Fast data structures for shortest path routing: a comparative evaluation

The execution time of Dijkstra's (1959) algorithm for shortest path routing strongly depends upon the underlying priority queue data structure used to store intermediate path costs for the graph during the algorithm's iterations. In particular, the efficiency is determined by how quickly the data structure can execute the operations of EXTRACT-MIN and DECREASE-KEY. Using Dijkstra's algorithm on graphs of various sizes and topologies, this paper compares the actual performance of the standard implicit binary heap with that of several proposed data structures: the binomial heap, relaxed heap, run relaxed heap, Fibonacci heap and splay tree. In particular, we compare the data structures' performance of the key operations, EXTRACT-MIN and DECREASE-KEY.