New synchronization protocol for distributed system with TCP extension

Process management is a fundamental problem in distributed systems. It is fast becoming a major performance and design issue for concurrent programming on modern architectures and for the design of distributed systems. One of the major duties in process management (synchronization) is mutual exclusion control. Previous studies related to the topic do not use enough messages in the distributed system. Hence, the presented solutions are unable to identify active processes that are currently related to critical sections and dead in the network communication. Specially, if a fault happened for a process that has a coordinator role or in a critical section, the inability to detect the faults would cause the system to crash. Faulty processes may cause some other active processes to become inactive during the queuing or within the using a critical section time. In addition, the system requires adding some messages to avoid starvation and deadlock. On the other hand, previous researches have focused only on time stamp, which is unfair because there are other critical parameters that are not being considered. The thesis shows that it is possible to model the processes of distributed system in the form of a three dimensional matrix, so as to optimize fault-tolerant for mutual exclusion and critical section management. In this regard, the research also presents a new approach of the race models for distributed mutual exclusion. The new components such as participation of time stamp, time action, and other parameters such as special priority are introduced. The matrix is defined as such based on weight, which is able to solve problems of critical sections in order to obtain better state of fault-tolerance. After embedding the presented solution, a new protocol is created and applied. Thus, the new protocol is available to the communicating packets across all computer networks. The new messages and parameters could add to the available option part in the TCP packet header format, which has some free places. Another aspect of distributed systems, which considered in the thesis, is process allocation to resources (such as critical sections). Optimization of process allocation causes to decrease the network traffic and fairer allocation, and therefore,optimization of fault-tolerance. The new protocol is simulated with using the OPNET and the MATLAB software. According to previous researches that evaluated performance to measure faulttolerance in this thesis, the network performances are evaluated and compared. The achieved results should continue to reach a steady state. These results show that the proposed algorithm, on average, has 7.97% higher fault-tolerance. In the worst condition, it optimized 3.99% higher than the most fault-tolerant previous works. The resulting algorithm does not involve variability in the hardware type nor is it based on specific distributed application software or databases. Finally, in sensitive industries, this protocol is capable to offer services on a fault tolerant infrastructure