The default configuration that is shipped with the server should work for most sites under normal load. For high-load systems, there are a few configuration items should be changed in raddb/radiusd.conf:

      max_requests = 1024

Change the number 1024 to a larger value, as suggested in the comments just above that entry.

For systems with a high load and many CPUs, the following configuration item should also be changed to a higher value:

      max_servers = 32

This number should be increased. The best value to use depends on local load and CPU power. Some sites set this number as high as 500.

That's it.

The same RADIUS server will easily transition from handling one request every few seconds, to handling thousands of requests per second. The only changes required to enable this scaling are outlined above. There is no need to purchase a "enterprise" product license to obtain better performance, more realms, more RADIUS clients, or any other feature.

In short, where commercial servers have "small business", "enterprise", and "carrier class" versions of their product, there is only one version of FreeRADIUS:

Carrier class stability and functionality for everyone.

We prefer to keep it simple.

As proof of it's scalability, the server is currently deployed in multiple Fortune 500 companies, and Tier 1 ISPs and telecommunications companies, as seen in the testimonials and survey pages. Many organizations with more than 10 million customers are dependant on FreeRADIUS for their AAA needs. For some of these large sites, and for most small sites (10 to 1000 users) it is the only RADIUS server that is used.