Dynamic resource allocation for shared data centers using online measurements

  title={Dynamic resource allocation for shared data centers using online measurements},
  author={Abhishek Chandra and Weibo Gong and Prashant J. Shenoy},
  booktitle={Measurement and Modeling of Computer Systems},
Since web workloads are known to vary dynamically with time, in this paper, we argue that dynamic resource allocation techniques are necessary to provide guarantees to web applications running on shared data centers. To address this issue, we use a system architecture that combines online measurements with prediction and resource allocation techniques. To perform resource allocation, we model a server resource that services multiple applications as a generalized processor sharing (GPS) server… 

Figures from this paper

Dynamic resource allocation for shared data centers using online measurements

The main advantage of the techniques is that they capture the transient behavior of applications while incorporating nonlinearity in the system model, and can judiciously allocate system resources, especially under transient overload conditions.

Predictive Control for Dynamic Resource Allocation in Enterprise Data Centers

This paper studies the effect of prediction on dynamic resource allocation to virtualized servers running enterprise applications and presents predictive controllers using three different prediction algorithms based on a standard auto-regressive (AR) model, a combined ANOVA-AR model, as well as a multi-pulse (MP) model.

Combined Power and Performance Management of Virtualized Computing Environments Serving Session-Based Workloads

This paper develops an online resource provisioning framework for combined power and performance management in a virtualized computing environment serving session-based workloads using limited lookahead control (LLC), a form of model-predictive control.

A Heuristic Approach for Scalability of Multi-tiers Web Application in Clouds

The results indicate that the performance model faithfully captures the behaviors of multi-tiers applications over a various range of workloads and configuration schemes and shows that the techniques can judiciously obtain the optimized configuration scheme effectively with modest computation.

An empirical model of adaptive cloud resource provisioning with speculation

The focus of the work is to study how speculative analysis can be used to predict exact resources for an application, whose accuracy demands solution for under/over-utilization of the resource.

A statistical approach to virtual server resource management

This work proposes a novel approach to resource management of virtualized data center components that incorporates a resource controller, based on Statistical Process Control, which permits the online management of a virtual machine's processor and memory capacity through real‐time analysis of its observed performance.

A Pattern-Based Prediction Model for Dynamic Resource Provisioning in Cloud Environment

A pattern-based prediction model for resource provisioning is proposed which facilitates best possible resource preparation by analyzing the resource utilization and deriving resource usage patterns and gives significant benefits over conventional models.

Self-Aware Resource Management in Virtualized Data Centers

Enterprise applications in virtualized data centers are often subject to timevarying workloads, i.e., the load intensity and request mix change over time, due to seasonal patterns and trends, or

Dynamic resource allocation of shared data centers supporting multiclass requests

This paper considers a shared data center with requests from multiple companies: each company with two classes of request: 1) streaming requests and 2) elastic requests, and develops a cost model to obtain a request admission control policy that utilizes the Web server's processing capacity efficiently.

Reconciling Dynamic System Sizing and Content Locality through Hierarchical Workload Forecasting

This work addresses the problem of reconciling dynamic system sizing and content locality by employing a hierarchical prediction framework that proactively provisions resources based on statistical models of the incoming workload and proposes two novel prediction-based locality-aware request distribution policies.



Dynamic resource allocation for shared data centers using online measurements

The main advantage of the techniques is that they capture the transient behavior of applications while incorporating nonlinearity in the system model, and can judiciously allocate system resources, especially under transient overload conditions.

QoS-driven server migration for Internet data centers

This paper develops a framework for QoS-driven dynamic resource allocation in IDCs, called QuID (quality of service infrastructure on demand), and develops an optimal off-line algorithm that bounds the advantage of any dynamic policy and provides a benchmark for performance evaluation.

Model-Based Resource Provisioning in a Web Service Utility

A new approach to utility resource management focusing on coordinated provisioning of memory and storage resources is presented, which incorporates internal models of service behavior to predict the value of candidate resource allotments under changing load.

An observation-based approach towards self-managing Web servers

This paper describes an observation-based approach for self-managing Web servers that can adapt to changing workloads while maintaining the QoS requirements of different classes and demonstrates the need to manage different resources in the system depending on the workload characteristics.

Managing energy and server resources in hosting centers

Experimental results from a prototype confirm that the system adapts to offered load and resource availability, and can reduce server energy usage by 29% or more for a typical Web workload.

An adaptive control framework for QoS guarantees and its application to differentiated caching

A new QoS-control paradigm based on adaptive control theory is introduced, which eliminates profiling and configuration costs ofQoS-aware software, by completely automating the process in a way that does not require user intervention.

A Resource Management Framework for Predictable Quality of Service in Web Servers

Experimental results obtained with a prototype implementation of this resource management framework on trace-based workloads show its effectiveness in providing desired QoS levels with high confidence, while achieving high average utilization of the hardware.

Application-aware admission control and scheduling in Web servers

  • Jakob CarlströmR. Rom
  • Computer Science
    Proceedings.Twenty-First Annual Joint Conference of the IEEE Computer and Communications Societies
  • 2002
A performance model, relying on online estimation of parameters describing client-server interaction, is presented, whereby a reward function corresponding to the service provider's objective is maximized using techniques for nonlinear optimization.

Adaptive Algorithms for Managing a Distributed Data Processing Workload

This paper presents algorithms developed to simplify performance management, dynamically adjust computing resources, and balance work across parallel systems that provide a single-system image to manage competing workloads running across multiple systems.

Admission control and dynamic adaptation for a proportional-delay diffserv-enabled web server

A web server that can provide differentiated services to clients with different QoS requirements is considered and two efficient admission control algorithms are proposed so that a web server can provide the QoS guarantees and classify each client to its "lowest" admissible class, resulting in lowest usage cost for the client.