Predicting Service Composition Costs with Complex Cost Behavior

Abstract

Nowadays, many companies expose their competencies as services on the Internet to facilitate the cooperation with their customers. This situation has created a new marketplace where services have been provided with similar functionality but different qualities such as cost, performance, and reliability. In this scenario, service composition providers have faced the challenge of choosing services that fulfill an expected quality without compromising a planned budget. This challenge is even more stringent when services have different cost behaviors. As a consequence, services can be less expensive than others in a scenario and be more expensive in others. Several approaches have been proposed to address cost analysis of service compositions. However, these approaches have not considered all classes of possible cost behaviors, e.g., fixed, variable, mixed and step cost. This paper addresses this limitation by proposing a solution to analyze costs of service compositions taking into account service reliability and all classes of cost behaviors. In order to evaluate the proposed solution, we carried out some experiments that show its effectiveness.

Cost-effective deployment of certified cloud composite services

“…The enrichment of traditional composition solutions with certification techniques evaluating non-functional properties of composite services introduces the need of rethinking the algorithms driving selection of component services. If, on one side, service selection has been already renewed to accomplish selection of services that prove a set of non-functional properties [4], on the other side, solutions to cost-based service selection need to depart from the assumption that costs are only due to service deployment and resource consumption [12].…”

“…The management of a certified service introduces a cost on the cloud provider that can be evaluated using the Deployment Compositional Matrix and depends on three main factors: direct (deployment) costs, indirect (certification) costs, and mismatch costs. Each of these cost factors can be characterized using one of the four different cost behaviors identified by Horngren [22], and later used for cloud services by de Medeiros et al [12]: i) fixed costs, a resource cost function that is completely independent from volume and time, indeed constant;…”

“…Qi et al [7] propose a QoS-aware composi- [32] propose a solution for moving application to the cloud, aiming to select the cloud service that matches at best application requirements and plan characteristics. Another line of research relevant for the work in this paper evaluated the costs of service composition algorithms [12,13,14,33,34]. Greenberg et al [23] analyzed the most relevant direct and indirect costs in cloud service data centers, identi-fying the following: i) server costs (45% of the total costs) depend on server utilization and optimization, ii) infrastructure costs (25% of the total costs) comprise all facilities for power delivery, air conditioning, ups, and the like, iii)…”

“…These costs, in fact, rapidly increase because the costs of continuous certification and verification become substantial. Current research on cloud computing has privileged solutions minimizing costs on the final users [12,13,14], neglecting the costs on the cloud providers that often represent a major source of fee increase.…”

“…It is inspired by our previous work in [4] and extends it according to the cloud challenges discussed in [11] and [15]. Differently from existing work [12,13], our service composition approach is driven by certificates awarded to single services and by a fuzzy-based cost evaluation methodology, and assumes certified properties as must-have requirements for service selection [16] and composition [17]. This methodology aims to decrease the costs of cloud providers, also analyzing those costs introduced by the need of keeping the composition continuously monitored and certified.…”