Predictable Cloud Provisioning Using Analysis of User Resource Usage Patterns in Virtualized Environment

Kim, Hyukho; Kim, Woongsup; Kim, Yangwoo

doi:10.1007/978-3-642-17625-8_9

Cited by 5 publications

(4 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The research of Bobroff, Kochut, and Beaty presented in has its focus on the dynamic placement of virtual machines, but workload forecasting is covered by the application of static ARMA processes for demand prediction. In , an approach is proposed and evaluated that classifies the gradients of a sliding window as a trend estimation on which the resource provisioning decision are then based. The authors Grunske, Aymin, and Colman of focus on QoS forecasting such as response time and propose an automated and scenario specific‐enhanced forecast approach that uses a combination of ARMA and GARCH stochastic process modeling frameworks for frequency‐based representation of time series data.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Self‐adaptive workload classification and forecasting for proactive resource provisioning

Herbst

Huber

Kounev

et al. 2014

Concurrency and Computation

View full text Add to dashboard Cite

As modern enterprise software systems become increasingly dynamic, workload forecasting techniques are gaining an importance as a foundation for online capacity planning and resource management. Time series analysis offers a broad spectrum of methods to calculate workload forecasts based on history monitoring data. Related work in the field of workload forecasting mostly concentrates on evaluating specific methods and their individual optimisation potential or on predicting QoS metrics directly. As a basis, we present a survey on established forecasting methods of the time series analysis concerning their benefits and drawbacks and group them according to their computational overheads. In this paper, we propose a novel self-adaptive approach that selects suitable forecasting methods for a given context based on a decision tree and direct feedback cycles together with a corresponding implementation. The user needs to provide only his general forecasting objectives. In several experiments and case studies based on real-world workload traces, we show that our implementation of the approach provides continuous and reliable forecast results at runtime. The results of this extensive evaluation show that the relative error of the individual forecast points is significantly reduced compared with statically applied forecasting methods, for example, in an exemplary scenario on average by 37%. In a case study, between 55 and 75% of the violations of a given service level objective can be prevented by applying proactive resource provisioning based on the forecast results of our implementation. N. R. HERBST ET AL.forecasting method executions on a machine with a Intel Core i7 CPU (2.7 GHz). The forecasting methods make use of only a single core as multi-threading is not yet fully supported by the existing implementations. APPROACH FOR WORKLOAD CLASSIFICATION AND FORECASTINGIn this section, we present a self-adaptive WCF process that can be used for selecting suitable forecasting methods at run-time.Over time, a WIB may change and develop in a way that affects its characteristics, that is, the class of the WIB is not fixed and needs to be updated periodically. Therefore, our classification process must be self-adaptive. Therefore, it must consider changes of the WIB and of given forecasting objectives to adapt the assignment of appropriate forecasting methods to given WIBs.An overview of the WCF process is sketched in Figure 2. Input of the WCF process is a trace of a WIB, a set of forecasting objectives, and possible feedback about the accuracy of the previous forecast. Essentially, we distinguish two phases in this process, the CLASSIFICATION phase and the FORECASTING phase. In the CLASSIFICATION phase, we extract the characteristics of a given WIB trace. Based on the identified characteristics, we use a decision tree (cf. 3) to classify the WIB and to select suitable forecasting methods. The assignment of forecasting methods also defines the class of the WIB. Then, in the FORECASTING phase, we apply the assigned forecasting...

show abstract

Section: Related Workmentioning

confidence: 99%

“…Related research in the field of proactive resource provisioning as it can be found in [3][4][5][6][7][8] mostly concentrates on single forecasting methods of the time series analysis and their individual optimisation potential. This way, reliable forecast results are achieved only in certain situations.…”

Section: Introductionmentioning

confidence: 99%

Self‐adaptive workload classification and forecasting for proactive resource provisioning

Herbst

Huber

Kounev

et al. 2014

Concurrency and Computation

View full text Add to dashboard Cite

show abstract

“…The research of Bobroff, Kochut and Beaty presented in [7] has its focus on the dynamic placement of virtual machines, but workload forecasting is covered by the application of static ARMA processes for demand prediction. In [20], an approach is proposed and evaluated that classifies the gradients of a sliding window as a trend estimation on which the resource provisioning decision are then based. The authors Grunske, Aymin and Colman of [3,2] focus on QoS forecasting such as response time and propose an automated and scenario specific enhanced forecast approach that uses a combination of ARMA and GARCH stochastic process modeling frameworks for frequency based representation of time series data.…”

Section: Related Workmentioning

confidence: 99%

Self-adaptive workload classification and forecasting for proactive resource provisioning

Herbst

Huber

Kounev

et al. 2013

Proceedings of the 4th ACM/SPEC International Conference on Performance Engineering

View full text Add to dashboard Cite

As modern enterprise software systems become increasingly dynamic, workload forecasting techniques are gaining in importance as a foundation for online capacity planning and resource management. Time series analysis offers a broad spectrum of methods to calculate workload forecasts based on history monitoring data. Related work in the field of workload forecasting mostly concentrates on evaluating specific methods and their individual optimisation potential or on predicting Quality-of-Service (QoS) metrics directly. As a basis, we present a survey on established forecasting methods of the time series analysis concerning their benefits and drawbacks and group them according to their computational overheads. In this paper, we propose a novel selfadaptive approach that selects suitable forecasting methods for a given context based on a decision tree and direct feedback cycles together with a corresponding implementation. The user needs to provide only his general forecasting objectives. In several experiments and case studies based on realworld workload traces, we show that our implementation of the approach provides continuous and reliable forecast results at run-time. The results of this extensive evaluation show that the relative error of the individual forecast points is significantly reduced compared to statically applied forecasting methods, e.g. in an exemplary scenario on average by 37%. In a case study, between 55% and 75% of the violations of a given service level agreement can be prevented by applying proactive resource provisioning based on the forecast results of our implementation.

show abstract

“…The SaaS providers also guarantee meeting quality of service (QoS) requirements of the SaaS users. Kim et al (2010) proposed a prediction-based resource provisioning model with which cloud system can analyze the resource usage history and predict the needed resource amount in advance before applications start requesting new or additional resources. This model employed resource usage history to find out the best-fit usage pattern at the given time window that determine whether cloud system allocates additional resources to guarantee performance or release resources to prevent resource overprovisioning.…”

Section: Resource Provisioningmentioning

confidence: 99%

A Value Based Dynamic Resource Provisioning Model in Cloud

Sood

2013

International Journal of Cloud Applications and Computing

View full text Add to dashboard Cite

Cloud computing has become an innovative computing paradigm, which aims at providing reliable, customized, Quality of Service (QoS) and guaranteed computing infrastructures for users. Efficient resource provisioning is required in cloud for effective resource utilization. For resource provisioning, cloud provides virtualized computing resources that are dynamically scalable. This property of cloud differentiates it from the traditional computing paradigm. But the initialization of a new virtual instance causes a several minutes delay in the hardware resource allocation. Furthermore, cloud provides a fault tolerant service to its clients using the virtualization. But, in order to attain higher resource utilization over this technology, a technique or a strategy is needed using which virtual machines can be deployed over physical machines by predicting its need in advance so that the delay can be avoided. To address these issues, a value based prediction model in this paper is proposed for resource provisioning in which a resource manager is used for dynamically allocating or releasing a virtual machine depending upon the resource usage rate. In order to know the recent resource usage rate, the resource manager uses sliding window to analyze the resource usage rate and to predict the system behavior in advance. By predicting the resource requirements in advance, a lot of processing time can be saved. Earlier, a server has to perform all the calculations regarding the resource usage that in turn wastes a lot of processing power thus decreasing its overall capacity to handle the incoming request. The main feature of the proposed model is that a lot of load is being shifted from the individual server to the resource manager as it performs all the calculations and therefore the server is free to handle the incoming requests to its full capacity.

show abstract

Predictable Cloud Provisioning Using Analysis of User Resource Usage Patterns in Virtualized Environment

Cited by 5 publications

References 13 publications

Self‐adaptive workload classification and forecasting for proactive resource provisioning

Self‐adaptive workload classification and forecasting for proactive resource provisioning

Self-adaptive workload classification and forecasting for proactive resource provisioning

A Value Based Dynamic Resource Provisioning Model in Cloud

Contact Info

Product

Resources

About