0027d request an additional layer or RPs called \u0027socket\u0027 or smth like that.\n\nGenerally, cross-socket memory access is considerably slower than in-socket.\nIf I have a multi-socket compute node, I may want to express a need to only use numas from the same socket in my VM.\nThis should work with the same `same_subtree` query of placement.\n\nAdditionally, we may want to look into SLIT https://stevescargall.com/blog/2022/11/linux-numa-distances-explained/ to at least sort candidates according to the SLIT distance, and try to choose the closest ones.","commit_id":"e290c1fcacea614fd0be5ad2dd8f25b19f5487bd"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"0a2aa5a78da77bdde635b2f9e5c45a594ff569b6","unresolved":true,"context_lines":[{"line_number":139,"context_line":"                                   +-------------------------------+"},{"line_number":140,"context_line":"                                   |  (no specific traits)         |"},{"line_number":141,"context_line":"                                   +--+---------------------------++"},{"line_number":142,"context_line":"                                      |                           |"},{"line_number":143,"context_line":"                                      |                           |"},{"line_number":144,"context_line":"               +-------------------------+                   +--------------------------+"},{"line_number":145,"context_line":"               | \u003cNUMA_NODE_O\u003e           |                   | \u003cNUMA_NODE_1\u003e            |"}],"source_content_type":"text/x-rst","patch_set":1,"id":"4fe9e2c4_16f93dda","line":142,"in_reply_to":"d3f3d1ba_d3663e90","updated":"2025-01-06 11:21:07.000000000","message":"this was approved for directionality but is not actully being worked on at preestnt.\n\nwe are aware of numa distance and that you can have multiple numna nodes per socket\n\nat this time we do not plan to proceed with modeling numa in placmenent in the next 12 months\n\nwe may consider numa distanct based modeling in nova in that time period but that unlikely. we do have some prototypes of that howeever.\n\nhttps://review.opendev.org/q/topic:%22bp/libvirt-smarter-cpu-placement%22\n\nagain that was depritied and no one is currenlty workign on that in this cycle.\nwe may look at that again next cycle.\n\ncurrently there are know scalablity problems with placment and nested resouce providers that prevent use form progressing numa in placment or at least raise some concerns https://review.opendev.org/c/openstack/nova-specs/+/938070\nso we are hesitent to do such a large refactor until we adress that.\n\nthe existing numa distance work does not depend on placement which is why that woudl be more viable to enabel then reviving this spec at this time.","commit_id":"e290c1fcacea614fd0be5ad2dd8f25b19f5487bd"}]}