Placement Groups

Quake AI concept: Server Groups

• OpenStack Nova server groups expose scheduling policies like affinity and anti-affinity at the API level, but Linode Placement Groups only support Anti-affinity today. The Linode doc says Affinity exists conceptually, yet "Currently, only the Anti-affinity Placement Group Type is supported." (Akamai TechDocs).
• Linode placement groups are region-scoped and capped at five Linodes per group, with the limit surfaced by the `GET /v4/regions/{regionid}` response as `maximum_linodes_per_pg`. OpenStack server groups are not documented this way in Linode's docs, and Linode's limit is a hard service constraint rather than a generic Nova scheduling abstraction (Akamai TechDocs).
• A Linode can only exist in one placement group, and placement groups support only dedicated and shared Linode plans. The doc also says specialty hardware such as GPUs are not supported, which is a narrower applicability model than Nova's general server-group concept (Akamai TechDocs).
• Linode placement groups are managed as an account-side resource that can be created, renamed, or deleted, and you add existing or new Linodes to them through Cloud Manager, API, or CLI. The docs also say that removing a Linode from the group leaves the Linode running as-is, which makes the grouping advisory for an existing instance rather than a persistent per-server scheduler rule (Akamai TechDocs).