Skip to content

Instantly share code, notes, and snippets.

Show Gist options
  • Save oxechicao/c2e4e4fd0fa2d5d643f7c80811d66be0 to your computer and use it in GitHub Desktop.
Save oxechicao/c2e4e4fd0fa2d5d643f7c80811d66be0 to your computer and use it in GitHub Desktop.
multi-cloud allocation resource
  1. Ayoub Alsarhan, Awni Itradat, Ahmed Y. Al-Dubai, Senior Member, IEEE Albert Y. Zomaya, IEEE, Fellow and Geyong Min

O que é feito

Proposta um novo framework SLA onde o parâmetro das demandas de QoS é o preço.

Como é feito

Esse framework usa aprendizagem por reforço (reinforcement learning, RL) para a política de alocação de VM que se adapta a mudanças no sistema. Parâmetros avaliados: custo do serviço, capacidade do sistema e demandas por serviços.

Objetivos chaves usados no para o modelo de RL.

  • Client satisfaction: Oferece uma adequada quantidade de VMs para os jobs dos usuários
  • System Grade of Service (GoS): Capacidade de bloquear uma requisição caso ela ultrapasse o limite de outro requisito. Mantendo sempre o melhor QoS.
  • Cloud Provider: O parâmetro para o ganho do CP é o preço.

No exemplo usuários paga por cada requisição

Para solucionar as políticas de controles de admição de requisições (Reqeust Admission Control, RAC) utiliza-se o processo de decisão de Markov. É utilizado o algoritmo Q-learning para implementação do RL que solucionará o problema de decisão do algoritmo de Markov.

Infraestrutura

Modelagem

Contribuições citadas

  1. Demonstra que utilizar o preço como parâmetro nos dar um controle de QoS quase que independente e contínuo de diferentes classes de clientes.
  2. Aumentando o custo de uma classe durante uma alta demanda reduz a taxa de alocação de VMs, o que auxilia a conhecer as limitações QoS para outras classes.
  3. O uso de Reinforcement Learning, RL, envolve a sintese de algoritmos de controle adaptativo para servir as requisições dos clientes, que é usado para medir as condições do cloud market.
  4. O uso da teoria de decisão de Markov foi implementada para computar os estados independentes das políticas de locação executadas pelo algoritmo RL. Considerando o preço da locação como parâmetro de decisão.

Quotes

Video streaming services require huge storage capacity. Hence, more than one data center should be used to support this service, which is called multi-cloud. Data centers should be monitored and controlled to support QoS.

...measurement techniques are hard to use in computer service performance prediction especially for cloud environments.

Trabalhos Relacionados

  • A. Alasaad, K. Shafiee, H. Behairy, H. M., and V. C. M Leung, Innovative Schemes for Resource Allocation in the Cloud for Media Streaming Applications, IEEE Transactions on Parallel and Distributed Systems, vol. 26, no. 4, pp. 1021-1033, 2015.

    • The authors in [27] propose a new scheme for resource reservation that maximally exploits discounted rates offered in the tariffs, while ensuring that sufficient resources are reserved in the cloud market. In order to make accurate resource allocation decisions, the scheme predicts
      the demand for streaming capacity.
  • G. Jia, G. Han, D. Zhang, L. Liu., and L. Shu, "An adaptive framework for improving quality of service in industrial systems‘,IEEE Access," vol. 3, no. , pp. 2129-2139, 2015.

    • The framework provides differentiated I/O service to various applications and ensures predictable performance for critical applications in multi-tenant cloud environment. An adaptive framework is proposed in [33] for Service Maximization Optimization (SMO). The framework is de-signed to improve the QoS of the soft real-time multimedia applications in multimedia cloud computing.
  • L. Wu, SK.Garg, S. Versteeg and R. Buyya, SLA-based resource provisioning for hosted software as a service applications in cloud computing environments, IEEE Transactions on services computing, vol. 99, no.1, pp. 465-485, 2013.

    • The work in [4] investigates various algorithms for resource provisioning in cloud computing systems. The main concern of the proposed algorithm is minimizing the penalty cost and improving customer satisfaction levels by minimizing QoS constraint violations.
  • D. Kusic, JO. Kephart, JE. Hanson, N. Kandasamy, and G. Jiang, “Power and performance management of virtualized computing environments via lookahead control,”Cluster Computing, vol.12,no 1, pp.1–15, 2009.

    • In [6], a new framework is proposed for dynamic resource provisioning in a virtualized computing environment. They consider switching costs and explicitly encode the notion of risk in the optimization problem.

The Xen ProjectTM is the leading open source virtualization platform that is powering some of the largest clouds in production today. Amazon Web Services, Aliyun, Rackspace Public Cloud, Verizon Cloud and many hosting services use Xen Project software. Plus, it is integrated into multiple cloud orchestration projects like OpenStack.

Used on

PRESS: PRedictive Elastic ReSource Scaling for cloud systems.

We have implemented the PRESS system on Xen and tested it using RUBiS and an application load trace from Google.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment