Workload modeling for resource usage analysis and simulation in cloud computing

Highlights

• We model a web application to support analysis and simulation of cloud environments.

• We implement the model as an extension of the CloudSim simulator.

• We found that the user behavior has a strong influence on the resource utilization.

• We found Generalized Extreme Value and Generalized Lambda distributions instead of Exponential distribution to represent session time.

Abstract

Workload modeling enables performance analysis and simulation of cloud resource management policies, which allows cloud providers to improve their systems’ Quality of Service (QoS) and researchers to evaluate new policies without deploying expensive large scale environments. However, workload modeling is challenging in the context of cloud computing due to the virtualization layer overhead, insufficient tracelogs available for analysis, and complex workloads. These factors contribute to a lack of methodologies and models to characterize applications hosted in the cloud. To tackle the above issues, we propose a web application model to capture the behavioral patterns of different user profiles and to support analysis and simulation of resources utilization in cloud environments. A model validation was performed using graphic and statistical hypothesis methods. An implementation of our model is provided as an extension of the CloudSim simulator.

Introduction

Clouds are being used as a platform for various types of applications with different Quality of Service (QoS) aspects, such as performance, availability and reliability. These aspects are specified in a Service Level Agreement (SLA) negotiated between cloud providers and customers. The failure to comply with QoS aspects can compromise the responsiveness and availability of service and incur SLA violations, resulting in penalties to the cloud provider. The development of resource management policies that support QoS is challenging and the evaluation of these policies is even more challenging because clouds observe varying demand, their physical infrastructure has different sizes, software stacks, and physical resources configurations, and users have different profiles and QoS requirements [1]. In addition, reproduction of conditions under which the policies are evaluated and control of evaluation conditions are difficult tasks.

In this context, workload modeling enables performance analysis and simulation, which brings benefits to cloud providers and researchers. Thereby, the evaluation and adjustment of policies can be performed without deployment of expensive large scale environments. Workload models have the advantage of allowing workload adjustment to fit particular situations, controlled modification of parameters, repetition of evaluation conditions, inclusion of additional features, and generalization of patterns found in the application [2], providing a controlled input for researchers. For cloud providers, the evaluation and simulation of resource management policies allow the improvement of their systems’ QoS. Finally, the simulation of workloads based on realistic scenarios enables the production of tracelogs, scarce in cloud environments because of business and confidentiality concerns [3], [4].

Workload modeling and characterization is especially challenging when applied in a highly dynamic environment, such as cloud data centers, for different reasons:(i) heterogeneous hardware is present in a single data center and the virtualization layer incurs overhead caused by I/O processing and interactions with the Virtual Machine Monitor (VMM); and (ii) complex workloads are composed of a wide variety of applications submitted at any time, and with different characteristics and user profiles. These factors contribute to a lack of methodologies to characterize the different behavioral patterns of cloud applications.

To tackle the above issues, a web application model able to capture the behavioral patterns of different user profiles is proposed to support analysis and simulation of resources utilization in cloud environments. The proposed model supports the construction of performance models used by several research domains. Performance models improve resource management because they allow the prediction of how application patterns will change. Thus, resources can be dynamically scaled to meet the expected demand. This is critical to cloud providers that need to provision resources quickly to meet a growing resource demand by their applications.

In this context, the main contribution of this paper is a model capable of representing resource demand of Web application supported by different user profiles in a context of cloud environment. The workload patterns are modeled in the form of statistical distributions. Therefore, the patterns fluctuate based on realistic parameters in order to represent dynamic environments. A model validation is provided through graphical and analytical methods in order to show that the model effectively represents the observed patterns. A secondary contribution of this paper is the validation and implementation of the proposed model as an extension of the CloudSim simulator [1], making the model available for the cloud research community.

The rest of the paper is organized as follows: Section 2 presents the challenges and importance of workload modeling in clouds. Section 3 describes related works. Section 4 details the adopted methodology and how it was achieved. Section 5 presents and discusses the modeling and simulation results. Section 6 concludes and defines future research directions.