VM Scaling and Load Balancing via Cost Optimal MDP Solution

Shifrin, Mark; Mitrany, Roy; Biton, Erez; Gurewitz, Omer

doi:10.1109/tcc.2020.3000956

Cited by 13 publications

(11 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…MDP models are useful when the whole system is a discrete-time stochastic control process. It is also applied in addressing different research challenges of cloud computing such as resource management [32], QoS [33], resource and service failure management [34], etc. As we need to make the decision on the type and number of replicas for each component, which is dynamic in nature due to the respective failure probability, it is highly essential to model the problem with a tool that can take the fully observable states of this controlled systems into consideration.…”

Section: Total Number Of Replica Optimizationmentioning

confidence: 99%

RRFT: A Rank-Based Resource Aware Fault Tolerant Strategy for Cloud Platforms

Dehury,

Sahoo,

Veeravalli

2021

Preprint

View full text Add to dashboard Cite

The applications that are deployed in the cloud to provide services to the users encompass a large number of interconnected dependent cloud components. Multiple identical components are scheduled to run concurrently in order to handle unexpected failures and provide uninterrupted service to the end user, which introduces resource overhead problem for the cloud service provider. Furthermore such resource-intensive fault tolerant strategies bring extra monetary overhead to the cloud service provider and eventually to the cloud users. In order to address these issues, a novel fault tolerant strategy based on the significance level of each component is developed. The communication topology among the application components, their historical performance, failure rate, failure impact on other components, dependencies among them, etc., are used to rank those application components to further decide on the importance of one component over others. Based on the rank, a Markov Decision Process (MDP) model is presented to determine the number of replicas that varies from one component to another. A rigorous performance evaluation is carried out using some of the most common practically useful metrics such as, recovery time upon a fault, average number of components needed, number of parallel components successfully executed, etc., to quote a few, with similar component ranking and fault tolerant strategies. Simulation results demonstrate that the proposed algorithm reduces the required number of virtual and physical machines by approximately 10% and 4.2%, respectively, compared to other similar algorithms.

show abstract

Section: Total Number Of Replica Optimizationmentioning

confidence: 99%

RRFT: A Rank-Based Resource Aware Fault Tolerant Strategy for Cloud Platforms

Dehury,

Sahoo,

Veeravalli

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…A nearoptimal solution was proposed in [17] for power switching of two dynamic servers to minimize power consumption, delay, and wear-and-tear costs using MDP and look-ahead approach. The MDP approach is also used in [18] to find the optimal policy for cost-performance trade-off in virtualized data centers, where VMs are modeled by process sharing queues. Moreover, the size of the proposed MDP is reduced using fixed thresholds to categorize the load of servers into into three light, moderate, and high levels.…”

Section: State Of the Art And Prior Workmentioning

confidence: 99%

Energy Management in Data Centers with Server Setup Delay: A Semi-MDP Approximation

Chitsaz¹,

Khonsari²,

Moradian³

et al. 2021

Preprint

View full text Add to dashboard Cite

The energy management schemes in multi-server data centers with setup time mostly consider thresholds on the number of idle servers or waiting jobs to switch servers on or off. An optimal energy management policy can be characterized as a Markov decision process (MDP) at large, given that the system parameters evolve Markovian. The resulting optimal reward can be defined as the weighted sum of mean power usage and mean delay of requested jobs. For large-scale data centers however, these models become intractable due to the colossal state-action space, thus making conventional algorithms inefficient in finding the optimal policy. In this paper, we propose an approximate semi-MDP (SMDP) approach, known as 'multi-level SMDP', based on state aggregation and Markovian analysis of the system behavior. Rather than averaging the transition probabilities of aggregated states as in typical methods, we introduce an approximate Markovian framework for calculating the transition probabilities of the proposed multi-level SMDP accurately. Moreover, near-optimal performance can be attained at the expense of increased state-space dimensionality by tuning the number of levels in the multi-level approach. Simulation results show that the proposed approach reduces the SMDP size while yielding better rewards as against existing fixed threshold-based policies and aggregation methods.

show abstract

“…A VM's memory operations are divided into four categories: disk write, disk read, memory write, and memory read. In the pre-copy and post-copy models, we summarize the performance impacts of various processes during migration [5].…”

Section: Introductionmentioning

confidence: 99%

Live Migration of Virtual Machines Using a Mamdani Fuzzy Inference System

Alyas¹,

Javed²,

Namoun³

et al. 2022

Computers, Materials &Amp; Continua

View full text Add to dashboard Cite

Efforts were exerted to enhance the live virtual machines (VMs) migration, including performance improvements of the live migration of services to the cloud. The VMs empower the cloud users to store relevant data and resources. However, the utilization of servers has increased significantly because of the virtualization of computer systems, leading to a rise in power consumption and storage requirements by data centers, and thereby the running costs. Data center migration technologies are used to reduce risk, minimize downtime, and streamline and accelerate the data center move process. Indeed, several parameters, such as non-network overheads and downtime adjustment, may impact the live migration time and server downtime to a large extent. By virtualizing the network resources, the infrastructure as a service (IaaS) can be used dynamically to allocate the bandwidth to services and monitor the network flow routing. Due to the large amount of filthy retransmission, existing live migration systems still suffer from extensive downtime and significant performance degradation in crossdata-center situations. This study aims to minimize the energy consumption by restricting the VMs migration and switching off the guests depending on a threshold, thereby boosting the residual network bandwidth in the data center with a minimal breach of the service level agreement (SLA). In this research, we analyzed and evaluated the findings observed through simulating different parameters, like availability, downtime, and outage of VMs in data center processes. This new paradigm is composed of two forms of detection strategies in the live migration approach from the source host to the destination source machine.

show abstract

VM Scaling and Load Balancing via Cost Optimal MDP Solution

Cited by 13 publications

References 28 publications

RRFT: A Rank-Based Resource Aware Fault Tolerant Strategy for Cloud Platforms

RRFT: A Rank-Based Resource Aware Fault Tolerant Strategy for Cloud Platforms

Energy Management in Data Centers with Server Setup Delay: A Semi-MDP Approximation

Live Migration of Virtual Machines Using a Mamdani Fuzzy Inference System

Contact Info

Product

Resources

About