In this project, you will implement the Chandy and Misra's (CM) algorithm using POSIX Threads (Pthreads). The algorithm is a distributed algorithm to solve a generalized dining philosopher problem, or a general resource sharing with arbitrary agents. For a detail description of the algorithm, refer the last page.
To solve a generalized dining philosopher problem, we have to specify a mutual exclusion graph using a simple text file consisting of two commands: node and edge. For example, the graph at the last page can be described by the following text file. A "node" command has three
(3) parameters: node name, an average thinking time, and an average eating time. An "edge" command has two node names. The first node will have a fork and the second node will have a token initially.
You will create a thread for each node. The nodes communicate to each other to exchange tokens and forks. To properly handle synchronization issue for shared variables if any, semaphore and/or mutex supported by pthread must be utilized. But, try to use such shared variables as little as you can. Note that you can scale down the time for fast simulation, but nodes (or threads) will think and eat at the real time of the possibly scaled-down clock.
While threads are running, they will collect logging data so that they can statistically analyze the actual timely behavior of the threads, and their overall statistical characteristics. Be creative to determine what statistical data to be shown in the project report.
The above graph is just an example. In the actual simulation run, create several connected graphs with different number of nodes, different edge density, and different average times and compare their statistical behavior.