Resource Allocation for Autonomic Data Centers using Analytic Performance Models

Bennani, Mohamed N.; Menascé, Daniel A.

doi:10.1109/icac.2005.50

Cited by 274 publications

(196 citation statements)

References 7 publications

Supporting

Mentioning

196

Contrasting

Order By: Relevance

“…We here instead focus on closed systems, where a fixed population of customers cyclically issues requests to the servers; in these models, the problem becomes even more difficult because, besides remaining non-convex, a single evaluation of the objective function is often much more expensive [17,24,38]. This holds because non-iterative expressions for the stationary performance indexes are lacking for multiclass closed models, thus most formulations require to use gradient-hill methods in combination with some model evaluation technique [10,17,24], e.g., approximate mean-value analysis (MVA) [15,33,43]. These formulations are popular in applications [5,16,31,39] but, as we show in the paper, they often fail to converge to a local optimum at the short timescales of minutes at which online capacity management systems operate today.…”

Section: Introductionmentioning

confidence: 99%

“…The gradient methods (bs-grad, aql-grad) have been used to establish the competitiveness of approaches that estimate the throughputs X ir by making an external call to an approximate MVA solver. Such approaches are probably the most common in applications [10], together with the basic ut formulation that is equivalent to reasoning based on utilization constraints only. Also, notice that the fpm formulation is immediately related to an open network of M/GI/1 processor sharing queues operating with an average population not greater than N r for each class.…”

Section: Solution Techniquesmentioning

confidence: 99%

See 1 more Smart Citation

Heavy-traffic revenue maximization in parallel multiclass queues

Anselmi

Casale

2013

Performance Evaluation

View full text Add to dashboard Cite

Motivated by revenue maximization in server farms with admission control, we investigate optimal scheduling in parallel processor-sharing queues. Incoming customers are distinguished in multiple classes and we define revenue as a weighted sum of class throughputs. Under these assumptions, we describe a heavy-traffic limit for the revenue maximization problem and study the asymptotic properties of the optimization model as the number of clients increases. Our main result is a simple heuristic that is able to provide tight guarantees on the optimality gap of its solutions. In the general case with M queues and R classes, we prove that our heuristic is (1 + 1 M −1 )-competitive in heavy-traffic. Experimental results indicate that the proposed heuristic is remarkably accurate, despite its negligible computational costs, both in random instances and using service rates of a web application measured on multiple cloud deployments. 1

show abstract

Section: Introductionmentioning

confidence: 99%

Section: Solution Techniquesmentioning

confidence: 99%

Heavy-traffic revenue maximization in parallel multiclass queues

Anselmi

Casale

2013

Performance Evaluation

View full text Add to dashboard Cite

show abstract

“…Analytical models were used for the first time in [24] for predicting performance and exercising QoS control for e-commerce sites [18,24] and for Web servers [25]. Bennani and Menascé used online analytic models for dynamic server allocation in autonomic data centres [2]. Others used the same approach and applied the idea to performance management of cluster-based Web services [16].…”

Section: Concluding Remarks and Future Workmentioning

confidence: 99%

QoS management in service-oriented architectures

Menascé

Ruan

Gomaa

2007

Performance Evaluation

View full text Add to dashboard Cite

The next generation of software systems will be highly distributed, component-based and service-oriented. They will need to operate in unattended mode and possibly in hostile environments, will be composed of a large number of 'replaceable' components discoverable at run-time, and will have to run on a multitude of unknown and heterogeneous hardware and network platforms. This paper focuses on QoS management in service-oriented architectures in which service providers (SP) provide a set of interrelated services to service consumers, and a QoS broker mediates QoS negotiations between SPs and consumers. The main contributions of this paper are: (i) the description of an architecture that includes a QoS broker and service provider software components, (ii) the specification of a secure protocol for QoS negotiation with the support of a QoS broker, (iii) the specification of an admission control mechanism used by SPs, (iv) a report on the implementation of the QoS broker and SPs, and (v) the experimental validation of the ideas presented in the paper.

show abstract

“…We use a function definition from earlier work on resource allocation in data centers [16] (which generates the curve illustrated in Fig. 2 for a target time of 50):…”

Section: B Utility Functions 1) Utility Based On Response Timementioning

confidence: 99%

“…Several researchers have reported the use of utility functions in autonomic computing, typically to support systems management tasks (e.g. [15], [16]); to the best of our understanding this is the first attempt to deploy utility functions for adaptive workflow execution.…”

Section: Introductionmentioning

confidence: 99%

Supporting IPv6 Interaction with Wireless Sensor Networks Using NP++

Jakeman

Hughes

Coulson

et al. 2008

Wireless Algorithms, Systems, and Applications

View full text Add to dashboard Cite

Abstract-Workflows are widely used in applications that require coordinated use of computational resources. Workflow definition languages typically abstract over some aspects of the way in which a workflow is to be executed, such as the level of parallelism to be used or the physical resources to be deployed. As a result, a workflow management system has responsibility for establishing how best to map tasks within a workflow to the available resources. As workflows are typically run over shared resources, and thus face unpredictable and changing resource capabilties, there may be benefit to be derived from adapting the task-to-resource mapping while a workflow is executing. This paper describes the use of utility functions to express the relative merits of alternative mappings; in essence, a utility function can be used to give a score to a candidate mapping, and the exploration of alternative mappings can be cast as an optimization problem. In this approach, changing the utility function allows adaptations to be carried out with a view to meeting different objectives. The contributions of this paper include: (i) a description of how adaptive workflow execution can be expressed as an optimization problem where the objective of the adaptation is to maximize some property expressed as a utility function; (ii) a description of how the approach has been applied to support adaptive workflow execution in grids; and (iii) an experimental evaluation of the resulting approach for alternative utility measures based on response time and profit.

show abstract

Resource Allocation for Autonomic Data Centers using Analytic Performance Models

Cited by 274 publications

References 7 publications

Heavy-traffic revenue maximization in parallel multiclass queues

Heavy-traffic revenue maximization in parallel multiclass queues

QoS management in service-oriented architectures

Supporting IPv6 Interaction with Wireless Sensor Networks Using NP++

Contact Info

Product

Resources

About