I want to enforce strict CPU quota limits on RD Session Host user sessions.

Question

Is relying on WSRM CPU allocation policies sufficient to control multi-threaded apps or is there a way to configure kernel driver parameters directly - meaning that when a user application spikes CPU usage on a high-density session host it will be hard-capped instead of taking up 100% of all logical processors?

Answer 1

Hi Lin,

No. WSRM is not the right solution for this requirement, and in fact WSRM was deprecated and removed from modern Windows Server releases, so it should not be relied upon for enforcing strict CPU limits on RDS session hosts.

Windows does not expose a supported kernel driver parameter that can hard-cap a user session's CPU consumption across all logical processors. The built-in mechanism on RD Session Hosts is Fair Share CPU Scheduling, which dynamically adjusts thread scheduling priorities to prevent session starvation, but it is a fairness mechanism rather than a hard quota. A heavily multi-threaded process can still consume substantial CPU when system load allows it.

If you need a true CPU ceiling, the supported approaches are process-level controls using Job Objects/CPU rate control APIs, third-party resource management products, or isolation through Hyper-V/VDI where CPU limits can be enforced at the VM boundary. Windows Server does not provide a native per-user or per-session CPU hard cap comparable to Linux cgroups.

So the direct answer is that neither WSRM policies nor kernel tuning can guarantee "this user application can never exceed X% of total CPU." For a high-density session host, the best Microsoft-supported option is Fair Share scheduling for contention management, and VM-level isolation if strict resource governance is a hard requirement.

Harry.