Voting–Priority-Based Deadlock Prevention in Multi-server Multi-CS Distributed Systems

Abstract

Concurrency and starvation control had always been a problem in distributed system. In the voting-based technique, a process receiving the greater part of votes will only be permitted to go into the critical section (CS). But, this particular technique has a shortcoming that if no process receives majority of votes, then the whole system will stay out in idle state, while a substantial amount of minority processes will continue in waiting queue. In the priority-based distributed mutual exclusion (MUTEX) algorithms, the process having uppermost priority among all the processes of distributed system is permitted to enter into the CS. Hence, leaving the other large number of lower-priority processes in the waiting queue may result a significant increase in the length of waiting queue. In order to solve the concurrency and starvation control-related problems of distributed systems, we have presented the voting- , priority- and non-maskable interrupt-based optimal and feasible solution in this research work. Here, an efficient and collective solution by including priority with voting-based approaches and certain other constraints like handling non-maskable interrupts and using shortest job sequencing first scheduling techniques will be useful for further improving the performance of the distributed system. The proposed technique of this paper permits the formation of multiple critical sections in single- and multi-server-based distributed systems, and hence, it prevents the processors from entering into inoperative state which leads to increase in overall throughput of system.