A Branch-and-Bound Algorithm for Autonomic Adaptation of Multi-cloud Applications

Author(s): André Almeida, Francisco Dantas, Everton Cavalcante, Thais Batista
Venue: 2014 14th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing
Date: 2014


With cloud computing platforms becoming bigger and more available, a new type of application is emerging which consists of multiple services of third-party platforms. Dynamic adaptation is an important concern in this scenario, as the cloud services can suffer from instability. The researchers suggest that using a branch and bound algorithm would fix this problem.

The researchers believed that branch and bound algorithms can determine the bounds as tight as possible and thus the optimal solution of the problem. The branching procedure consists in splitting the original solution set into subsets that are easier to solve. In turn, the bounding procedure will determine which subsets will be expanded and which ones will be bounded.

The researchers compared the results of their branching and bounding code to the current existing code used for dynamic adaptation. They found that the branch and bound algorithm is 22.47% faster than the existing code and has a greater capability of scaling for large configurations and options of services within the adaptation process.