Bug 1271852 - [RFE][nova]: Huge Pages per guest NUMA nodes
Summary: [RFE][nova]: Huge Pages per guest NUMA nodes
Alias: None
Product: Red Hat OpenStack
Classification: Red Hat
Component: openstack-nova
Version: 7.0 (Kilo)
Hardware: Unspecified
OS: Unspecified
Target Milestone: ga
: 11.0 (Ocata)
Assignee: Sahid Ferdjaoui
QA Contact: nlevinki
URL: https://blueprints.launchpad.net/nova...
Whiteboard: upstream_milestone_next upstream_defi...
Depends On:
Blocks: 1341176 1414580
TreeView+ depends on / blocked
Reported: 2015-10-14 21:33 UTC by Stephen Gordon
Modified: 2019-09-09 13:37 UTC (History)
12 users (show)

Fixed In Version:
Doc Type: Enhancement
Doc Text:
Clone Of:
Last Closed: 2016-09-23 12:12:33 UTC
Target Upstream Version:

Attachments (Terms of Use)

System ID Private Priority Status Summary Last Updated
OpenStack gerrit 374856 0 None ABANDONED virt: huge pages per guest numa nodes 2020-07-14 11:45:56 UTC

Description Stephen Gordon 2015-10-14 21:33:37 UTC
Cloned from launchpad blueprint https://blueprints.launchpad.net/nova/+spec/virt-hugepages-per-guest-numa-nodes.


The current huge Pages implementation build on top of NUMA takes advantage of memory placement to specify an amount of memory used on each NUMA nodes of guest.

This blueprint aim to improve that placement to make possible configuring different pagesize per guest NUMA nodes.

NUMA topologu will be enhanced with new attribute hw:numa_pagesize.N with N as an identify of guest NUMA node.

  hw:numa_pagesize.0 = large
  hw:numa_pagesize.1 = any

For a topology with several NUMA nodes it will be possible to set only page size of the desired node. Nodes without a specific page size will be considered to use default and smallest page size.


This example means that a guest will be configured with topology of 2 NUMA nodes and node identified 1 will have memory page size backed to large.


Specification URL (additional information):


Comment 11 Sahid Ferdjaoui 2016-09-28 15:02:07 UTC
Added the nova spec as reference.

Note You need to log in before you can comment on or make changes to this bug.