Optimizing Multi-objective Evolutionary Algorithms to Enable Quality-Aware Software Provisioning

Abstract

Abstract-Elasticity is a key feature for cloud infrastructures to continuously align allocated computational resources to evolving hosted software needs. This is often achieved by relaxing quality criteria, for instance security or privacy because quality criteria are often conflicting with performance. As an example, software replication could improve scalability and uptime while decreasing privacy by creating more potential leakage points. The conciliation of these conflicting objectives has to be achieved by exhibiting trade-offs. Multi-Objective Evolutionary Algorithms (MOEAs) have shown to be suitable candidates to find these trade-offs and have been even applied for cloud architecture optimizations. Still though, their runtime efficiency limits the widespread adoption of such algorithms in cloud engines, and thus the consideration of quality criteria in clouds. Indeed MOEAs produce many dead-born solutions because of the Darwinian inspired natural selection, which results in a resources wastage. To tackle MOEAs efficiency issues, we apply a process similar to modern biology. We choose specific artificial mutations by anticipating the optimization effect on the solutions instead of relying on the randomness of natural selection. This paper introduces the Sputnik algorithm, which leverages the past history of actions to enhance optimization processes such as cloud elasticity engines. We integrate Sputnik in a cloud elasticity engine, dealing with performance and quality criteria, and demonstrate significant performance improvement, meeting the runtime requirements of cloud optimization.

A multi-objective and evolutionary hyper-heuristic applied to the Integration and Test Order Problem

“…and c is a child in C. One comparison is done for each parent and child and the As far as we know, there are no reward measures for credit assignment that use solely the concept of Pareto dominance and only the individuals involved in the mating. In some works, such as [28,36], the reward is computed by comparing the generated solutions with the whole population using quality indicators for each generation or after n (parameter) generations. One of the advantages of our measure is its straightforward implementation using explicitly the wellknown Pareto dominance concept.…”

“…HITO was implemented using jMetal [17], an object-oriented framework for . For all experiments we defined NSGA-II [14] as the MOEA used by HITO, since it is one of the common MOEAs used for solving this problem and within the hyper-heuristics field for comparison [4,28,36,47]. For executing the described algorithms, we used the chromosome representation and the constraints adopted by the approach presented in [4], since this approach was used as a standard approach for the MOEA due to its good results.…”

“…HITO was designed to be flexible and robust, hence it has the following main characteristics: i) a set of steps to select and apply LLHs using MOEAs; ii) a set of parameters to allow the user to personalize its functionality; iii) selection of the best LLH at each mating, in contrast to other works (such as [28,31,36]) that perform the selection at each generation; iv) a rewarding function that is based on the Pareto dominance concept and uses solely the parents and children involved in the mating, rather than using quality indicators such as in [36]; v) score based selection methods, e.g. Choice Function (CF) [36] and Multi-Armed Bandit (MAB) [19], which can also be chosen by the user.…”

“…The experimental evaluation compares the results of their hyper-heuristic approach to the results of state-of-the-art techniques and to the best known results of the literature. The hyper-heuristic performed well on constrained and unconstrained problems in several instances.Kateb et al[28] propose Sputnik, a hyper-heuristic designed to online select mutation operators for MOEAs in order to improve the performance and reduce the cost of cloud infrastructures. Sputnik uses a score-based selection mechanism and apply the best mutation operator for a whole generation before evaluating and selecting the operators again.…”

“…As far as we are aware, this is the only hyper-heuristic besides HITO for this kind of problem, and one of the few hyper-heuristics ever applied to software testing.We treat the ITO problem as multi-objective in order to increase the accuracy in measuring the real effort needed to develop the stubs. In the literature, some works such as[1,20,28,31,36,39,42,43,44,46] proposed multi-objective hyper-heuristics using evolutionary algorithms and LLH selection. However, most of these works are concerned in selecting construction LLHs, meta-heuristics, heuristics to generate new heuristics, or using offline approaches to train the algorithms.…”