Project presentation

15
Resource usage on different Linux distributions hosting KVM An application perspective Nitin Reddy Ayyagari Akash Reddy Malkannagari Group C4

Transcript of Project presentation

Resource usage on different Linux distributions hosting KVM

An application perspective

Nitin Reddy AyyagariAkash Reddy Malkannagari

Group C4

Motivation

Virtualization & Cloud based products-very popular

Eg.: Online Drive, Virtualized systems Facilitated in Data Centers- Millions of Servers Companies migrating applications to Cloud Why? Cost reduction, Scalability and Reliability Identifying suitable configuration- Suitable Host

OS for KVM hypervisor based on applications Optimum utilization of computing resources,

lower installation costs and better environment

Research Problem

Different type of applications used in companies

Eg: Database, ERP etc. Linux – preferred choice as host OS in Kernel based

Virtual Machine, KVM Plethora of Linux distributions available, Usage of

single distribution not ideal for all applications Which implementation can lead to optimum

utilization of available computing resources?

Research Question

Which Linux distribution hosting a KVM, would be better suited for running different-grade applications in terms of the resources they consume?

Linux distributions – Ubuntu, CentOS and openSUSE

Identification based on the type of application(low, medium and high) being run on guest OS.

Memory utilization and CPU utilization- key criteria Usage statistics only on Host OS, not guest OS.

Main contribution

Hosting KVM on each of Ubuntu, CentOS and openSUSE operating systems

Run three applications Skype, YouTube and Bash script on a uniform guest OS on KVM

Collect Memory utilization and CPU utilization statistics on host OS

Analyze the results, average utilization tabulated and graphs plotted

Procedure

Collect statistics on Host system

Low Level

• Bash Script

Medium level

• YouTube

High level

• Skype

KVM Architecture

Basis for our work

Hypervisor: KVM is open source, uses virtualization extensions of Linux kernel

Distribution choice: Ubuntu, CentOS and openSUSE are widely used Linux distributions

Applications chosen: Based on compatibility with the three Operating Systems

Parameters chosen: CPU Utilization and Memory Utilization, generally used parameters for resource usage

Observations I

Bash You Tube

Skype

Ubuntu 52.513 68.571 69.396

CentOS 58.565 94.312 97.514

openSUSE

90.323 93.755 97.667

Bash Youtube Skype0

10

20

30

40

50

60

70

80

90

100

Ubuntu

CentOS

OpenSUSE

Aver

age

Mem

ory

Utiliz

ation

(in %

)

Table I. Average Memory Utilization(in %)

Bash You Tube SkypeUbuntu 25.831 31.545 54.159

CentOS 24.387 55.689 51.842openSUSE 54.912 64.516 58.148

Bash Youtube Skype0

10

20

30

40

50

60

70

80

90

100

Ubuntu

CentOS

OpenSUSE

Aver

age

CPU

Utiliz

ation

(in %

)

Table II. Average CPU Utilization(in %)

Observations II

In terms of memory utilization, Ubuntu proved to be efficient for Medium and High level Applications.

CentOS had better performance for Low and High level Applications in terms of CPU Utilization

Conclusions I

Based on experimental results, for: Low level applications : Ubuntu and CentOS show

similar usage statistics. Medium Level applications : Ubuntu is the ideal OS. High Level applications : Ubuntu has better

performance in terms of memory utilization while CentOS is a better choice in terms of CPU utilization.

Conclusions II

The End

Main contribution ambiguity. We are intrested in collecting statistics of host Operating Sysytem, OS. Guest OS remains the same throughout the experiment. “Ideal configuration plays vital role” was used. Configuration of virtual machine and the

CPU, memory allocation were neglected. In paper we mentioned the sentence as ”In a data-center capable of hosting thousands of virtual instances, identification of ideal configuration plays a vital role in meeting up Service-Level Agreements, SLA’s”. With this we meant identification of a suitable/ideal OS. Our main idea was to select a OS based on the resources being utilized by the applications. We have kept VM CPU & memory allocation same for all the three OS. Title is too long: Suggestion was given to change “Different Linux distribution” with “Different OS”. But we wanted to be more precise with our title as we used different Linux distributions. And a sub-tittle was given as it conveys that we are specifically dealing with application. “Resource usage” variable was not described in abstract. Yes, it not was not described. But we have explictily mentioned in Introduction and Main Contribution sections of the paper.The unit for statistics values in tables and figure is not specified.The average utilizations are in ”%”. We will specify it in the next version.

Reviewer Comments & our Answers

Future work not mentioned.“This work can be extended by testing customized applications and also other hypervisors like Xen and Vmware”. This was mentioned at the end conclusion section. Bash script,not a web application is used.Yes, we chose that as it runs on all three OS and serves the best purpose to depict a low level application. Screen Shots of data collections were not placed in the paper. Due to lack of space we could not provide in the paper. If the reviewer insists we are happy to provide in the appendix. Table I & II got interchanged. Yes, they were interchanged(but it was only the order) and we will change it in next version. Research question is not clear.Research question can be found in section III of the paper. The number of values taken into consideration for the calculation of these average

values has not been mentioned.We felt specifying average values would be sufficient. We will accomodate them in the next version.There were few other vocabulary corrections which were found by the reviwers and we appreciate that. They will be corrected appropriately in the next version of our paper.

Reviewer Comments & our Answers