Info-based approach in distributed mutual exclusion algorithms

In this paper, we propose a token-based fully distributed algorithm with token-asking method for Distributed Mutual Exclusion (DME) in a computer network composed of NN nodes that communicate by message exchanges. The main goal is to introduce a new class of token-based DME algorithms called info-based algorithms. In some previous algorithms, the request to enter a critical section is sent to all nodes because the token-holding node is unknown, but in this info-based algorithm some nodes know the token-holding node and lead critical section entering requests to it, directly. This algorithm uses a logical structure in the form of a wraparound two-dimensional array which is imposed on the interconnecting network. Usually, a request message for entering the critical section is sent vertically down in the array, and eventually sent to the token-holding node with the assistant of an informed-node (common node between the row consisting of the token-holding node and the column consisting of the requester node). The nodes invoking the critical section can obtain the token with fewer message exchanges in comparison with many other algorithms. Typically, the number of message exchanges is 4N+1 under light demand which reduces to approximately 2 message exchanges under heavy demand. A correctness proof is provided.

► The main goal is to introduce a new class of token-based DME algorithms called info-based.
► The algorithm uses a logical topology to decrease the number of message exchanges.
► The algorithm is distributed and all processes receive approximately equal workload.
► Two interesting principles with considerable influence on the performance are used.
► 4N+1 messages under light demand and 2 messages under heavy demand are obtained.