Comparison between two approaches to overload control in a Real Server: “local” or “hybrid” solutions?

This work analyzes the performances of two algorithms candidate to overload control within a server. An algorithm works according to a “local” strategy, using the concept of measuring the processor occupancy. The other algorithm takes advantage of measuring both the processor occupancy and the call losses. The latter fits in the “hybrid” family, because it exploits a further element (Load Balancer) to reduce the charge of the cpu of the server under protection. The job assumes the throughput, i.e. the number of calls managed within one second, as the unit of comparison.