Reinforcement learning based autonomic virtual machine management in clouds

Habib, Arafat; Khan, Muhidul Islam

doi:10.1109/iciev.2016.7760166

Cited by 8 publications

(3 citation statements)

References 7 publications

(14 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Nonetheless, they may have drawbacks such as high costs and extensive computing power requirements. The Markov decision process, optimized with reinforcement learning (RL), is another technique applied in this context [36]. Additionally, deep learning can be combined with RL to explore data characteristics and learn scheduling strategies, or it can be applied directly to the VM placement problem [37].…”

Section: Selection Of Destination Server For Vm Migrationmentioning

confidence: 99%

Queuing Model with Customer Class Movement across Server Groups for Analyzing Virtual Machine Migration in Cloud Computing

Kushchazli,

Safargalieva,

Kochetkova

et al. 2024

Mathematics

View full text Add to dashboard Cite

The advancement of cloud computing technologies has positioned virtual machine (VM) migration as a critical area of research, essential for optimizing resource management, bolstering fault tolerance, and ensuring uninterrupted service delivery. This paper offers an exhaustive analysis of VM migration processes within cloud infrastructures, examining various migration types, server load assessment methods, VM selection strategies, ideal migration timing, and target server determination criteria. We introduce a queuing theory-based model to scrutinize VM migration dynamics between servers in a cloud environment. By reinterpreting resource-centric migration mechanisms into a task-processing paradigm, we accommodate the stochastic nature of resource demands, characterized by random task arrivals and variable processing times. The model is specifically tailored to scenarios with two servers and three VMs. Through numerical examples, we elucidate several performance metrics: task blocking probability, average tasks processed by VMs, and average tasks managed by servers. Additionally, we examine the influence of task arrival rates and average task duration on these performance measures.

show abstract

Section: Selection Of Destination Server For Vm Migrationmentioning

confidence: 99%

Queuing Model with Customer Class Movement across Server Groups for Analyzing Virtual Machine Migration in Cloud Computing

Kushchazli,

Safargalieva,

Kochetkova

et al. 2024

Mathematics

View full text Add to dashboard Cite

show abstract

“…RL has been used to solve many routing problems over the past few years. For RL-CH, we have used Q-learning [8,17,18] to accommodate multiple CHs in the model. Q-learning is a model free RL algorithm where an agent interacts with the environment and learns the optimal policy by yielding some scalar rewards.…”

Section: Rl-ch Data Aggregation Mechanismmentioning

confidence: 99%

A Novel Data Aggregation Mechanism using Reinforcement Learning for Cluster Heads in Wireless Multimedia Sensor Networks

Uddin¹

2022

AETiC

View full text Add to dashboard Cite

Wireless multimedia sensor networks (WMSNs) are getting used in numerous applications nowadays. Many robust energy-efficient routing protocols have been proposed to handle multimedia traffic-intensive data like images and videos in WMSNs. It is a common trend in the literature to facilitate a WMSN with numerous sinks allowing cluster heads (CHs) to distribute the collected data to the adjacent sink node for delivery overhead mitigation. Using multiple sink nodes can be expensive and may incur high complexity in routing. There are many single-sink cluster-based routing protocols for WMSNs that lack in introducing optimal path selection among CHs. As a result, they suffer from transmission and queueing delay due to high data volume. To address these two conflicting issues, we propose a data aggregation mechanism based on reinforcement learning (RL) for CHs (RL-CH) in WMSN. The proposed method can be integrated to any of the cluster-based routing protocol for intelligent data transmission to sink node via cooperative CHs. Proposed RL-CH protocol performs better in terms of energy-efficiency, end-to-end delay, packet delivery ratio, and network lifetime. It gains 17.6% decrease in average end-to-end delay and 7.7% increase in PDR along with a network lifetime increased to 3.2% compared to the evolutionary game-based routing protocol which has been used as baseline.

show abstract

“…Diferentemente, neste artigo os estados são baseados inteiramente na banda contratada. Habib e Khan adotam uma proposta de ações similar a de Mera-Gomez et al, com foco na alocação de recursos em nuvens, porém comparando com outro algoritmo de aprendizado por reforço, o SARSA (State-Action-Reward-State-Action) [Habib e Khan, 2016]. O trabalho conclui que o algoritmo Q-learning apresenta melhor desempenho em relação ao tempo de convergência.…”

Section: Os Trabalhos Relacionadosunclassified

Agentes Inteligentes baseados em Aprendizado por Reforço para Alocação Dinâmica de Tráfego em Nuvens

Filho¹,

Mattos²,

Medeiros³

2020

Anais XXXVIII Simpósio Brasileiro De Redes De Computadores E Sistemas Distribuídos (SBRC 2020)

View full text Add to dashboard Cite

A alocação eficiente do tráfego em nuvens é desafiadora devido ao compartilhamento de recursos entre os clientes. Isso pode implicar recursos ociosos caso os clientes estejam limitados a utilizar somente a banda contratada. O uso da nuvem pode ser otimizado provendo recursos aos clientes dinamicamente de acordo com a demanda. Agentes de aprendizado por reforço promovem respostas adaptáveis a ambientes variantes no tempo. Este artigo propõe um mecanismo baseado em Q-learning com múltiplos agentes para gerenciar o acesso aos recursos por cada cliente da nuvem. A proposta é analisada em um ambiente emulado, no qual um controlador é responsável pela alocação de tráfego aos clientes. Os resultados mostram que o mecanismo reduz a ociosidade da nuvem, permitindo que clientes com baixa priorização utilizem a banda disponível, ao mesmo tempo que garante a banda contratada a clientes prioritários. O mecanismo exige baixo comprometimento de processamento total, mesmo variando o número de estados e espaço de ações, ao passo que o custo em memória por agente aumenta, alcançando um máximo de 300 kB para 200 estados e ações.

show abstract

Reinforcement learning based autonomic virtual machine management in clouds

Cited by 8 publications

References 7 publications

Queuing Model with Customer Class Movement across Server Groups for Analyzing Virtual Machine Migration in Cloud Computing

Queuing Model with Customer Class Movement across Server Groups for Analyzing Virtual Machine Migration in Cloud Computing

A Novel Data Aggregation Mechanism using Reinforcement Learning for Cluster Heads in Wireless Multimedia Sensor Networks

Agentes Inteligentes baseados em Aprendizado por Reforço para Alocação Dinâmica de Tráfego em Nuvens

Contact Info

Product

Resources

About