Top Banner
Market-Oriented Cloud Computing: Vision, Hype, and Reality for Delivering IT Services as Computing Utilities Rajkumar Buyya(1,2), Chee Shin Yeo(1), and Srikumar Venugopa(l) 1.Grid Computing and Distributed Systems (GRIDS) Laboratory Department of computer Science and Software Engineering The University of Melbourne, Australia 2.Manjrasoft Pty Ltd, Melbourne, Australia HPCC '08. 10th IEEE 1
38

Rajkumar Buyya (1,2), Chee Shin Yeo (1), and Srikumar Venugopa (l)

Feb 23, 2016

Download

Documents

Tyrell

Market-Oriented Cloud Computing: Vision, Hype, and Reality for Delivering IT Services as Computing Utilities. Rajkumar Buyya (1,2), Chee Shin Yeo (1), and Srikumar Venugopa (l) - PowerPoint PPT Presentation
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

1

Market-Oriented Cloud Computing:Vision, Hype, and Reality for Delivering IT

Services as Computing Utilities

Rajkumar Buyya(1,2), Chee Shin Yeo(1), and Srikumar Venugopa(l)1.Grid Computing and Distributed Systems (GRIDS) Laboratory Department of computer

Science and Software Engineering The University of Melbourne, Australia

2.Manjrasoft Pty Ltd, Melbourne, AustraliaHPCC '08. 10th IEEE

Page 2: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

2

Outline Introduction Market-Oriented Cloud Architecture Commercial offering of market-oriented Clouds

requirement and Qos issue Emerging cloud platform Amazon EC2 intro&pricing Google App Engine intro&pricing Microsoft Anzure platform intro&pricing Possible pricing strategy(by Ming Lung) Conclusions&comments

Page 3: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

3

Introduction:definition Definition of cloud:

A Cloud is a type of parallel and distributed system;

Consisting of a collection of interconnected and virtualized computers That are dynamically provisioned and presented as

one or more unified computing resources ,based on service-level agreements established

through negotiation between the service provider and consumers.”

Page 4: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

4

Introduction:trend Web Search Trends:

[C]. Google and Salesforce.com in Cloud computing deal, Siliconrepublic.com - Apr 14 2008

Page 5: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

5

Market-Oriented Cloud Architecture Cloud providers will need to consider and

meet different QoS parameters of each individual consumer as negotiated in specific SLAs.

Traditional system-centric resource management architecture are no longer fit Do not provide incentives for them to share

their resources. Regard all service requests to be of equal

importance.

Page 6: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

6

Market-Oriented Cloud Architecture

Page 7: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

7

Market-Oriented Cloud Architecture

Pricing: The Pricing mechanism decides how service requests

are charged. Ex. submission time (peak/off-peak)

, pricing rates (fixed/changing) Accounting:

Maintains the actual usage of resources by requests and historical information usage. Final cost to charge users. Improve resource allocation decisions.

Service Request Examiner and Admission Control: Interprets the submitted request for QoS requirements

before determining whether to accept or reject the request.

Page 8: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

8

Market-Oriented Cloud Architecture

Dispatcher: starts the execution of accepted service

requests on allocated VMs. Service Request Monitor:

keeps track of the execution progress of service requests.

VM monitor: Keep track of the availability of VMs and their

resource entitlements.

Page 9: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

9

Qos parameter issue In cloud there are critical QoS parameters to

consider in a service request time, cost, reliability and trust/security.

In particular, QoS requirements cannot be static and need to be dynamically updated over time. Due to continuing changes in business

operations and operating environments. But , there are no or limited support for

dynamic negotiation of SLAs. Recently, we have developed negotiation

mechanisms based on alternate offers protocol for establishing SLAs [8].

[8]S. Venugopal, X. Chu, and R. Buyya. using the Alternate Offers Protocol (IWQoS 2008), A Negotiation Mechanism for Advance Resource Reservation

Page 10: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

10

Commercial offering of market-oriented Clouds requirement Customizable

Support customer-driven service management based on customer profiles and requested service requirements.

Market-based resource management Contain computational risk management to

sustain SLA-oriented resource allocation. Incorporate autonomic resource

management models: Effectively self-manage changes in service

requirements to satisfy both new service demands and existing service obligations.

Page 11: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

11

Emerging cloud platform

Page 12: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

12

Amazon EC2

Page 13: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

13

Amazon EC2 Instances Types (Memory / *ECU / Storage /

Platform) Standard Instances

Small (default): 1.7 GB / 1 / 160 GB / 32-bit Large: 7.5 GB / 4 / 850 GB / 64-bit Extra Large: 15 GB / 8 / 1690 GB / 64-bit

High-Memory Instances Double Extra Large: 34.2 GB / 13 / 850 GB / 64-bit Quadruple Extra Large: 68.4 GB / 26 / 1690 GB / 64-bit

High-CPU Instances Medium: 1.7 GB / 5 / 350 GB / 32-bit Extra Large: 7 GB / 20 / 1690 GB / 64-bit

http://aws.amazon.com/ec2/

Page 14: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

14

About Measuring Compute Resources(quote from Amazon)

*ECU – EC2 Compute Unit, providing the equivalent CPU capacity of a 1.0 – 1.2 GHz 2007 Opteron or 2007 Xeon processor

“Amazon EC2 uses a variety of measures to provide each instance with a consistent and predictable amount of CPU capacity.” We use several benchmarks and tests to manage the consistency

and predictability of the performance of an EC2 Compute Unit. Over time, we may add or substitute measures that go into the

definition of an EC2 Compute Unit, if we find metrics that will give you a clearer picture of compute capacity.

“To find out which instance will work best for your application, the best thing to do is to launch an instance and benchmark your own application.” pay by the hour

Page 15: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

15

On-Demand Instances

US – N. Virginia EU – IrelandStandard Instances Linux/UNIX Windows Linux/UNIX WindowsSmall (default) $0.085 $0.12 $0.095 $0.13Large $0.34 $0.48 $0.38 $0.52Extra Large $0.68 $0.96 $0.76 $1.04High-Memory Instances

Linux/UNIX Windows Linux/UNIX Windows

Double Extra Large $1.20 $1.44 $1.34 $1.58Quadruple Extra Large $2.40 $2.88 $2.68 $3.16High-CPU Instances Linux/UNIX Windows Linux/UNIX WindowsMedium $0.17 $0.29 $0.19 $0.31Extra Large $0.68 $1.16 $0.76 $1.24

Unit: Per Hour

Page 16: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

16

Reserved InstancesLinux/UNIX One-time fee US – N.

VirginiaUS – N. California & EU – Ireland

Standard Instances 1 yr 3 yr Usage ( /hr)

Usage ( /hr)

Small (default) $227.50

$350 $0.03 $0.04

Large $910 $1400 $0.12 $0.16Extra Large $1820 $2800 $0.24 $0.32High-Memory Instances

1 yr 3 yr Usage ( /hr)

Usage ( /hr)

Double Extra Large $3185 $4900 $0.42 $0.56Quadruple Extra Large $6370 $9800 $0.84 $1.12High-CPU Instances 1 yr 3 yr Usage (

/hr)Usage ( /hr)

Medium $455 $700 $0.06 $0.08Extra Large $1820 $2800 $0.24 $0.32

Page 17: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

17

Spot Instances Spot Instances enable you to bid for unused Amazon EC2 capacity. To use Spot Instances, you should set

(instance type, region, amount, maximum price)US – N. Virginia EU – Ireland

Standard Instances Linux/UNIX Windows Linux/UNIX WindowsSmall (default) $0.085 $0.12 $0.095 $0.13Large $0.34 $0.48 $0.38 $0.52Extra Large $0.68 $0.96 $0.76 $1.04High-Memory Instances

Linux/UNIX Windows Linux/UNIX Windows

Double Extra Large $1.20 $1.44 $1.34 $1.58Quadruple Extra Large $2.40 $2.88 $2.68 $3.16High-CPU Instances Linux/UNIX Windows Linux/UNIX WindowsMedium $0.17 $0.29 $0.19 $0.31Extra Large $0.68 $1.16 $0.76 $1.24

*fluctuates periodically depending on the supply of and demand for Spot Instance

Page 18: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

18

Data TransferInternet Data TransferData Transfer InAll Data Transfer Free through June 30, 2010*Data Transfer OutFirst 10 TB per Month $0.17 per GBNext 40 TB per Month $0.13 per GBNext 100 TB per Month $0.11 per GBOver 150 TB per Month $0.10 per GB

Data transferred between two Amazon Web Services within the same zone is free of charge.Data transferred between AWS services in same regions but different zone will be charged $0.01 per GB in/out.

Page 19: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

19

Amazon add-on services Amazon Elastic Block Store

Amazon EBS volumes provide off-instance storage that persists independently from the life of an instance.

Charged per GB/month and I/O request Amazon CloudWatch (bundle with Auto Scaling)

Amazon CloudWatch is a web service that provides monitoring for AWS cloud resources. such as CPU utilization, disk reads and writes, and network traffic.

Auto Scaling allows you to automatically scale your Amazon EC2 capacity up or down according to conditions you define.

Charged per instance-hour Elastic Load Balancing

Elastic Load Balancing automatically distributes incoming application traffic across multiple Amazon EC2 instances.

Charged per hour and GB of data processed

Page 20: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

20

Google App Engine

Page 21: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

21

Google App Engine Run web applications on Google’s

infrastructure Programming language support: python, java Pricing:

Quota Fixed quota (for free)

Disable billing Enable billing

Billable quota Budget

http://code.google.com/intl/en/appengine/docs/whatisgoogleappengine.html

Page 22: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

22

Requests

Page 23: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

23

Datastore

Page 24: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

24

URL Fetch

Page 25: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

25

Mail

Page 26: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

26

Image Manipulation

Page 27: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

27

Memcache

Page 28: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

28

Billable Quota Unit Cost

http://code.google.com/intl/en/appengine/docs/billing.html

Page 29: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

29

Microsoft Windows Azure

Page 30: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

30

Microsoft Windows Azure Windows Azure platform

Provides a scalable environment with compute, storage, hosting, and management capabilities.

SQL Azure A Relational Database for the

Cloud(Windows Azure platform).

Page 31: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

31

Microsoft Windows Azure During Community Technology Preview (CTP),

services included in Windows Azure will be available without charge Total compute usage: 2000 VM hours/month Cloud storage capacity: 50GB Total storage data transfers: 20GB/day

Once launched for commercial use, Windows Azure would be priced and licensed Jan 1, 2010 First month without charge

Page 32: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

32

Pricing unit Compute Instances:

(Instance Size, CPU, Memory, Storage, I/O Performance )Small --------1.6 GHz ,1.75 GB, 225 GB, ModerateMedium --2 x 1.6 GHz , 3.5 GB, 490 GB, HighLarge----- 4 x 1.6 GHz, 7 GB, 1,000 GB, HighExtra large-8 x 1.6 GHz, 14 GB, 2,040 GB, High

Instance hour transformation: Instance Size Elapsed Hour Small Instance

Hours Small 1 hour 1 hourMedium 1 hour 2 hours Large 1 hour 4 hours Extra large 1 hour 8 hours

Page 33: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

33

Pricing Consumption:

Compute = $0.12 / small instance hour Storage = $0.15 / GB stored / month Storage transactions = $0.01 / 10K Data transfers = $0.10 in / $0.15 out / GB - ($0.30

in / $0.45 out / GB in Asia) Reserved(Development Accelerator Core):

750 hours (small compute instance) 10 GBs of storage 1,000,000 storage transactions 7 GB in / 14 GB out(2.5 GB in / 5 GB out in Asia) For 6 month = $59.95 (42% off from consumption)

Page 34: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

34

Pricing Web Edition: Up to 1 GB relational database =

$9.99 / month Business Edition: Up to 10 GB relational

database = $99.99 / month Data transfers = $0.10 in / $0.15 out / GB -

($0.30 in / $0.45 out / GB in Asia)

Page 35: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

35

Possible Strategies Cost-based pricing

Flat pricing Tiered-pricing Performance-based pricing User-based pricing Usage-based pricing

Page 36: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

36

Possible StrategiesAmazon EC2

Google App Engine

Windows Azure

Low-price leader O O OExperience-curve pricing ?Bundling O OPrice signalingReference pricing ?Image/prestige pricing O O OCost-plus pricingComplementary pricingPremium pricingRandom discounting ?Periodic discounting ?Second-market discounting *

Page 37: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

37

Possible Strategies Other effects

Similar prices (competing situation?)

Page 38: Rajkumar Buyya (1,2),  Chee  Shin  Yeo (1), and  Srikumar Venugopa (l)

38

Conclusion&Comments In this paper, we have proposed architecture

for market-oriented allocation of resources within Clouds.

We have discussed some representative platforms for Cloud computing covering the state-of-the-art.

Comments: This paper has a simple but clear architecture

that we can use. (need add something detail) Some of the information of the cloud platform are

out of date, but the comparison is good.