Modeling Node and Link Failure/Recovery
An important aspect of many performance studies is the effect of network components that fail during operation. Therefore, simulations provide the capability to enable and disable communication nodes and links. This section describes the mechanisms available to model failures and recoveries and presents some example applications of these capabilities.
Each node and link object has a “condition” attribute which specifies whether it is in an operative state or not. When a node or link changes from an operative state to an inoperative one, the Simulation Kernel may generate failure interrupts. Similarly, when a node or link changes from an inoperative state to an operative one, the Simulation Kernel may generate recovery interrupts. These interrupts will be delivered to processor and queue modules that request them. These interrupts are delivered forced, which means that the current event which is changing the node’s “condition” attribute, is suspended and each of the interrupts is delivered during that event.
When a node fails, no interrupts destined for modules within that node are delivered. However, just before the node fails, failure interrupts (if there are any) are the last interrupts delivered to modules within that node until the node recovers. Just after the node recovers, the first interrupts delivered to modules within that node are the recovery interrupts, if any are to be delivered. Typically this does not happen, but the possible exception to failure interrupts being the last interrupts, and recovery interrupts being the first, to be delivered to modules within the node, are other forced events that are generated from other interrupts within the same failure or recovery event.