To search, Click below search items.

 

All Published Papers Search Service

Title

Overload Avoidance Algorithm for Real-Time Distributed System

Author

A. F. M. Suaib Akhter, Mahmudur Rahman Khan, Md. Shariful Islam

Citation

Vol. 12  No. 9  pp. 121-124

Abstract

For real-time distributed systems dynamic scheduling has always been a challenging problem. Most of the systems handle overloaded condition after failure of some processes. Thus after recovery the system has to be involved in repairing the lost jobs. In the proposed algorithm the system will never enter in the overloaded state which will ensure that, none of the process will be failed which have already entered in the system. Proposed algorithm is a combination of a dynamic and a static scheduling algorithm. It uses EDF (Earliest Deadline First) which is the most widely used algorithm for dynamic scheduling and RM algorithm is the static algorithm for scheduling in such a condition when the probability of entering in overloaded condition is high. In regular load system will use EDF, but when the load reached in a certain level system will use RM algorithm until the system back to the safe state. A variable is assigned with each process to check whether the process will enter in the system and at the same time it decides when the system has to switch from one algorithm to another. The value of the variable depends on the current system load and remaining capacity of the system. By this way the algorithm will ensure resource utilization, efficiency and high performance of the system in any condition.

Keywords

Distributed system, EDF, Real-Time System, Scheduling algorithm.

URL

http://paper.ijcsns.org/07_book/201209/20120919.pdf