docs/doc/source/node_management/kubernetes/host_memory_provisioning/about-host-memory-provisioning.rst

About Host Memory Provisioning

starlingx

For each node on a host, you can adjust the amount of memory reserved for platform use, and the size and number of memory pages allocated for use by applications.

partner

start-after

introduction-text-begin

end-before

introduction-text-end

Due to limitations in Kubernetes, only a single huge page size can be used per host.

The amount of platform memory reserved differs between and standard configurations.

Note

For Simplex deployments implemented on an Intel Xeon D-based platform, the default platform memory is reduced automatically to limit the number of worker threads, reducing the platform resource usage. You can increase the allotment if required to restore the default maximum number of worker threads.

starlingx

Memory not reserved for platform use is made available as application memory. You can provision individual nodes to use either 2 MiB or 1 GiB hugepages. Due to limitations in Kubernetes, only a single huge page size can be used per host. Using larger pages can reduce page management overhead and improve system performance for systems with large amounts of virtual memory and many running instances.

partner

start-after

memory-text-begin

end-before

memory-text-end

You can use the system host-memory-list and system host-memory-show commands to see how much memory is available for applications. This information is also shown on the Memory tab of the Host Inventory page (see Memory Tab <memory-tab>).

For individual containers (or hosted applications if you are running OpenStack), you can specify which page size is required.

partner

start-after

memory-text-end