Browse Prior Art Database

Performance prediction technique by combining the logical resource consumption model and physical device models in virtualized environment

IP.com Disclosure Number: IPCOM000234701D
Publication Date: 2014-Jan-29
Document File: 3 page(s) / 106K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a new method of constructing performance models for cloud applications. The solution separates the components of the performance model into two parts: a logical application resource consumption model and a physical device model. The method also includes a performance analysis module.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 51% of the total text.

Page 01 of 3

Performance prediction technique by combining the logical resource consumption model and physical device models in virtualized environment

This article addresses the performance prediction problem in the virtualized /cloud environment. In the virtualized/cloud environment, it is critical that administrators or the management systems be able to predict the performance of live user applications running inside virtual machines (VM) because of the dynamic nature of the cloud . VMs

are constantly created/deleted and migrated, so resources such as the Central Processing Unit (CPU), memory, and disk must be adjusted. Without having good information about the performance implications of such actions , however, the VMs' performance may be impacted and result in costly Service Level Agreement (SLA) violations.

Existing techniques to address the issue include various works in the area of performance modeling. However, performance modeling techniques are difficult to apply because of heterogeneity of the physical servers. One performance model trained and tuned for one machine does not translate well to the different physical machines. Since application characteristics vary from one VM to another , it is not feasible to build and maintain the performance models whose number equals the number of target applications multiplied by the hardware types .

The novel contribution is a method of constructing performance models for cloud applications. The solution separates the components of the performance model into two parts: a logical application resource consumption model and a physical device model. The logical resource consumption model is a model that describes how much CPU, memory, disk, and network resources are consumed per each request type of a given application (or a VM that hosts it). This model is independent of hardware (HW) configurations. The second model, physical device mode, is the performance model of individual HW devices within each physical server. This is independent of the

workloads or applications within the VM.

These models can be built using any known techniques, but act independently of any applications that might be running on top of it . The logical resource consumption model explains how to combine such device models. Without having the logical resource consumption model, it is difficult to determine how each device is going to be used for a given application. For example, it is not clear whether the online bookstore application

will be is...