Feature Modeling Method for IaaSs. The goal of IaaS feature modeling is to describe IaaSs and user demands for an IaaS effectively and accurately.
In fact, there is a consistent functional and nonfunctional characteristics category existing in IaaSs. Therefore, an IaaS domain model is created based on the metamodel and the IaaS field knowledge so as to unify IaaS FMs' information structure and terminology.
Second, it is necessary to avoid too much refinement of the abstract features that may have a variety of representations, so as to meet the actual needs for describing IaaSs. Third, the important common attributes of the IaaS function characteristics should be defined in a uniform manner, because these attributes will be the main parameters that are evaluated and compared when performing the supply and demand matching and horizontal comparison of IaaSs.
Second, our IaaS description method has good extensibility and compatibility and can continuously update the domain model to improve the conceptual repository according to the development of IaaSs. The concept ontology of IaaS FMs acts as the bridge between the different terms with the same meaning in different IaaS FMs.
Based on the formal IaaS description presented in Section 3, we propose a systematic discovery and selection method for IaaSs. An architecture of the system implementing the method is designed as Figure 6.
Important Nonfunctional Metrics of IaaSs. The establishment of a unified and effective QoS-evaluating metric system is the premise of scientific decision-making.
Different IaaSs having the same VM configuration usually show a significant difference in performance testing results.
Therefore, it is very difficult to obtain available reputations for IaaSs. Although the market share of an IaaS cannot substitute for its reputation, it can basically reflect the user's choice tendency and the service credibility to a certain extent, so we use the market-share data provided by CloudHarmony as a measure of the IaaS reputation.
We obtain the best matching instances from the three IaaSs and their costs by matching the IaaS service and demand models.
First, the IaaS discovery process implements the matching of supply and demand on IaaSs with a configuration instance granularity, in which the search for the matching instances covers the complete configuration space for each IaaS.
This paper systematically studied the description, discovery, and selection methods of IaaSs in cloud computing.
In reality, a combination of multiple IaaSs can sometimes meet the needs of some applications better, such as global ecommerce.