Strategic Provision of Cloud Computing Services
Models, Methods, and Techniques of Service Provisioning
The used methods, models, and techniques for successful and efficient provisioning of services have a paramount importance. For these discussions, we use several acronyms and please refer to the Abbreviations Section. Hence, we discuss the criteria of service provisioning as follows.
Service Provisioning Models
The service measuring index (SMI) is a service measurement model based on business model of the International Standard Organization (ISO). The SMICloud model is proposed by Garg et al. which lets users compare different cloud offerings, according to their priorities and along several dimensions, and select whatever is appropriate to their needs. Several challenges are tackled in realizing the model for evaluating QoS and ranking cloud providers. The SMICloud systematically measures all the QoS features proposed by cloud service measurement index consortium (CSMIC) and ranks the cloud services based on these business services. Again, Li et al. introduce CloudCmp, a systematic cloud service performance and cost comparator system. It assesses the elastic computing, storage, and networking services promised by cloud metrics, which have an impact on performance of customer applications. CloudCmp safeguards the fairness, participation, and compliance of these assessments, compromising limiting measurement cost. By using CloudCmp, most of the cloud customers today can find offered services from cloud service provider very widely in terms of efficiency and price and calculate the need for a thoughtful provider selection. Han et al. present the cloud service selection framework which employed a recommender system (RS) that selects the best services from different cloud providers according to the customer needs. The Cloud Provider (CP) registers their services in the web portal through CSRS system and users put their requirements and get recommendation by the web portal. If a CP wants to register, the request is passed to the request manager through the web portal after the evaluation process and then sends to resource register after calculating the S-rank and QoS values of CPs and storing in resource repository.
Brokerage Aided Provisioning
Sundareswaran et al. describe the problems of finding the best price and services are addressed from the huge pool of services from the cloud service providers. A novel brokerage based architecture is proposed to use a unique indexing technique for efficient service selection by employing algorithms which aggregate and select the optimal option. In this model, the cloud broker collects the service provider's properties and similar properties are analyzed and indexed by the cloud broker and using it when receiving request from customer for the best matched service. Le Duy et al. introduce a new benchmark to evaluate and compare cloud brokers. Cloud broker challenge (CBC) explains the cloud providers, cloud consumers, and goals with five variety levels of complexities. CBC benchmark is useful for evaluation and comparison of unbiased brokers and feasible for real-life cloud brokers. The design and development of software agents for cloud service discovery, service negotiation, and service composition play an important role. Sim introduces an agent-based paradigm for constructing software tools and test beds for cloud resource management. He develops Cloudle – an agent-based search engine for cloud service discovery to show effective agent-based negotiation mechanisms and agent-based cooperative problem solving techniques effectively adopted in automatic cloud service composition. Moreover, agent-based problem solving techniques such as acquaintance networks and the contract net protocol are employed by Gutierrez-Garcia and Sim for creating a self-organizing service composition framework. In this model, cloud service composition framework provides generic agent behaviours to handle ad hoc web service workflow specifications. Finally, self-organizing service composition is supported by contracts among cloud participants which mapped to service level agreements in cloud computing environments. Amato et al. in the mOSAIC project design knowledge based representing resources and domain concepts of semantic web ontologies and rule based support tool, the semantic engine. It aids the user to abstract the requirements in a vendor independent way to compare the different offers of providers with the user proposal and retrieves the best offer.
Policy Ensured SLA
In cloud computing, contracts between users and traders are recognized as service level agreements (SLAs), mentioning the terms and conditions of service usage. Service level agreements are established between service consumers and providers and define a number of obligations and rights for both sides. However, the increasing number of service offerings is so rapid and there is a lack of a standard for specification; manual service selection is an expensive task, averting the successful implementation of on demand ubiquitous computing. Therefore, automatic methods for matching SLAs are essential. Redl et al. propose a method to select semantically equal SLA elements from differing SLAs by employing several machine learning algorithms. In addition, this method enables autoselection of optimal service offerings for cloud services. A framework is presented to automatic SLA management by a simulation-based study to establish several significant advantages of this approach for cloud customers. Research on SLA management focuses on SLAs with rights for consumers and obligations for providers, keeping the two parties balanced interest. Spillner and Schill present a solution of monitoring data at runtime and feeding it back into the service registry to adjust descriptions and make contract template derivation as a more realistic process. Emeakaroha et al. introduce detection SLA violation infrastructure (DeSVi), the novel architecture for monitoring and detecting SLA violations in cloud computing infrastructures. The main components of the architecture are the automatic VM deployer, responsible for the allocation of resources and for mapping of tasks, application deployer, responsible for the execution of user applications, and LoM2HiS framework, which monitors the execution of the applications and translates low-level metrics into high-level SLAs. However, this proposed system is capable of monitoring only a single cloud data center. On the one hand, SLA violation should be prevented to avoid costly penalties and on the other hand providers have to efficiently utilize resources to minimize cost for the service provisioning. Few approaches are limited to simple workflows and single task applications. Bouchenak introduces a systematic and synchronized integration by the definition of SLA aware cloud and explains the automated cloud control for building SLA aware dynamic elastic clouds. Finally, Chi et al. describe a framework by employing novel data structure. Pearson and Sander introduce a policy based mechanism of service provider assessing the risk-based semiautomated system which drastically reduces the transaction to lower the cost of selecting desired CSP. This ensures compliance and trustworthiness of service providers.
Heuristic and Holistic Perspective
Song et al. introduce a framework for task selection and allocation to enhance resource utilization for PCP by exploiting an adaptive filter. To optimize the goal of a heuristic algorithm for optimizing the tradeoff between QoS of the tasks and utilization of resources a tradeoff metric is introduced. A VM-based overall resource structure for computing resource utilization is presented. Simulation study shows that algorithm performs better than other existing algorithms. Moreover, Beloglazov et al. define an architectural framework and principles for energy aware heuristics provision data center resources to client applications in cloud computing. Energy efficient resource allocation policies and scheduling algorithms are introduced by considering QoS expectations and power usage characteristics. The approach is validated by a performance evaluation study using the CloudSim toolkit. Again, Casalicchio and Silvestri propose autonomic service provisioning and resource management of cloud-based systems especially for IaaS providers. Thus, this system has four alternatives implementation which has different degree of control on the various components of the autonomic cycle.
Ferrer et al. incorporate the service provider (SP) and infrastructure provider (IP) with toolkit which optimize the whole service life cycle. Each core component of the toolkit provides common services, which are needed for service deployment and execution. Again, the criteria followed by IP are past performance and legal and security aspects. The calculation is conducted by considering a value for each criterion from 0 to 1. Finally, the assessment is achieved by implementing by Dempster Shafey analytical hierarchy process (DS-AHP) as the service provisioning system is important in the user's perspective of cloud service performance. Hence, for network virtualization it is a vital attribute of next-generation Internet-based service provisioning approach to integrating networking and cloud computing. Duan introduces a holistic approach of the application of the SOA in network virtualization for composing network and cloud services and studied modelling and performance analysis on network virtualization for composite network-cloud service provisioning.
Cloud Service Provisioning Based on MCDM
Multicriteria decision making (MCDM) is a well-established area in the field of operations research and has proven its effectiveness in addressing different complex real-world decision-making problems. Rehman et al. present a comparative case study involving infrastructure as a service cloud and use MCDM techniques to select the best service on the basis of actual performance measurements by a third party monitoring service against five different criteria. We present a comparative study of service provisioning techniques in Table 2.
| Service provisioning techniques | Features | Solving approaches | Attributes | References |
| Algorithmic | A cloud-based computing services scheduling with collaborative QoS requirements | Binary integer programming method | Optimization and fairness | Wei et al. |
| Considering qualitative effects of cost and strategy model | Nash equilibrium under different formulations | Capacity and probability | Rao et al. | |
| Run time management in service provisioning in IaaS | Distributed algorithm | Equilibrium efficiency | Ardagna et al. | |
| Cost based multi-QoS job scheduling model | Soft deadline and penalty cost | Better scheduling | Dutta and Joshi | |
| MCDM | Task oriented resources allocation | Reciprocal and induced bias matrix | Bandwidth, task costs, and time | Ergu et al. |
| A distributed resource management | Considering SLA and QoS | Realizing user needs | Khaddaj | |
| A dynamic autonomous resource management and scalability | PROMETHEE architecture |
Suitable for large data centers | Yazir et al. | |
| SLA and Policy based brokering | For autoselection of SLA from different offerings | Machine learning algorithms | SLA mapping | Redl et al. |
| Knowledge based sources and services in mOSAIC project | Semantic web Ontologies rule based support tool | Requirements and services | Amato et al. | |
| Compare and evaluate cloud broker by CBC benchmark | Cloud service selection (CSS) algorithm | Query encoding, k-nearest neighbor | Le Duy et al. | |
| Best offering selection by brokerage based architecture | Indexing technique B+-tree | Encoded and analyzed, index key | Sundareswaran et al. | |
| Monitoring and detecting SLA violation | DeSVi architecture | Low level metrics to high level SLAs | Emeakaroha et al. | |
| SLA aware cloud considering by data structure SLA tree | SLaaS, SLA tree | SLA aware provisioning | Bouchenak, Chi et al. | |
| A policy based mechanism of service provider selection | Assessing risk by semiautomated system | Low cost with trust and compliance | Pearson and Sander | |
| Heuristic and holistic | Energy aware heuristics provision of data center resources | Energy efficient allocation policies and algorithms | Power usage, QoS | Beloglazov et al. |
| Four architectural schemas for autonomic resource allocation | Four alternative degree of control | Autonomic management | Casalicchio and Silvestri | |
| Optimized service life cycle for dependable adaptive dynamic service provisioning | Dempster Shafey analytical hierarchy process (DS-AHP) | Past performance, maintenance, security, and legal | Ferrer et al. | |
| Accessing federated architecture dynamically | Meta brokering concept | Heterogeneous IaaS aggregation | Kertesz et al. |
Table 2 Comparison of service provisioning techniques.
Analytical hierarchy process (AHP) is a tool for decision makers to be able to do more informed decisions regarding their investment in such technologies. The AHP is a multiobjective, multicriteria decision-making approach, which employs a pair-wise comparison procedure to arrive at a scale of preferences among a set of alternatives. AHP enhances the decision making towards transform subjective judgments into objective measures. In AHP, an input is asked to give ratios for each pairwise comparison between issues for each criterion in a hierarchy and also between the criteria. The pairwise comparison results are displayed in a hierarchy with a weight for each criterion, providing both qualitative as well as quantitative characteristics here. In addition, the technique is for order preference by similarity (OPS) to an ideal solution (TOPSIS) to aid service consumers and providers for analyzing available services with fuzzy opinions. Fuzzy TOPSIS methods are now popular in dealing with imprecise information.
Algorithmic Techniques
In the area of multiobject optimization, genetic algorithm is a famous search method. Genetic algorithm based search is employed to find out the best service variant among the current context by Vanrompay et al. Then, the choice services are deployed in an optimal way which serves the demands of the service running on mobile systems. Their model has several requirements such as deployment of an intelligent artificial learning mechanism, scalable variants, resource constraints of mobile devices, bandwidth limitation, and run time QoS properties that should be considered. Zhao et al. propose an optimized task scheduling algorithm based on genetic algorithm to schedule independent and divisible tasks to adapt to different computation and memory requirements. The algorithm is in the heterogeneous system, and dynamic scheduling is also considered and accordingly GA is designed to solve combinational optimization problem. Again, Dutta and Joshi propose a genetic algorithm based approach to cost based multi-QoS job scheduling model in cloud computing. It guarantees the best solution in finite time. A genetic algorithm has been developed to provide a better scheduling in a cloud environment. Analysis and a number of results show that it ensures a good profit for the different cloud providers. The real execution time of job in different system as well as soft deadline and penalty cost in the algorithm is also considered.
Wei et al. present a game-theoretic method for scheduling cloud-based computing services with collaborative QoS requirements. Game theory is employed to solve the problem of resource allocation. A binary integer programming method is proposed to solve the independent optimization, and an evolutionary mechanism is designed to minimize their efficiency losses. The algorithms consider both optimization and fairness into account which finally reveal that Nash equilibrium always exists for game which has feasible solutions for resource allocation. Again, Rao et al. propose a game-theoretic approach for the provisioning and operation of the infrastructure by considering qualitative effects of cost and strategy model. The Nash equilibrium under different formulations computes in polynomial time and derives provisioning choices to ensure the capacity C with probability PS. In addition, Ardagna et al. introduce a game theory based model for the run time management in service provisioning problem especially for IaaS provider capacity among multiple competing SaaS.