An Approach to Exploiting Skewed Associative Memories in Avionics Systems

• Venue: ICPADS '02 Proceedings of the 9th International Conference on Parallel and Distributed Systems
• Year: 2002

There are two main types of process schedulingalgorithms commonly used in aircraft/spacecraftavionics systems. The first category consists ofdynamic algorithms, which dynamically assignpriorities to processes on the basis of runtimeparameters. The second category consists of staticalgorithms, which statically determine prioritiesbefore runtime. The main disadvantage ofapplying dynamic process scheduling algorithmsto avionics systems is the extra runtime overheadproduced by these algorithms. This overhead ismainly related to the time required to sort activeprocesses in the ready queue upon each processpreemption or the arrival of each new process.The mentioned overhead encourages the use ofstatic algorithms. But static algorithms have theirown disadvantages. In fact, these algorithmsbound the maximum available CPU utilization andhave difficulties with non-periodic processes. Thispaper proposes and evaluates an approach toexploiting skewed associative memories in orderto replace the time-consuming sorting operationby an efficient search operation. Both analyticalmodels and simulation results show that theproposed approach can reduce the timecomplexity of the runtime overhead of dynamicscheduling algorithms (in terms of n the number ofactive processes) from O(nlogn) to O(n). This canconsiderably increase the performance of dynamicscheduling algorithms and make them much morefeasible to be used in aircraft/spacecraft avionicssystems.