for Cloud Computing

beeuppityΤεχνίτη Νοημοσύνη και Ρομποτική

19 Οκτ 2013 (πριν από 3 χρόνια και 5 μήνες)

63 εμφανίσεις

SLA
-
Oriented Resource Provisioning
for Cloud Computing


Challenges,
Architecture
, and
Solutions

1

Author

2

Content


Abstract


Introduction


Challenges and Requirements


SLA
-
Oriented Cloud Computing Vision


State
-
of
-
the
-
art


System Architecture


SLA Provisioning in Aneka


Performance Evaluation


Future Directions

3

Abstract


Need to offer differentiated services to users and meet
their quality expectations.


Existing resource management systems are yet to support
SLA
-
oriented resource allocation.


No work has been done to collectively incorporate
customer
-
driven service management, computational risk
management, and autonomic resource management into a
market
-
based resource management system to target the
rapidly changing enterprise requirements of Cloud
computing.


This paper presents vision, challenges, and architectural
elements of SLA
-
oriented resource management.

4

Introduction

5


There are dramatic differences between developing
software for millions to use as a service versus
distributing software for millions to run their PCs





--

Professor David Patterson



New Computing
Paradigms


Cloud Computing


Grid Computing


P2P Computing



Utility Computing


Challenges and Requirements
-

1

6

Challenges and Requirements
-

2


Customer
-
driven Service Management


Computational Risk Management


Autonomic Resource Management


SLA
-
oriented Resource Allocation Through
Virtualization


Service Benchmarking and Measurement


System Modeling and Repeatable Evaluation


7

SLA
-
Oriented Cloud Computing Vision

The resource provisioning will be driven by market
-
oriented
principles for efficient resource allocation depending on user
QoS

targets and workload demand patterns.


Support for customer
-
driven service management based on customer profiles and
QoS

requirements;


Definition of computational risk management tactics to identify, assess, and manage
risks involved in the execution of applications;


Derivation of appropriate market
-
based resource management strategies that
encompass both customer
-
driven service management and computational risk
management to sustain SLA
-
oriented resource allocation;


Incorporation of autonomic resource management models;


Leverage of Virtual Machine technology to dynamically assign resource shares;


Implementation of the developed resource management strategies and models into
a real computing server;

8

State
-
of
-
the
-
art


Traditional Resource Management Systems(Condor,
LoadLeveler
, Load Sharing Facility, Portable Batch System)


adopt system
-
centric resource allocation approaches that focus on
optimizing overall cluster performance


Increase processor throughput and utilization for the cluster


Reduce the average waiting time and response time for jobs


Assume that all job requests are of equal user importance and neglect
actual levels of service required by different users.


Virtual Machine management platform solutions(Eucalyptus,
OpenStack
, Apache VCL, Citrix Essentials)


Main goal
is to provide automatic configuration and maintenance of the
centers


Market
-
based resource management


Not considered and incorporated customer
-
driven service management,
computational risk management, and autonomic resource management
into market
-
driven resource management

9

System Architecture

10

High
-
level system architectural framework

SLA Provisioning in Aneka
-
1

11

Aneka architecture

SLA Provisioning in Aneka
-

2

12

SLA Provisioning in Aneka
-

3

13

SLA Provisioning in Aneka
-

4

14

Performance Evaluation
-

1


Static resource


1 Aneka master
-

m1.large(7.5GB memory, 4 EC2
compute units, 850GB instance storage, 64bit platform,
US0.48 per instance per hour) Windows
-
based VM


4 Aneka workers


m1.small(1.7GB memory, 1 EC2
compute unit, 160GB instance storage, 32bit platform,
US0.085 per instance per hour)
L
inux
-
based VM


Dynamic resources


m1.small Linux
-
based instances

15

Performance
Evaluation
-

2


CPU
-
intensive application


SLA is defined in terms of user
-
defined deadline


execution time of each task was set to 2 minutes


Each job consists of 120 tasks

16

Conclusions and Future Directions


The need for a deeper investigation in SLA
-
oriented resource allocation strategies that
encompass:


Customer
-
driven service management


Computational risk management


Autonomic management of Clouds

In order to:


Improve the system efficiency


Minimize violation of SLAs


Improve profitability of service providers




17

Thanks !

18