Realization And Comparison Of Cloud-computing Mapping Algorithms

Nowadays, instead of developing one’s own network, IT companies tend to allocate IT resources from the cloud operators, which provide them with a flexible service hosting and allow them to save buying expensive IT infrastructures. Compound service deployed by IT companies can be composed of a set of basic services, the total installation cost is denoted by the summation of the deploying cost on each requested service plus the possible connectivity cost. However, since there are lots of possibilities to map from requested services layer to underlying substrate layer, the mapping cost varies a lot based on different mapping algorithms. This paper’s objective is to find out a feasible and cost-efficient mapping solution, which can achieve all services’ requirements with adequate user experiences.It should be noted that this paper considers the topology of abstract and substrate layer as two independent graphs, which are composed of nodes and edges. The total mapping cost, should thus be calculated as a cumulated product of all matched pairs of abstract and substrate nodes/edges. Therefore, this paper introduces an algorithm called STRM, which is based on mapping the maximum spanning tree of service graph to the substrate graph. In this way, a mapping solution is determined by taking all abstract nodes and most of the important edges into consideration. Then an enhanced version called IGM is designed after STRM by adding the ignored abstract edges of STRM, and adjusting them to the previous mapping solution. IGM is focusing on optimizing the shortage of STRM, which is thus considered to be a more advanced and more matured algorithm.In the simulation part, this paper has compared STRM/IGM with two other algorithms, one is ViNEYard—an algorithm based on linear-programming with integer relaxed, another is Isomorphism—a heuristic-based algorithm which attempts all the feasible possibilities of mapping for a generated set of abstract nodes. Consequently, among the four algorithms in comaparison, it turns out that in term of mapping cost, IGM performs superior than STRM and ViNEYard, while Isomorphism being the most costly one; result of execution time tests shows that STRM is also the fastest one while ViNEYard being the slowest as it has introduced the linear programming.