Read a RAS GRE benchmark in the context of its TCP session count

Source facts

The cited Microsoft page reports a non-SDN test environment using Windows Server version 1709. Its gateway VMs are configured with ten cores and eight gigabytes of memory, with one active and one passive gateway. When the experiment changes from multiple TCP sessions to one TCP session, Microsoft reports that only one CPU core reaches maximum capacity on the gateway VMs. Microsoft Learn.

Applicability

Treat the page as a versioned experiment rather than a present-day throughput promise. Compare its gateway role, VM resources, tunnel type, and session count with the system being assessed. Identify whether the actual application uses one sustained flow or many independent flows.

DSE recommendation

Write a benchmark plan that records the TCP session count alongside the GRE gateway configuration. Ask the application owner which flow pattern represents normal and peak use. Keep the active and passive gateway observations separate, and collect per-core evidence instead of only an overall CPU average. Use the historical result to shape questions for a current pilot, not to select an obsolete deployment.

Verification

Run the approved single-flow and multi-flow cases on the current supported platform. Compare per-core utilization and completed application traffic with the stated test conditions. Record the operating-system build and resource settings with every result. Explain differences from the historical experiment rather than reporting its numbers as measurements of the new system.

Official references

Microsoft Learn: RAS Gateway GRE Tunnel Throughput and Performance. Source reviewed September 8, 2026.