Cloud Economics in Training and Simulation

CLOUD ECONOMICS IN TRAINING AND SIMULATION

Nane Kratzke

1 Prof. Dr. rer. nat. Nane Kratzke Computer Science and Business Information Systems

The next 25 minutes are about ...

•  What is cloud computing?

•  (Economical) characteristics of cloud computing

•  Postulated use cases for cloud computing

•  Some data from real world

•  Decision making is not always obvious => How to decide?

•  Findings

Prof. Dr. rer. nat. Nane Kratzke Computer Science and Business Information Systems 2

What is a cloud computing (definition)


„Cloud computing is a model for enabling ubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management effort or service provider interaction.“

National Institute of Standards and Technology, NIST: „The NIST definition of cloud computing“; Peter Mell, Timothy Grance, 2011

http://csrc.nist.gov/publications/nistpubs/800-145/SP800-145.pdf

Essential Characteristics of Clouds

Prof. Dr. rer. nat. Nane Kratzke Praktische Informatik und betriebliche Informationssysteme 4

•  No human interaction necessary

•  Programmable data center

On-demand self-service

•  Remote access via thin or fat client platforms

•  No physical access

Network access

•  Resources are pooled to serve multiple consumers

•  Little control or knowledge over exact location

Resource pooling

•  Rapid provisioning •  Autoscaling •  Ressources are

virtually unlimited

Rapid elasticity

•  Pay-per-use business model

•  Resource usage can be monitored, controlled, and reported

Measured Service

Programmable

Data Center

Business Characteristics

Pay as you go Fixed costs become variable

Cost are associative •  100 servers for one hour •  1 server for 100 hours •  Same price

Business gains flexibility •  no long-term financial

commitment to resources


Economical Cloud Usage Patterns have to do with peak loads


„In other words, even if cloud services cost, say, twice as much, a pure cloud solution makes sense for those demand curves where the peak-to-average ratio is two-to-one or higher.“

Weinman, Mathematical Proof of the Inevitability of Cloud Computing, 2011

http://www.joeweinman.com/Resources/Joe_Weinman_Inevitability_Of_Cloud.pdf

Peak loads are cloud economics best friends

Postulated use cases

•  Research shows that cost advantages of cloud computing are deeply use case specific

•  Be aware of comparing non comparable use cases

•  This contribution presents some data of educational use cases (similar usage characteristics of simulational use cases)

hosting websites support software development cycles

short-term system demonstrations

data storage, disaster recovery

and business continuity

Training and education

media processing and rendering

overflow processing or large-scale scientific data

processing simulation


These use cases (among others) are postulated to be cloud compatible:

Analyzed use case

•  Web technology lecture/practical course for computer science students (bachelor) in summer 2011 and summer/winter 2012.

•  Projects: Development of web information systems (Drupal based)

•  All groups were assigned cloud service accounts provided by Amazon Web Services (AWS).

•  Analysis of billing as well as usage data provided by AWS.


Cost analysis


Total costs: 846.99 $

Total students: 49

Cost per student: 17.28 $ CW 13 CW 14 – CW 17 CW 18 – CW 21 CW 22 – CW 25

(A)Costs per Month (aligned to Weeks)

Calendar Weeks (CW)

Cos

ts in

US

D

0100

200

300

400

500

instancehour (62%)

datastorage (34%)

adressing (3%)datatransfer (0%)

(B)Main Cost Drivers

B 1 (5%)

B 2 (7%)

B 3 (7%)B 4 (4%)B 5 (6%)

A 1 (10%)

A 2 (31%)

A 3 (10%)

A 4 (19%)

(C)Costresponsibilty of Groups

(D) Histogram of Costs per Group

Cost Ranges in USD

# G

roup

s

0 50 100 150 200 250 300

01

23

4

Main identified cost drivers:

(1)  Server uptime (2/3)

(2)  Data storage (1/3)

Server Usage Analysis


13 14 15 16 17 18 19 20 21 22 23 24 25

Average Box UsageMaximum Box Usage in an hour

(A)Maximum and Average Box Usage

Calendar Week

Use

d S

erve

r Box

es

010

2030

4050

13 14 15 16 17 18 19 20 21 22 23 24 25

(B)Accumulated Processing Hours per Week

Calendar Week

Pro

cess

ing

Hou

rs

0500

1000

1500

2000

14 16 18 20 22 24

0.0

0.2

0.4

0.6

0.8

1.0

(C)Average Box to Maximum Box Ratio

according to Weinman

Calendar Week

Avg

to M

ax B

ox U

sage

Rat

io

Training

Project 24x7 Migration

Average to Peak Ratio per Week


13 14 15 16 17 18 19 20 21 22 23 24 25



Calendar Week

Use

d S

erve

r Box

es

010

2030

4050

13 14 15 16 17 18 19 20 21 22 23 24 25


Calendar WeekP

roce

ssin

g H

ours

0500

1000

1500

2000

14 16 18 20 22 24

0.0

0.2

0.4

0.6

0.8

1.0



Calendar Week

Avg

to M

ax B

ox U

sage

Rat

io

Cloud computing is economical reasonable

Cloud computing might be reasonable

Cloud computing is economical not reasonable

Economical Decision Analysis A four step process to decide for or against cloud based solutions

Determine your atp ratio

Determine your dedicated costs

Determine your maximal cloud costs

Determine appropriate cloud ressources


Max instances: 49

Processing hours: 7612

Average: 7612 / (26 * 7 * 24) = 1.74

Overall atp ratio: 1.74 / 49 = 0.035

13 14 15 16 17 18 19 20 21 22 23 24 25



Calendar Week

Use

d S

erve

r Box

es

010

2030

4050

13 14 15 16 17 18 19 20 21 22 23 24 25


Calendar Week

Pro

cess

ing

Hou

rs

0500

1000

1500

2000

14 16 18 20 22 24

0.0

0.2

0.4

0.6

0.8

1.0



Calendar Week

Avg

to M

ax B

ox U

sage

Rat

io







Example Server: 500 US Dollar Amortization: 3 years

€

d3years(500$) =500$

3• 365 • 24h= 0.019$ h

„In other words, even if cloud services cost, say, twice as much, a pure cloud solution makes sense for those demand curves where the peak-to-average ratio is two-to-one or higher.“ Weinman, Mathematical Proof of the Inevitability of Cloud Computing, 2011







€

cMax =0.019$ h0.035

= 0.54 $h

According to Weinman the peak-to-average ratio should be greater than the ratio between the variable costs c and your (assumed) dedicated costs d:







€

cMax =0.019$ h0.035

≈ 0.54 $h

Exam

ple:

Am

azon

Web

Ser

vice

s EC

2-Pr

icin

gs fo

r EU

regi

on, 1

9th

Mar

ch, 2

012

Economical Decision Analysis A four step process to decide for or against cloud based virtual labs


A cloud based solution provides a more than 25 times cost advantage.

The measured ATP ratio of 0.035 means in fact a 1/0.035 == 28.57 times cost advantage.

This means for the presented use case:

Compared to necessary investment efforts for a classical dedicated system implementation.

Why this big cost advantage?


13 14 15 16 17 18 19 20 21 22 23 24 25



Calendar Week

Use

d S

erve

r Box

es

010

2030

4050

13 14 15 16 17 18 19 20 21 22 23 24 25


Calendar Week

Pro

cess

ing

Hou

rs

0500

1000

1500

2000

14 16 18 20 22 24

0.0

0.2

0.4

0.6

0.8

1.0



Calendar Week

Avg

to M

ax B

ox U

sage

Rat

io

How to dimensionize the data center? peak load

What is the need?

average load

And the delta?

Measures the overdimension of a data center

In other words ...


13 14 15 16 17 18 19 20 21 22 23 24 25



Calendar Week

Use

d S

erve

r Box

es

010

2030

4050

13 14 15 16 17 18 19 20 21 22 23 24 25


Calendar Week

Pro

cess

ing

Hou

rs

0500

1000

1500

2000

14 16 18 20 22 24

0.0

0.2

0.4

0.6

0.8

1.0



Calendar Week

Avg

to M

ax B

ox U

sage

Rat

io

You have to finance a really big house ...

... knowing that you

will inhabit only some rooms of it.

Findings

•  Cloud computing loves peak load scenarios (be happy) •  25 times cost advantage (analyzed use case)

•  Cloud generated costs are use case specific (be carefull) •  Decision making must not be obvious •  Four step decision making model (to determine your ATP ratio)

•  Main cost drivers are (try to minimize) •  Server uptime •  Data storage (server volumes) •  Data transfer (in communication intensive use cases)

•  Uneconomical use cases (try to avoid) •  24x7 and •  constant loads

•  So if you have to deal with peak load scenerios it is likely that cloud based solutions might be an economical option ...


Thank you for listening


Prof. Dr. Nane Kratzke Computer Science and Business Information Systems Lübeck University of Applied Sciences Mönkhofer Weg 239 23562 Lübeck Germany

Mail: [email protected]

Slideshare: i21aneka

XING: Nane_Kratzke

LinkedIn: nanekratzke

WEB: http://praktische-informatik.fh-luebeck.de

Find this presentation here: http://www.slideshare.net/i21aneka/itis-ws-2013

Twitter: @nanekratzke