Cloud Brokering

Felipe, Díaz Sánchez

• Publisher: Ediciones USTA
• Year: 2016
• eBook ISBN: 9789586319621
• Topic: Technology & Engineering
• Subtopic: Electrical Engineering & Telecommunications
• Index: Technology & Engineering

El servicio de intermediación en la nube es un paradigma de servicio que proporciona interoperabilidad y portabilidad de aplicaciones a través de múltiples proveedores de nube. El atractivo de la intermediación en la nube depende de los nuevos servicios y de las instalaciones informáticas ampliadas que mejoran o complementan los ya ofrecidos por los proveedores de nube aislados. Estos servicios brindan un nuevo valor a las pequeñas y medianas empresas (PYMES) y grandes empresas y hacen que los proveedores de nube sean más competitivos. Hoy en día, a nivel de infraestructura, los intermediarios de servicios en la nube actúan como intermediarios entre los usuarios finales y los proveedores de nube. El intermediario de servicios en la nube proporciona un punto único para el consumo de servicios con el fin de evitar el bloqueo del proveedor, aumentar la resiliencia de las aplicaciones, proporcionar facturación unificada y simplificar los procesos de gobernanza, adquisición y liquidación a través de múltiples proveedores en la nube. En el futuro, los intermediarios de servicios en la nube ofrecerán servicios avanzados de valor agregado y usarán modelos de precios atractivos para capturar a posibles consumidores de nube. La finalidad este libro es proponer servicios avanzados de valor agregado y un modelo de precios para los intermediarios de servicios en la nube. Este libro tiene tres objetivos: • El primero es presentar una figura única de mérito del rendimiento de las máquinas virtuales de la nube basado en el perfil de aplicación. • El segundo es proponer un enfoque exacto para la asignación de máquinas virtuales a través de múltiples proveedores de nube con base en diferentes criterios de optimización. • El tercero es describir el modelo de precios para la intermediación de servicios en la nube, llamado payasyoubook.

Part I

Value-added services in cloud brokering

Chapter 1

Cloud performance and placement in cloud brokering

Introduction

As the number of cloud computing services increases, so does the interest of consumers to be able to compare these services in order to choose those best adapted to their needs. This chapter focuses on the performance issues related to cloud provider evaluation and on the role of cloud brokers in the automatic optimization of resource allocation across multiple cloud providers. This chapter is structured as follows. Section "cloud performance evaluation". presents a survey of the current studies related to cloud performance evaluation. The motivation for and challenges behind the evaluation of cloud provider performance is also detailed in this section. Section "Placement in cloud brokering" describes the current methods used to allocate resources in cloud brokering. The studies are classified into two categories: placement based on non-functional requirements and application-aware placement.

Cloud performance evaluation

Motivations and challenges

The current cloud computing landscape hinders a straightforward comparison of cloud provider service offerings. In the case of computing resources, this is mainly due to the heterogeneity of VM configurations and prices. On one hand, traditional cloud providers such as Amazon, Rackspace and WindowsAzure sell fixed-size VMs. These VM configurations often vary from one cloud provider to another, making it impossible to make a direct comparison. On the other hand, new cloud providers in an effort to attract consumers, look to differentiate their services through technology by allowing consumers to freely configure the size of the computing resources to be purchased.

VM performance evaluation adds another layer of complexity to the comparison of cloud providers. Firstly, consumers have little knowledge and control over the infrastructure hosting their applications. Due to the virtualization of hardware used in cloud computing, cloud providers may use resource sharing practices (e.g. processor sharing, memory overcommit, throttling or under-provisioned network [4]) that degrade the performance of a cloud application. Secondly, cloud provider’s data centers are equipped with hundreds of thousands of servers with different qualities of hardware and software. The evaluation of performance cross all the data centers of multiple cloud providers, implies a trade-off between thoroughness, time and cost of the evaluation [5]. Thirdly, cloud providers may continually upgrade or extend their hardware and software infrastructures, and new commercial services and technologies may gradually enter the market [6]. Therefore, performance evaluations become quickly out of date and the tools for performance measurement must be continuously re-designed. Finally, there are no cloud-specific benchmarks to evaluate all VM features [7]. However, traditional benchmarks can partially satisfy the requirements for cloud performance evaluation.

Cloud performance evaluation would be beneficial for both consumers and cloud providers [5]. Consumers testing their applications across multiple cloud providers can choose the cloud provider that represents the best performance-cost trade-off. Also, performance evaluations can serve as a recommendation for the performance of a particular system [4] or can give technical arguments to consumers to put pressure on cloud providers to use better practices [7]. A provider may identify its market positioning in order to improve its services or to adjust its prices [5].

Studies related to cloud providers performance evaluation

An exhaustive study about the academic approach to commercial cloud services evaluation has been carried out by the Australian National University [6]. A Systematic Literature Review (SLR) was the methodology employed to collect the relevant data to investigate the evaluation of cloud services. As a result, 82 relevant cloud service evaluation studies were identified. The key findings of this study represent a state-of-practice when evaluating cloud services and are as follows:

• 50% of the relevant studies investigated applying cloud computing to scientific issues, while only 16% of the studies focused on the evaluation of business applications in the cloud.
• 21 cloud services over 9 cloud providers were identified. 70% of the relevant studies evaluated cloud services provided by Amazon Web Services (AWS).
• Three main aspects and their properties for cloud services evaluation have been investigated (performance, economics and security), performance being the most studied aspect (78 studies).
• There is no consensus regarding the definition and the usage context of metrics. Some metrics with the same name were used for different purposes, some metrics with different names were essentially the same. The study identified more than 500 metrics including duplications.
• There is a lack of effective metrics vis-à-vis elasticity and security aspects in cloud computing. Therefore, it is hard to quantify these aspects.
• There is not a single or a small set of benchmarks that provides a holistic evaluation of cloud services. The SLR identified around 90 different benchmarks in the selected studies of cloud services evaluation. These benchmarks can be grouped in three main categories: application, synthetic and micro-benchmarks, as explained below.
• 25 basic setup scenarios for constructing complete cloud service evaluation experiments have been identified and classified.
• The cloud service evaluation is getting more and more attention from the research community. The number of relevant studies was 17 times larger in 2011 (34 studies) than in 2007 (2 studies).

Cloud performance evaluation is done by running application benchmarks, synthetic benchmarks or micro-benchmarks in single or multiple cloud providers. Application benchmarks correspond to real-world software that provides an overall view of the performance of a specific application. Synthetic benchmarks simulate application behavior by imposing a workload on the system. Similarly, micro-benchmarks impose a workload with the aim of measuring hardware-specific VM features. Since there are no cloud-specific benchmarks, cloud performance has been measured through widely used benchmarks such as TPC-W (a transactional web e-Commerce benchmark) [8], HPCC (a software suite consisting of 7 basic benchmarks) [4, 9, 10], NPB (set of parallel benchmarks to evaluate the performance of parallel supercomputers) [4, 11] or common measurement tools such as ping or iperf [12,13]. Also, specific benchmarks have been developed to measure cloud performance of CPU, memory, disk and network [14, 15] further the VM provisioning or deprovisioning time [10, 12, 16]. Details about the studies related to cloud providers’ performance evaluation are presented in Table Appendix A.

Recent st...

Table of contents

• Portada
• Título
• Derechos de autor
• About this book
• Cloud brokering
• Part I. Value-added services in cloud brokering
• Part II. A new pricing model in cloud brokering
• Conclusion and future works
• Appendix A. Studies related to cloud providers performance
• Appendix B. Cloud performance evaluation: details and extended results
• Bibliography