ARE CLOUD ENABLED VIRTUAL LABS ECONOMICAL?

A Case Study Analyzing Cloud based Virtual Labs for Educational Purposes

Nane Kratzke

ubeck University of Applied Sciences, M

onkhofer Weg 239, 23562 L

ubeck, Germany

Keywords:

Cloud, Computing, Virtual Labs, Higher Education, IaaS, Decision, Making, Cost, Estimation.

Abstract:

Cost efﬁciency is an often mentioned strength of cloud computing (Talukader et al., 2010). In times of de-

creasing educational budgets virtual labs provided by cloud computing might be an interesting alternative

for higher education organizations or IT training facilities. This contribution analyzes the cost advantage of

virtual educational labs provided via cloud computing means and compare these costs to costs of classical ed-

ucational labs provided in a dedicated manner. This contribution develops a four step decision making model

which might be interesting for colleges, universities or other IT training facilities planning to implement cloud

based training facilities. Furthermore this contribution provides interesting ﬁndings when cloud computing

has economical advantages in education and when not. The developed four step decision making model of

general IaaS

∗

applicability can be used to ﬁnd out whether an IaaS cloud based virtual IT lab approach is more

cost efﬁcient than a dedicated approach.

1 INTRODUCTION

Cloud computing is one of the latest developments

within the business information systems domain and

describes a delivery model for IT services based on

the Internet, and it typically involves the provision of

dynamically scalable and often virtualized resources.

Nowadays cloud computing is used in e-learning sce-

narios as well because it ﬁts very well to e-learning

requirements. When e-learning has a distinctive re-

mote aspect why delivering necessary educational re-

sources like labs has to be delivered still in a classical

and dedicated manner?

Accompanying the increasing relevance of cloud

computing in research literature and media there

arise manifold publications covering the application

of cloud computing to e-learning. E.g. (Cayirci

et al., 2009) present a training and education cloud.

But only some publications covering aspects how to

use cloud computing to provide virtual labs. (Oberg

et al., 2011) presents a ”virtual lab” system architec-

ture for academic high-performance computing but

without educational purposes. And a lot of other au-

thors reﬂect about to use cloud computing as tech-

nical infrastructure for providing e-learning systems

∗

Infrastructure as a Service. This contribution follows

the IaaS, PaaS and SaaS deﬁnition of NIST (Mell and

Grance, 2011).

(Dong et al., 2009), (Ko and Young, 2011), (Masud

and Huang, 2011) digital campus systems (Li and Li,

2011) or personalized learning environment systems

(Liang and Yang, 2011) but this has nothing to do with

virtual laboratories in the understanding of this con-

tribution. Only (Thi

ebaut et al., 2011) show a very

interesting example of processing wikipedia dumps

by applying cloud computing technologies and com-

pare them to dedicated cluster solutions in practical

college courses. This could be named a ”virtual lab”

but it is not done by the authors due to a different

focus of their paper. So this contribution is about a

case study how to use cloud computing in computer

science study programs providing virtual IT labs for

practical courses in the higher education domain or

for IT training facilities. According to (Barr, 2010)

we deﬁne:

A Virtual Educational IT Lab is a collec-

tion of compute, storage and net-working re-

sources provided by an educational organiza-

tion for educational or research purposes. A

virtual lab can be provided to a single or a

small group of students to support these stu-

dent(s) in solving practical problems by pro-

viding a necessary IT infrastructure. Provided

resources are available for short-term use

and are billed by actual resource consump-

tion generated by educational or research ac-

353

Kratzke N..

ARE CLOUD ENABLED VIRTUAL LABS ECONOMICAL? - A Case Study Analyzing Cloud based Virtual Labs for Educational Purposes.

DOI: 10.5220/0003888503530362

In Proceedings of the 4th International Conference on Computer Supported Education (CSEDU-2012), pages 353-362

ISBN: 978-989-8565-07-5

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

tivities. Typically all provided resources are

rented by the educational organization from a

cloud service provider.

Beside use cases like hosting websites, support soft-

ware development cycles, short-term system demon-

strations, data storage, disaster recovery and business

continuity, media processing and rendering, overﬂow

processing or large-scale scientiﬁc data processing

(Barr, 2010) mentions training use cases as very cloud

compatible and economical use cases. This paper

does not denial this postulation in general but ad-

vocates a more critical view like (Weinmann, 2011)

or (Mazhelis et al., 2011) do. Ongoing research

(Kratzke, 2011a), (Kratzke, 2011b), (Kratzke, 2012a)

shows that cost advantages of cloud computing are

deeply use case speciﬁc and you should be aware

of comparing non comparable use cases. And nev-

ertheless we have to consider some shortcomings of

cloud computing which are mentioned in literature

and should be considered in planning phases of vir-

tual labs.

Shortcomings of Cloud Computing. (Kratzke,

2012a) showed by analysing popular IT management

frameworks like COBIT, TOGAF or ITIL that – from

an IT management point of view – cloud computing

provides more beneﬁts than disproﬁts. Nevertheless

Kratzke showed also the presence of so called sub-

stantial show stoppers for cloud based approaches.

Especially security and compliance management as-

pects may come along with substantial ”showstop-

pers”. But these aspects are of minor relevance re-

garding cloud based virtual labs.

According to (Truong and Dustdar, 2010) or

(Kratzke, 2012a) cloud computing is also character-

ized by an ex ante cost intransparency. This very im-

portant weakness (from an IT management and de-

cision making point of view) is even little reﬂected

in literature so far. To answer the question whether

a cloud-based approach is more cost efﬁcient than a

dedicated approach it has to be answered the ques-

tion what costs will be generated per month before a

cloud based approach enters operation (Walker et al.,

2010). This is very difﬁcult to answer ex ante because

it is inﬂuenced by a bunch of interdependent parame-

ters. But for profound decision making for or against

cloud based virtual labs exactly this question has to

be answered before a virtual lab is established in a

dedicated or cloud based manner.

Outline. Therefore this contribution has the follow-

ing outline. The research methodology is described in

section 2. An economical decision making model for

or against cloud based virtual labs is described brieﬂy

in section 3. A corresponding case study of a com-

puter science study program is analyzed in section 4.

Section 5 shows derived ﬁndings from the analyzed

case study which are useful to plan virtual lab sup-

ported practical courses in higher education. Section

5 shows furthermore how to transfer the presented use

case to other computer science related lectures and

related practical courses and operationalizes simple

rules for controlling costs pragmatically. The contri-

bution ends with a summary and outlook in section

2 RESEARCH METHODOLOGY

Especially the work of (Weinmann, 2011) is very in-

teresting from this decision making point of view be-

cause it shows how to decide for or against a cloud

based approach very pragmatically. This contribution

is about using the work of Weinman to build up a sim-

ple decision making model for or against cloud based

approaches. This decision making model has been ap-

plied in a presented case study covering the higher ed-

ucation domain.

For practical courses in higher university or col-

lege education we want to ﬁnd out whether it is more

economical to provide classical dedicated educational

labs or to use IaaS

providing virtual labs for student

practical courses.

The analyzed case study was a web technology

lecture for computer science students being held at

the L

ubeck University of Applied Sciences in 2011.

During the practical courses of this lecture students

formed groups of 5 or 6 persons in order to build up

a website for a scientiﬁc conference on robotic sail-

ing

(project 1) or establish a google map based auto-

matic sailbot tracking service (project 2) for the same

conference. All groups were assigned cloud service

accounts provided by Amazon Web Services. The

groups were asked to use these accounts in order to

fulﬁll their projects in a complete cloud based man-

ner. The resource consumption of all groups were

measured by analyzing billing as well as usage data

provided by the cloud service provider Amazon Web

Services (AWS).

Both projects had three main phases (see table

1). The training phase (calendar week 13 to 15) was

about to get all students into touch with the cloud ser-

vice provider tool suite. In the project phase (calendar

week 16 to 23) all groups had to develop a cloud based

solution for project 1 or project 2. All groups had to

This contribution follows the IaaS, PaaS and SaaS def-

inition of NIST (Mell and Grance, 2011).

World Robotic Sailing Conference 2011, see

http://www.wrsc2011.org.

CSEDU2012-4thInternationalConferenceonComputerSupportedEducation

354

proof that their solutions were adequate for a 24x7 op-

eration within a 24x7 phase (calendar week 21 to 24).

So project and live phase overlapped within a three

week period. Calendar week 24 was used for the sys-

tem presentations. The practical course of the lecture

ﬁnished with a migration phase. The best solutions

were used as website (project 1) or google maps based

automatic sailbot tracking service (project 2) for the

conference and were migrated to the destination envi-

ronment (which also was a cloud based environment)

in calendar week 25.

24x7

Training Project P M

CW 13

CW 14

CW 15

CW 16

CW 18

CW 19

CW 20

CW 21

CW 22

CW 23

CW 24

CW 25

P = Presentation M = Migration

CW = Calendar Week

Figure 1: Project phases.

3 DECISION MAKING MODEL

(Weinmann, 2011) is stressing the following interest-

ing fact which is a crucial input for pragmatic decision

making for or against cloud based system implemen-

tations especially on a IaaS level of cloud computing:

A pay-per-use solution obviously makes sense

if the unit cost of cloud services is lower than

dedicated, owned capacity. [...]

[...] a pure cloud solution also makes sense

even if its unit cost is higher, as long as

the peak-to-average ratio of the demand curve

is higher than the cost differential between

on-demand and dedicated capacity. 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. (Wein-

mann, 2011)

According to Weinman the peak-to-average ratio is

the essential indicator whether a cloud based ap-

proach is economical reasonable or not. So it is

not necessary to estimate costs per month of a cloud

based solution exactly. It is sufﬁcient to proof that

cloud based costs are smaller then a dedicated sys-

tem implementation. And this can be ﬁgured out by

analysing the peak usage as well as the average usage

of a system.

Due to page limitations we will use in this con-

tribution the average to peak ratio at p in a simpliﬁed

form and refer to (Kratzke, 2012b) where the decision

making model is presented more detailly.

at p :=

averageusage

maximumusage

pta

(1)

According to (Weinmann, 2011) we have to compare

the costs of a classical dedicated approach with the

costs of a cloud based approach. On the IaaS level

it is common to be billed per service usage with a

granularity of the atomic timeframe t level. Which

would be in case of Amazon Web Service that you

are billed for a server instance per complete (or par-

tial) hour usage. Let us name our dedicated costs per

atomic timeframe d and our cloud costs per atomic

timeframe c. Cloud costs c can be easily ﬁgured out

because they are provided as pricing by their cloud

service providers

. Dedicated costs per atomic time-

frame d are a little more complex to calculate. Never-

theless for estimations we can assume, that they can

be deﬁned via their amortizations. If a dedicated in-

stance can be procured for p value units

their dedi-

cated costs per atomic timeframe (ATF) can be calcu-

lated as follows

AT F

(p) =

AT F

3year

(p) =

3 · 365 · 24h

5year

(p) =

5 · 365 · 24h

(2)

Typical amortization timeframes are a 3 year or a 5

year hardware regeneration interval (see equation 2).

So a 500 $ server would have the following dedicated

costs per atomic timeframe of 1h over a amortization

interval of 3 years.

3year

(500$) =

500$

3 · 365 · 24h

≈ 0.019

(3)

According to (Weinmann, 2011) the peak-to-average

ratio pta should be greater than the relation between

the variable costs per atomic timeframe c and the ded-

icated costs per atomic timeframe d which can be ex-

pressed in the following form:

pta >

⇔ pta · d > c ⇔ c <

at p

(4)

E.g. Amazon Web Services publishes

these cloud costs per atomic timeframe here:

http://aws.amazon.com/de/ec2/#pricing.

E.g. US Dollars $ or Euro e.

Be aware – this assumption do not account typical ad-

ditional operational efforts like administration, cooling or

electricity. Nevertheless we do not want to calculate exact

costs we only want to know whether a cloud based approach

is more economical than a dedicated one. In this case it is

OK to give the dedicated side an advantage by not account-

ing aspects like administration, cooling, electricity, etc. al-

though these costs are included in the variable costs on the

cloud service provider side.

ARECLOUDENABLEDVIRTUALLABSECONOMICAL?-ACaseStudyAnalyzingCloudbasedVirtualLabsfor

EducationalPurposes

355

In other words equation 4 provides a clear decision

criteria to decide for or against a cloud based ap-

proach. By knowing your average to peak ratio

at p, your hardware procurement costs per instance

p as well as your hardware amortization timeframes

(which is typically 3 or 5 years) it is possible to calcu-

late a maximum of cloud costs per atomic timeframe

MAX

until a cloud based approach is economical (see

equation 5). Whenever a cloud service provider can

realize instance pricings below c

MAX

we decide for a

cloud based approach

– in all other cases we should

realize the system in a dedicated approach

MAX

at p

(5)

4 ANALYSED CASE STUDY

All data used in this section has been collected by the

supporting cloud service provider AWS. The billing

data as well as usage data are standard informations

provided by AWS for each client. They are provided

by AWS in CSV or XML format and can be therefore

easily downloaded and processed by anyone. We used

billing as well as EC2 usage data provided by AWS.

No special preprocessing of the data were performed

by AWS for this special use case analysis. So anyone

can run the same analysis by using a standard toolset

4.1 Cost Analysis

Table 1 shows all costs per group within the analysed

timeframe. In total the L

ubeck University of Applied

Sciences had to spend 847.01$ in providing (virtually)

unlimited amount of server instances to 49 students

organized in 9 groups within a timeframe of 13 cal-

endar weeks. This sounds impressive but says in fact

nothing about how cost efﬁcient this performed cloud

based approach was. Could we had reached the same

results with a classical dedicated approach?

Let us have a look at the cost analysis provided in ﬁg-

ure 2. Figure 2(A) shows the costs per month. AWS

is only providing billing data on a monthly basis so

we had to align months to the corresponding analysed

calendar weeks. By interpreting these graphs we must

have in mind the general project phases performed by

all student groups (see table 1). So please keep the

following phases in mind:

• Week 13 - 15: initial training phase.

Only from an economical point of view.

Also only from an economical point of view.

We used only open source software. MySQL as a

database and R as a analytical and visualization tool.

CW 13 CW 14 – CW 17 CW 18 – CW 21 CW 22 – CW 25

(A)

Costs per Month (aligned to Weeks)

Calendar Weeks (CW)

Costs in USD

0 100 200 300 400 500

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

# Groups

0 50 100 150 200 250 300

0 1 2 3 4

Figure 2: Cost analysis.

• Week 16 - 23: project/development phase.

• Week 21 - 24: (overlapping) 24x7 phase.

• Week 24: presentation.

• Week 25: migration.

So ﬁgure 2(A) shows that most of the costs were gen-

erated in 24x7 phase (calendar week 21 - 24). So 24x7

CSEDU2012-4thInternationalConferenceonComputerSupportedEducation

356

Table 1: Group overview.

Group Size Project Costs in $

A 1 5 WRSC Website 88.39$

A 2 6 WRSC Website 265.37$

A 3 4 WRSC Website 88.14$

A 4 6 WRSC Website 162.88$

B 1 6 Sailbot Tracking 41.17$

B 2 6 Sailbot Tracking 57.58$

B 3 6 Sailbot Tracking 57.46$

B 4 5 Sailbot Tracking 37.42$

B 5 5 Sailbot Tracking 48.58$

seems to be expensive.

Figure 2(B) shows the main cost drivers. Almost

2/3 of the costs were generated by instancehours –

that means running servers and being billed for per

hour. Almost 1/3 of the costs were generated by data

storage – that means all costs which have to do with

the provision of server hard drives, storing backups or

other data storage services. Other costs like address-

ing (requesting IP addresses, DNS names, etc.) or

even data transfer had no relevant impact to the total

costs. So the main cost driver were instance hour

costs – or box usage as it is sometimes called.

Figure 2(C) and (D) shows that analyzed groups

had quite different cost responsibilities. The expected

value of cost responsibility would be 100%/9 ≈

11.11%.

Nevertheless the most cost efﬁcient group was

only responsible for about 4% and the most cost ”lav-

ish” group was responsible for about 31% of the total

costs. It turned out that the A groups – groups re-

sponsible for setting up the website – consumed more

resources than the sailbot tracking B groups (check

table 1). So cloud generated costs are use case

speciﬁc. Different problems result in different archi-

tectural solutions which results in different cost be-

haviour.

4.2 Usage Analysis

As we have found out in our analyzed use case – main

cost driver was server/box usage (or instance hours

– see ﬁgure 2(B)). Thats why a detailed box usage

analysis has been performed and is shown in ﬁgure 3.

Figure 3(A) shows the maximum and average box

usage per calendar week measured within the ana-

lyzed timeframe (calendar week 13 - 25). Figure 3(B)

shows the total sum of all processing hours generated

by all used server boxes/instances per calendar week.

Figure 3(C) shows the average to peak load ratio (ac-

cording to equation 1) per calendar week.

Within the initial training phase (calendar week

13 - 21) the usage characteristic shows an extremely

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

Average Box Usage

Maximum Box Usage in an hour

(A)

Maximum and Average Box Usage

Calendar Week

Used Server Boxes

0 10 20 30 40 50

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

(B)

Accumulated Processing Hours per Week

Calendar Week

Processing Hours

0 500 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 Max Box Usage Ratio

Figure 3: Analysed box usage.

high maximum box usage but an astonishing low av-

erage usage. This characterizes an extreme peak load

situation and results in extreme low at p ratios (see

equation 1 and ﬁgure 3(C)). According to equation

4 or the deﬁnition of c

MAX

(equation 5) this shows

a very ideal cloud computing (peaky) situation. So

training phases seems to be very economical in-

teresting cloud computing use cases which is also

stated by (Barr, 2010)

Within the project phase (calendar week 13 - 15)

the usage characteristic shows dramatically reduced

maximum as well as average box usages. Neverthe-

less the at p ratios (see equation 1 and ﬁgure 3(C))

(Barr, 2010) is chief evangelist of Amazon Web Ser-

vices – so this source can not be rated objective. Neverthe-

less our case study hardens his point.

ARECLOUDENABLEDVIRTUALLABSECONOMICAL?-ACaseStudyAnalyzingCloudbasedVirtualLabsfor

EducationalPurposes

357

stay in a very comfortable situation for cloud com-

puting approaches. So we still have a peaky usage

situation but on a dramatical lower level. So also de-

velopment phases seems to be very economical in-

teresting cloud computing use cases which is also

stated by (Barr, 2010)

The 24x7 phase (calendar week 21 - 24) shows

raised maximum as well as average box usages (see

ﬁgure 3(A)). Also the at p ratios are raising within

this phase – nevertheless the peaky usage character-

istic remains but less distinctive. The 24x7 phase can

be clearly seen in the accumulated processing hours

(see ﬁgure 3(B)) which shows a clear peak in calendar

weeks 22 - 24. We have stated already in section 4.1

that 24x7 seems to be expensive and we can harden

this conclusion by our usage analysis. This might sur-

prise some readers

The migration phase (calendar week 25) was

characterized by transferring the best solutions for the

website and sailbot tracking service into the opera-

tional environment. Within this phase the systems run

in a steady load scenario as can be seen in ﬁgure 3(A)

(almost the same average box as well as maximum

usage) and in ﬁgure 3(C) (an at p ratio near 1.0). Ac-

cording to equation 5 this shows an extreme uncom-

fortable situation for cloud computing situations – so

steady loads seem to be no economical interesting

cloud computing use cases.

4.3 Economical Decision Analysis

As we have seen in sections 4.1 and 4.2 we can iden-

tify different phases which are more cloud compatible

than others from an ecomomical point of view. Train-

ing and development phases show very low at p ratios

(see ﬁgure 3(C)) and therefore indicate peaky usage

characteristics of resources which advantages cloud

computing realizations

(check equation 5). Other

phases with less peaky usage characteristics (like our

24x7 or migration phase) disadvantage cloud comput-

ing realizations.

So we have identiﬁed pro and cons for a cloud

Remember (Barr, 2010) is chief evangelist of Amazon

Web Services. He is not objective. Nevertheless our case

study hardens this point too.

It is often stated that cloud computing is good for host-

ing websites (a 24x7 job). This is true but only if not pro-

vided on IaaS level or when provided on IaaS level with a

clear peaky usage characteristic. Steady loads are seldom

economical IaaS use cases. So you should be very clear

about your use case, the related (un)peaky usage character-

istic and the IaaS, PaaS or SaaS provision level. No one

states that economical decision making for or against cloud

computing is obvious.

From an economical point of view.

based realization of our educational labs. But how

to decide? Now we are going to apply our decision

model presented in section 3.

Step 1: Determine the at p Ratio.

First of all we have to calculate our overall average

to peak load ratio at p. Our analysis timeframe cov-

ered the calendar weeks 13 - 25. So an intuitive time-

frame for average building would be 13 weeks – but

this implicates a continual usage of an educational lab

over a complete year which is very uncommon. In

the university or college business educational labs are

typically used one time per semester. In most cases

an educational lab can be used only one time a year

(per semester – that means average building over 26

weeks) or even only one time per year (every second

semester – that means average building over 52 weeks

which is a year). In our analyzed timeframe 7612

hours of instance usage were generated (the sum of

all bars of ﬁgure 3(B)). So equation 7 shows the av-

erage amount of servers which would be necessary to

provide 7612 processing hours within a 26 or 52 week

timeframe.

avg

26w

7612h

26 · 7 · 24h

≈ 1.74

avg

52w

7612h

52 · 7 · 24h

≈ 0.87

(6)

Now we can build up our average to peak ratio. Our

maximum server usage within an atomic timeframe of

1 hour was 49 servers (please check ﬁgure 3(A)). So

we get the following at p ratios for a 26 or 52 week

timeframe.

at p

26w

1.74

≈ 0.035

at p

52w

0.87

≈ 0.018

(7)

Step 2: Determine your Dedicated Costs.

First of all we have to ﬁnd out how much would cost

us a dedicated server. At the L

ubeck University of

Applied Sciences our procurement ofﬁce could pur-

chase the smallest possible server version

for about

3055$ (2167e

). Equation 2 told us to calculate our

dedicated costs per atomic timeframe (1h) in the fol-

lowing way for a 5 year amortization interval:

5year

(3055$) =

3055$

5 · 365 · 24h

≈ 0.0697

(8)

Dell PowerEdge Server R610, 2.13 GHz Intel Xeon

processor, 8GB memory, 140 GB hard drive (valid on

28th October 2011) with approximately 2 ECU – so the

AWS instance type pendant would be something between

a Standard Small (1 ECU) or Large (4 ECU) instance type.

Check out detailed instance type informations of AWS here:

http://aws.amazon.com/de/ec2/instance-types/.

Exchange rate 1.41$ = 1.00e from 28th October 2011.

CSEDU2012-4thInternationalConferenceonComputerSupportedEducation

358

Step 3: Determine your Maximal Economical

Cloud Costs.

Equation 5 told us to calculate our c

MAX

costs in the

following way:

26w

MAX

5year

(3055$)

at p

26w

0.0697

0.035

≈ 1.99

52w

MAX

5year

(3055$)

at p

52w

0.0697

0.018

≈ 3.87

(9)

In other words. In our analyzed case study a cloud

service provider (regarding a 5 year amortization time

frame for servers) could be

• 28.57 times more expensive in case of a one time

per semester usable educational lab or even

• 55.56 times more expensive in case of a only one

time per year usable educational lab

then own dedicated costs – a cloud based educational

lab would be still more economical then a dedicated

one. Ok these ﬁgures are really impressive but do we

ﬁnd appropriate resources within our maximal costs?

We have to ﬁgure this out in our last step 4 to make

a profound decision for or against a cloud based ap-

proach for our virtual lab.

Step 4: Determine Appropriate Cloud Resources.

Now we know our maximal cloud costs and have to

look if our cloud service provider can deliver appro-

priate and comparable resources. In our case this is

Amazon Web Services, but it could be any other IaaS

cloud service provider as well. We do this exemplar-

ily for a 26 week timeframe

. But it works abso-

lutely the same for all other timeframes or IaaS cloud

service providers as well.

Table 2 shows all instance types of AWS and their al-

located costs. Remember section 4.3 told us, that all

instance types cheaper than 1.99 $/h result into cloud

based solutions which are more economical than ded-

icated approaches.

As you can see in table 2 all provided instance types

(except one

) of AWS in the EU Region are econom-

ical in the sense of section 3 and equations 4 and 5.

The most similar instance types listed in table 2 are

marked as ’o’. (’-’ stands for worser, ’+’ for better

and ’++’ for much better than a dedicated reference

system

So in our analysed use case the Medium (High CPU)

or may be even the Small (Standard) AWS instance

types (see table 2) are the most comparable systems

Because in our special case we can use our educational

lab every semester – so two times a year.

4x XL (High Memory) instance type is not economical

reasonable but this instance type is not comparable to our

reference system because it is much more powerful.

In our case the Dell PowerEdge Server R610

Table 2: AWS Instance types and pricings, according to

AWS pricing informations on 28th Oct. 2011, EU (Ireland)

region, on-demand instances, Linux/UNIX operating sys-

tem.

AWS Instance Type ECU Price/h

Economical

Comparable

Micro < 1 0.025$ yes -

Small (Standard) 1 0.095$ yes o

Large (Standard) 4 0.38$ yes o

XL (Standard) 8 0.76$ yes +

XL (High Memory) 6.5 0.57$ yes +

2x XL (High Memory) 13 1.14$ yes ++

4x XL (High Memory) 26 2.28$ no ++

Medium (High CPU) 5 0.19$ yes o

XL (High CPU) 20 0.76$ yes ++

to our dedicated reference system (Dell PowerEdge

Server R610). Both provide variable cloud costs

clearly below our maximum costs of 1.99$/h (see

equation 9). So in our analysed use case a virtual IT

lab for education is much more economical than

a dedicated approach. Thats why we decided to

implement cloud based IT labs for our practical

courses instead of classical dedicated educational

labs.

5 CONCLUSIONS

In section 5.1 we derive some general ﬁndings from

our use case analysis which might be useful to ﬁnd in-

teresting practical courses in higher education of com-

puter science which are likely to show similar cost

characteristics like the analyzed use case. Neverthe-

less our analysis shows also in section 5.2 that the

here presented approach is transferable to other lec-

tures and related practical courses as well. We op-

erationalize some general advices in section 5.3 for

setting up cloud based virtual labs and develop some

rules for virtual lab users to control costs pragmati-

cally.

5.1 General Findings

Cloud Computing Economics Love Peak Load

Scenarios. As equation 4 showed cloud computing

becomes more and more economical as the average to

peak ratio nears 0 - this indicates that peaky as well

as seldom usage of educational labs argue for cloud

based virtual lab approaches, extreme continual us-

age of labs argue against cloud based approaches from

an economical point of view. Nevertheless we always

ARECLOUDENABLEDVIRTUALLABSECONOMICAL?-ACaseStudyAnalyzingCloudbasedVirtualLabsfor

EducationalPurposes

359

have to analyze the individual characteristic of a prac-

tical course by measuring its average to peak ratio in

real world.

Cloud Generated Costs are use Case Speciﬁc. Dif-

ferent problems result in different realization architec-

tures generating different costs as well as usage char-

acteristics. So be aware of comparing non comparable

use cases! You should run for each practical course

the here mentioned cost analysis and decide for or

against a cloud computing based virtual lab approach

after you have ﬁgured out your practical course spe-

ciﬁc average to peak ratio.

24x7 Seems to be Expensive and of no Economical

Interest. for cloud computing if not associated with

a peaky usage characteristic. So try to avoid 24x7

tasks in practical courses whereever possible. This

will save a lot of money.

One of the Main Cost Driver is Box Usage per

Complete (or Partial) Hour. Second Relevant Cost

Driver is Data Storage. Data transfer seems to have

a minor impact to costs. This might be only valid

for our analyzed use case and should be handled with

care. Please check out other publications like (Mazhe-

lis et al., 2011) for communication intensive use cases

if you plan to use communication/data transfer inten-

sive use cases in your practical courses.

Constant Loads Over Time Seem to be no Eco-

nomical Interesting use Cases for Cloud Comput-

ing. If you have to deal with this kind of use cases

in practical courses (which is typically very unlikely)

think about classical dedicated approaches. Cloud

computing approaches might be only feasible in these

cases due to their convenience but not necessary eco-

nomical reasonable.

5.2 Transferability of the Approach

Because of the identiﬁed cost advantages in the ana-

lyzed use case there was one more question arising.

Is it possible to transfer made experiences to other

lectures and practical courses as well? Therefore all

module descriptions of computer science (near) study

programs of the L

ubeck University of Applied Sci-

ences were analyzed. The following lectures and re-

lated practical courses showed potential being sup-

ported by virtual labs

• Industrial Networks and Databases.

• Informationtechnology.

• Webtechnologies.

• Database Management (and Engineering).

Most of the mentioned lectures and related practical

courses are in german. So lecture titles are translated.

• Integrated Informationsystems.

• Distributed Systems.

• Operating Systems (if Linux/UNIX based).

It turned out that in each analyzed computer science

related study program between 8.33% up to 23.10%

of all practical courses could be supported by virtual

labs according to the module descriptions of the study

programs. Whether these practical courses will show

similar cost advantages is up for ongoing research.

Nevertheless it seems very likely from made experi-

ences. So it should by obvious to the reader that the

presented approach is easily transferable to other lec-

tures and related practical courses.

To identify interesting lectures and practical courses

it is good to know what is making a practical course

virtual lab compatible. Our identiﬁed courses have

the following virtual lab relevant educational require-

ments in common:

• a practical course requires databases,

• a practical course requires distributed processing

instances,

• a practical course requires webtechnologies on

server instances,

• a practical course requires linux/unix based in-

stances,

• a practical course requires parallel processing,

• a practical course requires large-scale data pro-

cessing.

Whenever there exist a practical course with one of

the above mentioned educational requirements a vir-

tual lab might be a reasonable option.

5.3 Advices for Transferring

We learned a lot about providing virtual labs to stu-

dents via cloud computing means. Some of our made

experiences are operationalized as advices for every-

one who is planning similar approaches.

Advice Nr. 1: It is likely that Cloud Computing

will be new for students. So plan an initial training

phase with the cloud tooling of the service provider

which is necessary to solve the problems of the prac-

tical course. We did this with by providing very de-

tailed step by step manual. This seems to be an effec-

tive way. And force your students to be present in the

ﬁrst initial training phase.

Advice Nr. 2: If you are planning to use virtual

clouds in more than one practical course you might

think about a virtual lab training course in one of

the ﬁrst semesters of a study program. This might

avoid double and triple trainings.

CSEDU2012-4thInternationalConferenceonComputerSupportedEducation

360

Advice Nr. 3: Let your students play within their

virtual lab in presence phases, at home, at university,

wherever they want. Cloud computing gives ﬂexibil-

ity. Use it! You are not bounded to presence phases

any more if training phase is passed.

Advice Nr. 4: Watch your costs! We identiﬁed two

main cost drivers. First cost driver is instance oper-

ating hours. Whenever an instance is running money

ﬂows. Make this clear to students by setting up rules.

Rule Nr. 1: Shut down all instances after ﬁnishing

your experiments. Some cloud providers have APIs

to control resources. You may think about developing

garbage collector scripts shutting down every running

instance at midnight? This will discipline even lazy

students. Most cloud service providers provide differ-

ent instance types which can be formalised in another

rule. Rule Nr. 2: Use the smallest instance types for

experiments. It is always possible to upscale.

Second most relevant cost driver is data storage gen-

erated by not running instances. Rule Nr. 3: Use as

few instances as possible. Every instance produces

storage costs. Some cloud providers provide detailed

access control settings. Use them to let no group (ac-

count) instantiate more than a limited amount of in-

stances in parallel.

Advice Nr. 5: And to increase a general cost

awareness it is good to establish another rule. Rule

Nr. 4: If two groups show the same performance. The

one with less costs will get the better grade. So re-

source consumption will not become a key criteria for

grading. But our experiences show that this rule dis-

ciplines a lot.

Table 3 summarizes all relevant rules to be considered

by virtual lab users.

Table 3: Cost control rules for Virtual Lab users.

Rule Description

1 Shut down all instances after ﬁnishing experiments.

2 Use the smallest possible instance types for experiments.

3 Delete all unnecessary stopped instances.

4 Generated costs are considered for grading.

6 SUMMARY AND OUTLOOK

Not mentioned so far but very important. Cloud

based virtual labs are scalable. The here presented

approach is working with 10 students. It is also work-

ing with 100 or even 1000 students. We only have to

provide more virtual labs but do not have to invest into

dedicated hardware with a typical three or ﬁve year

or even longer ﬁnancial commitment. That provides

ﬂexibility and provides the possibility to manage per-

iods of time which a signiﬁcant step-up or step-down

of students.

Nevertheless not all practical courses are virtual

lab compatible. This might be of technical, functional

or of economical reasons. This contribution presented

a pragmatical four step decision making model for

or against IaaS based virtual IT labs for educational

purposes primarily from an economical point of view.

The decision making is pragmatical and inspired by

(Weinmann, 2011). We applied this decision making

model (see section 3) in a concrete use case of prac-

tical educational labs in the higher education domain

(colleges, universities, etc., see section 4) and showed

that it can be very economical to use cloud based ed-

ucational labs. It turned out that cloud based educa-

tional labs (in the analysed use case) have a more than

25 to 50 times cost advantage (see section 4.3 [step

3]) to classical dedicated approaches. So cloud com-

puting seems to be a very promising and economical

variant of providing educational labs for university

or college practical courses which is mainly due to an

inherent peaky usage characteristics of practical uni-

versity or college courses (see ﬁgure 3). This is a very

essential ﬁnding for the L

ubeck University of Applied

Sciences and might be of interest for other colleges,

universities or IT training facilities as well.

Furthermore this contribution reﬁned some gen-

eral ﬁndings to identify cloud compatible lectures and

practical courses (see section 5.1) and showed that the

approach of virtual labs is transferable to a signiﬁ-

cant amount of practical courses or lectures in com-

puter science study programs (see section 5.2). Other

study programs were not analyzed but the here pre-

sented virtual lab approach is not necessarily bound to

computer science study programs. Our transferability

analysis showed that all courses with requirements of

providing databases, distributed processing capabili-

ties, parallel processing or large-scale data processing

are interesting candidates. These requirements might

be also typical for engineering study programs.

Finally this contribution reﬁned some advices for

applying virtual labs in practical courses and proposes

some easy mind-keeping rules for efﬁcient cost con-

trol (see section 5.3 and table 3).

Outlook. In our ongoing research we plan to transfer

the here mentioned cloud based virtual lab approach

to more lectures and their related practical courses in

order to double check our conclusions and to harden

our postulations. By collecting more usage as well

as billing data from different courses we hope to get

an even better understanding of relevant cost driving

parameters. By collecting this data we hope to im-

prove a more precise cost estimation. It is furthermore

planned to develop a kind of management software

ARECLOUDENABLEDVIRTUALLABSECONOMICAL?-ACaseStudyAnalyzingCloudbasedVirtualLabsfor

EducationalPurposes

361

for setting up and managing virtual labs. A key fea-

ture would be an automatic average to peak ratio cal-

culation based on provided usage data by a cloud ser-

vice provider in order to answer the question whether

a cloud based approach is more economical than a

dedicated one.

ACKNOWLEDGEMENTS

Thanks to Amazon Web Services for supporting our

ongoing research with several research as well as ed-

ucational grants. Thanks to my students and Michael

Breuker for using cloud computing in practical educa-

tion. This contribution would not exist without their

engagement. Let me thank Alexander Schlaefer and

Uwe Krohn for organizing the World Robotic Sailing

Championship 2011 in L

ubeck and their conﬁdence

in our students.

REFERENCES

Barr, J. (2010). Host your Web Site in the Cloud. sitepoint.

Cayirci, E., Rong, C., Huiskamp, W., and Verkoelen, C.

(2009). Snow leopard cloud: A multi-national educa-

tion training and experimentation cloud and its secu-

rity challenges. In LNCS CloudCom 2009, pages 57 –

68. Springer.

Dong, B., Zheng, Q., Qiao, M., Shu, J., and Yang, J. (2009).

Bluesky cloud framework: An e-learning framework

embracing cloud computing. In LNCS CloudCom

2009, pages 577–582. Springer.

Ko, C.-C. and Young, S. S.-C. (2011). Explore the next

generation of cloud-based e-learning environment. In

LNCS Edutainment 2011, pages 107–114. Springer.

Kratzke, N. (2011a). Cloud-based it management impacts.

In Proceedings of CLOSER 2011, pages 145–151.

Kratzke, N. (2011b). Overcoming ex ante cost intrans-

parency of clouds. In Proceedings CLOSER2011 (spe-

cial session on Business Systems and Aligned IT Ser-

vices - BITS 2011), pages 707–716.

Kratzke, N. (2012a). Cloud computing costs and bene-

ﬁts (to be published). In Ivanov, van Sinderen, and

Shishkov, editors, Cloud Computing and Service Sci-

ence, Lecture Notes in Business Information Process-

ing. Springer.

Kratzke, N. (2012b). What cost us cloud computing? how

to decide for or against iaas using the average to

peak ratio atp (to be published). In Proceedings of

CLOSER2012.

Li, N. and Li, G. (2011). Research on digital campus based

on cloud computing. In Proceedings of CESM2011,

pages 213–218. Springer.

Liang, P.-H. and Yang, J.-M. (2011). Virtual personalized

learning environment (vple) on the cloud. In Proceed-

ings of WISM 2011 (LNCS), pages 403–411. Springer.

Masud, A. H. and Huang, X. (2011). Esaas: A new edu-

cation software model in e-learning systems. In Pro-

ceedings of ICCIC 2011, pages 468–475. Springer.

Mazhelis, O., Tyrv

ainen, P., Eeik, T. K., and Hiltunen, J.

(2011). Dedicated vs.. on-demand infrastructure costs

in communications-intensive applications. In Proc. of

the 1st international Conference on Cloud Computing

and Services Science (CLOSER2011), pages 362–370.

Mell, P. and Grance, T. (2011). The nist deﬁnition of cloud

computing (draft). Technical report, National Insti-

tute of Standards and Technology (U.S. Department

of Commerce).

Oberg, M., Woitaszek, M., Voran, T., and Tufo, H. M.

(2011). A system architecture supporting high-

performance and cloud computing in an academic

consortium environment. Computer Science - Re-

search and Development.

Talukader, A. K., Zimmermann, L., and Prahalad, H.

(2010). Cloud economics: Principles, costs and bene-

ﬁts. In Cloud Computing - Computer Communications

and Networks, volume 4, pages 343–360. Springer.

Thi

ebaut, D., Li, Y., Jaunzeikare, D., Cheng, A., Recto,

E. R., Riggs, G., Zhao, X. T., Stolpestad, T., and

Nyguyen, C. L. T. (2011). Processing wikipedia

dumps - a case-study comparing xgrid and mapre-

duce approaches. In Proc. of the 1st international

Conference on Cloud Computing and Services Science

(CLOSER2011).

Truong, H.-L. and Dustdar, S. (2010). Cloud computing for

small research groups in computational science and

engineering. Computing.

Walker, E., Brisken, W., and Romney, J. (2010). To lease

or not to lease from storage clouds. Computer, pages

44–50.

Weinmann, J. (2011). Mathematical proof

of the inevitability of cloud computing.

http://www.JoeWeinman.com/Resources/Joe Wein-

man Inevitability Of Cloud.pdf.

CSEDU2012-4thInternationalConferenceonComputerSupportedEducation

362