Integrated overload control for overload control for distributed real time systems

DETAILED DESCRIPTION In this document, we describe an overload control consisting of two inter-related mechanisms: short-term and long term.
In many situations, short-term control actions have smaller impact on the system performance while long-term actions have more severe performance consequences.
Thus, it is desirable to distinguish between short-term and long-term.
In particular, it is important that a unified control method that does not distinguish between the two should not be applied to the system under overload.
The term overload state will be used to include the zero overload state, i.
e.
, the normal state wherein there is no overload in a module.
In the autonomous control method described first, individual switch components (nodes or modules) measure their performance and take action based on these measurements.
All nodes are autonomous and independent and their action may not have relevance to the states and/or actions of other nodes.
In contrast, in the "integrated" overload approach used, in one embodiment, when one or more modules enter a long-term overload level, the states of all nodes are simultaneously considered in determining the course of action during the control.
The short-term control as the name implies is in reaction to measurements of system performance over short periods of time.
This is intended to capture the transient and short-term overload conditions that may not be long lasting and thus, may not require actions that result in heavy penalties (e.
g.
, high call blocking rates.
) Typical short-term control may consist of deferring processing of non-critical tasks; these tasks will be processed after the transient overload has disappeared.
The severity of the action in this case is not critical and is transient.
For example, a non-critical task may be to respond to a request to re-set registers and memory blocks that are allocated to switch maintenance and administration and have no immediate implications to the main tasks such as processing calls in progress in the switch