Modeling and Performance Analysis of Scalable Web Servers Deployed on the Cloud

Aljohani, A.M.; Holton, D. R. W.; Awan, Irfan-Ullah

doi:10.1109/bwcca.2013.45

Cited by 7 publications

(3 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Aljohani et al [1] propose a solution based on queuing theory. Its distinctive feature is that it considers that requests queue up in the application servers rather than in the load balancer.…”

Section: Related Workmentioning

confidence: 99%

Using Performance Forecasting to Accelerate Elasticity

Moura

Kon

Voulgaris³

et al. 2015

Adaptive Resource Management and Scheduling for Cloud Computing

View full text Add to dashboard Cite

Abstract. Cloud computing facilitates dynamic resource provisioning. The automation of resource management, known as elasticity, has been subject to much research. Monitoring of a running service plays a crucial role, and adjustments are made when certain thresholds are crossed. On such occasions, it is common practice to simply add or remove resources. In this paper we ask ourselves how we can predict the performance of a service in order to dynamically adjust allocated resources based on predictions. In other words, instead of "repairing" because a threshold has been crossed, we attempt to stay ahead and allocate a best amount of resources in advance. To do so, we need to have accurate predictive models that are based on workloads. We present our approach, based on the Universal Scalability Law, and discuss initial experiments.

show abstract

Section: Related Workmentioning

confidence: 99%

Using Performance Forecasting to Accelerate Elasticity

Moura

Kon

Voulgaris³

et al. 2015

Adaptive Resource Management and Scheduling for Cloud Computing

View full text Add to dashboard Cite

show abstract

“…A similar study is performed by Dai et al [13] where a sequential queuing network is used as the fundamental model. Aljohani et al [14] introduce a simple queueing model to analyze the performance metrics of error-free elastic cloud servers under varying traffic loads. In the presented model, two thresholds were used to control the action of elastic scaling.…”

mentioning

confidence: 99%

On the performance and power consumption analysis of elastic clouds

Guo

Pang

et al. 2016

Concurrency and Computation

View full text Add to dashboard Cite

Cloud computing is a novel paradigm capable of rationalizing the use of computational resources by means of outsourcing and virtualization. Elasticity is one of the most attractive features of cloud computing. Elastic clouds are able to adapt to workload changes by provisioning and de-provisioning resources in an autonomic manner, such that at each point in time the available resources match the current demand as closely as possible. However, elasticity adds complexity, which makes quantitative analysis of cloud performance and power consumption difficult. Such analysis is required to evaluate and quantify the cost-benefit of a strategy portfolio and the quantitative runtime performance and power consumption experienced by cloud-users. In this study, we present a comprehensive analytical approach to performance and power consumption analysis of elastic clouds. Several metrics are defined and evaluated: expected task completion time, power consumption rate, and task rejection rate under different load conditions, elasticity intensities, and error intensities. To validate the proposed approach, we obtain experimental data through a real-world cloud and conduct a confidence interval analysis. The analysis results suggest the perfect coverage of theoretical results by corresponding experimental confidence intervals. 4368 K.-Y. GUO ET AL.up/out as long as the workload is high, and scaling back in/down when possible, which potentially brings energy saving and operational cost reduction. Elasticity is one of the essential characteristics of scalable cloud, which distinguish the cloud paradigm from traditional computing systems such as gird and cluster computing. However, elasticity is not always welcomed because it also leads to nonnegligible performance degradations caused by lack of available computing resources when facing bursty/high workload and extra operational cost needed by up-scaling/down-scaling activities.Another notable feature influencing performance and power consumption of elastic cloud is machine error. Such errors are caused by accidental storage unavailability, component malfunction, software bug, software aging, storage space fragmentation accumulation, memory leak, extra-longresponse message, connection failure, message loss, wrong input sequence, external intrusion, and so on. Errors are a major reason causing performance and availability degradation in cloud computing systems. Such degradation could be further amplified because of the fact that elastic cloud usually maintains a subset of active machines rather than put all machines online. Such reduced number of active machines are obviously more susceptible to machine errors than those of non-elastic clouds where all machines are kept active.Errors occur at different levels, and not all errors especially hardware and platform-related errors can be detected and repaired by operating system-level (OS-level) detectors. The complexity of error hierarchy and the fact that lower-level hardware/platform errors are more rare than higherlevel software/tas...

show abstract

“…Aljohani et al[AHA13] model a server as an M/M/n queue (a queue with exponential arrival rate and service time distributions, and n service units), where n is the pool size, and uses thresholds based on the number of requests in the system.A dierent strategy to improve the resource management is learning from history on-the-y.Vasi¢ et al [VNM + 12] experimented with numerous o-the-shelf machine-learning techniques, re-3.2 porting good results with Bayesian models and decision trees. Gong et al [GGW10] use signal processing techniques to nd patterns in workload and resource usage.…”

mentioning

confidence: 99%

Alocação dinâmica de recursos em sistemas elásticos baseada em modelos de escalabilidade

Moura¹

View full text Add to dashboard Cite

First, I would like to thank my advisor, Fabio Kon, for the support and guidance along this thesis. He taught me a lot in this period. I also would like to thank professors Maarten van Steen and Spyros Voulgaris, who received me in the Netherlands, rst at VU Amsterdam, and later at the University of Twente. I learned a lot from their advices and they also provided me access to the infrastructure in which I executed most of the experiments in this thesis. I thank both of them for the valuable time they invested in helping me with my research. Finally, thank you all to my colleagues from our research group at IME and my family. Specially to my wife, Nathália.

show abstract

Modeling and Performance Analysis of Scalable Web Servers Deployed on the Cloud

Cited by 7 publications

References 13 publications

Using Performance Forecasting to Accelerate Elasticity

Using Performance Forecasting to Accelerate Elasticity

On the performance and power consumption analysis of elastic clouds

Alocação dinâmica de recursos em sistemas elásticos baseada em modelos de escalabilidade

Contact Info

Product

Resources

About