An Effective Voting and Priority based Technique for Deadlock Prevention in Distributed & Cloud Systems

Abstract

Deadlock, starvation and concurrency have always been a problem in single, multiple, distributed and cloud computing performance. Voting based approach is a process of getting the majority of votes and will only be allowed to enter into the critical section (CS). But disadvantage of this method is that if no process remains in the pre-requisite range of majority of votes then the system will remain in idle state and a considerable number of minority processes will remain in waitlist. In prioritized distributed mutual exclusion based algorithms, the process having highest amount of priority will be allowed to enter into the critical section (CS while the processes having lower priority in waitlist will result in an increase in the length of waitlist. Hence, for solving the concurrency and starvation related problems of distributed and cloud computing systems, we have tried to present an NMI based voting and priority algorithm with optimal & feasible solution through our study. Here, we have tried to propose a clubbed solution for voting and priority based approach by adding some other factors such as space for ‘non-mask-able interrupts’ and ‘Shortest Job Scheduling First (SJSF)’ methods has been used for upgrading the throughput of the system. The method will allow the creation of multiple critical sections in cloud, single and multi-servers based distributed systems preventing the serving processors from entering into starvation and in achieving optimum output from the system.