Ensuring robust call throughput and fairness for SCP overload controls

Abstract

This paper analyzes the performance of two ditYerent types of service control point (SCP) overload control algorithms, table-driven controls and an adaptive control. As performance measures, we use aggregate transaction throughput and fair- ness (whether the control algorithm allocates SCP capacity to indkidual traffic sources as desired). We find that even when parameters are weil-matched to source characteristics, tabie- driven controls do not perform as well as the adaptive control. In addition, the adaptive control is more robust to traffic patterns and SCP internal operation than the table-driven controls. I. INTRODumlON s ERVICE control points (SCP) are critical focal points in the Advanced Intelligent Network (AIN) architecture. They serve as logical hubs for AIN because they will contain the data and service logic for applications that will run in many end offices and tandems. SCP's are also physical focal points for traffic because each SCP pair will serve a large number of SSP'S (service switching point; a switch capable of common channel signaling), e.g., all those in a U.S. LATA or a larger area. The central role of SCP'S in AIN makes it likely that they will at some time become the targets of overload and essential to network integrity that they have the means to cope effectively with overloads. SCP overload controls (1 )-(4) use automatic code gapping (ACG). ACG consists of procedures in SCP'S for measuring SCP overload, messages for communicating the SCP conges- tion level back to sources of traffic, and algorithms in the sources for throttling back traffic on the basis of telephone numbers. An ACG request to a source tells it to apply a rate control of the form "send at most one message per g seconds" for a specified duration of time. The key issue in this paper is how to determine the gap interval g. A version of ACG that we call "standard ACG" chooses gaps from a table populated with allowed values from (2). The integer-valued SCP congestion level is the index into the table. Every source uses the same table. We find that standard ACG shows the following. " It can yield low throughput because the control alternately turns traffic on and off and the off period is too long. Many sources may be throttled, even though the SCP is idle. At the next measurement interval, the SCP will again become overloaded. Standard ACG uses gap interval values that are too large.