Decision Support for Application Migration to the Cloud

Challenges and Vision

Vasilios Andrikopoulos, Steve Strauch and Frank Leymann

Institute of Architecture of Application Systems, University of Stuttgart, Stuttgart, Germany

Keywords:

Application Migration, Decision Support, Cloud-enabling Applications.

Abstract:

The success of Cloud computing has encouraged many application developers to consider migrating their

applications to the Cloud. Given the early market dominance of the IaaS service model, many existing works

focus on selecting the best service provider for a set of criteria related to the virtualization and hosting of the

application. In this work, we aim to progress the State of the Art by formulating a vision of a decision support

system that incorporates multiple dimensions and different analysis tasks in feedback relationships with each

other. The research challenges that need to be addressed towards this direction are identiﬁed and related to the

different aspects of migration of applications to the various Cloud service models.

1 INTRODUCTION

The steadily increasing domination of Cloud comput-

ing in the software market means that existing appli-

cations may need to migrate to this environment in

order to reap the beneﬁts of reduced infrastructural

costs and dynamic access to computational resources.

While software has already started being developed

speciﬁcally for the Cloud (forming Cloud-native ap-

plications), existing systems must be adapted to be

suitable for the Cloud, requiring to make them Cloud-

enabled (Andrikopoulos et al., 2013). Existing work,

e.g. (Li et al., 2010; Khajeh-Hosseini et al., 2012;

Menzel and Ranjan, 2012), focuses on the IaaS ser-

vice model (Mell and Grance, 2011), building on the

success and early market dominance of infrastructure

virtualization solutions like Amazon Web Services.

In this context, providing decision support for the

migration of applications to the Cloud is reduced to

the selection of the appropriate IaaS provider that can

support the best application performance for the least

cost.

Beyond this VM-based migration of applications

to the Cloud however, new PaaS and SaaS offerings of

Cloud providers enable alternative migration options

for applications. Data stores like the Amazon Rela-

tional Database Service (Amazon RDS), and business

logic executing environments like the Google App

Engine allow for partially or fully migrating the ap-

plication. Decision support in this case needs also to

include different options for segmenting and distribut-

ing the application in the Cloud, ﬁguring out which

parts of the application should scale, and how multi-

tenancy can be implemented to maximize the return

of investment for application developers.

The position that this work takes is that migrat-

ing to the Cloud is a multi-dimensional problem with

multiple decision points that may create feedback

loops with each other, and with various analysis tasks

related to them. The contribution of this paper is to

identify these decision points and analysis tasks, and

deﬁne their dependencies. Based on these results,

a vision of a Cloud migration decision support sys-

tem considering these aspects is provided. In order

to reach this point however, a comprehensive analysis

of the challenges related to migrating to the Cloud is

required.

The rest of this paper is structured as follows: Sec-

tion 2 summarizes the State of the Art in decision sup-

port systems for Cloud migration. Section 3 discusses

the issues in migrating applications to the Cloud and

identiﬁes the research challenges for supporting the

decisions related to them. Section 4 presents our vi-

sion for how a decision support system can address

these challenges; Section 5 concludes this paper.

2 STATE OF THE ART

Existing work on decision support for application mi-

gration to the Cloud identify a number of criteria to

be considered, with the cost being the most impor-

149

Andrikopoulos V., Strauch S. and Leymann F..

Decision Support for Application Migration to the Cloud - Challenges and Vision.

DOI: 10.5220/0004407001490155

In Proceedings of the 3rd International Conference on Cloud Computing and Services Science (CLOSER-2013), pages 149-155

ISBN: 978-989-8565-52-5

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

tant one. The work presented in (Khajeh-Hosseini

et al., 2011) discusses two decision support tools im-

plemented as spreadsheets: a cost calculation one,

and a risk-beneﬁt analysis one. The former is geared

towards IaaS solutions, and allows users to calculate

the cost for their infrastructure when virtualized and

ran in one or more VMs on the Cloud. The latter sur-

veys and categorizes a number of beneﬁts and risks

identiﬁed in the literature (also for IaaS solutions),

and informs about potential consequences, mitiga-

tion approaches and indicators for each risk. The

two tools are then combined in the Cloud Adoption

Toolkit (Khajeh-Hosseini et al., 2012) that supports

decisions on technology suitability, cost, energy con-

sumption, stakeholder impact and operational feasi-

bility.

The (MC

)

framework (Menzel et al., 2011),

later incorporated in the CloudGenius toolkit (Menzel

and Ranjan, 2012), also focuses on IaaS solutions and

provides a multi-criteria approach in decision sup-

port for application migration. Both approaches al-

low users to deﬁne multiple quantitative and quali-

tative requirements that are then matched against a

knowledge base of Cloud service providers. Various

criteria to evaluate the possible alternatives are iden-

tiﬁed, and a series of cost categories are considered.

Beyond the emphasis on the IaaS service model, the

main criticism of all of the above works is the lack

of connecting the outcomes of each analysis step with

the particular decisions to be made. In this sense, sys-

tem users are able to evaluate their design, but they

do not have any insight into how each design decision

affects the overall system.

In particular with respect to selecting appropriate

Cloud service providers for the migration, Brebner

and Liu (Brebner and Liu, 2011) report on the costs

of running applications with different work loads in

various Amazon, Google and Microsoft offerings. A

similar calculation is reported in (Li et al., 2010)

and (Suleiman et al., 2011) for anonymized Cloud ser-

vices. The benchmarking of (Li et al., 2010) takes

into account storage and networking costs, while the

analysis of (Suleiman et al., 2011) focuses on the elas-

ticity options of the offerings. While these works can

be used as part of the decision support process, they

do not consider the characteristics of the application

itself it terms of how it needs to be adapted to operate

in the Cloud environment.

3 CLOUD MIGRATION

CHALLENGES

As the ﬁrst step towards a decision support system

beyond the State of the Art, in the following we use

the issues and concerns identiﬁed in previous work on

application migration (Andrikopoulos et al., 2013) as

the basis to identify the research challenges affecting

decision support for Cloud migration.

Application Distribution. Logically and physi-

cally distributing the data and the computational as-

pect of applications over the Cloud creates a series of

economic, performance, and legal issues. The calcu-

lations in (Armbrust et al., 2009) illustrate the trend

of computational power becoming cheaper faster than

disk storage, which becomes cheaper faster than net-

work bandwidth in turn. It is therefore in principle

good practice to keep data and computation close to

each other. The actual physical proximity of the mi-

grated application components and their data is how-

ever beyond the control of the application developer

— except from the location of the Cloud provider data

center.

Similarly, it becomes very difﬁcult to ensure com-

pliance to legal and regulatory requirements, concern-

ing for example the privacy of users. EU regulations

for instance require the physical location of the stored

data to be inside the EU borders (Reese, 2009). Even

if it is possible to deﬁne the region for deploying the

application, the cost of using them can vary due to dif-

ferent charging policies (Suleiman et al., 2011). Fur-

thermore, previous work has shown that Cloud ser-

vices performance varies signiﬁcantly for different re-

gions (Schad et al., 2010), or even for different data

centers inside the same region (Li et al., 2010).

It is important to notice that distributing the appli-

cation may be necessary in order to address the need

for business resiliency (Andrikopoulos et al., 2013),

while aiming to minimize the risk (both in terms of

security and QoS assurances) of the application mi-

gration to the Cloud. This requires backup solutions

and use of multiple Cloud providers as alternatives

(see also (Badger et al., 2012)), translates into a mul-

tiplication of the costs, and may require application

re-design.

Research Challenges. How to estimate the effect of

distributing logically and physically the application

in terms of performance and cost? How to address

compliance across different jurisdictional domains?

What is the trade-off between business resiliency and

additional incurred costs?

Elasticity. Elasticity provides the means for opti-

mizing resource usage in case of ﬂuctuating and/or

unknown application work loads. Ideally, the con-

sumer perceives an inﬁnite number of resources avail-

able in any quantity, at any time. Existing works

CLOSER2013-3rdInternationalConferenceonCloudComputingandServicesScience

150

like (Brebner, 2012; Suleiman et al., 2011; Vaquero

et al., 2011) however connect the beneﬁts from elas-

ticity mechanisms not only with the Cloud solutions

themselves, but also with the particular characteris-

tics of the application in terms of its work load. Two

types of scalability are identiﬁed in (Vaquero et al.,

2011): horizontal, i.e. addition of more application

instances where required, and vertical, that is, addi-

tion of more computational resources to the applica-

tion. While vertical scalability is possible in principle

for all applications, it largely depends on the service

provider to offer the mechanisms to implement scal-

ing dynamically. Horizontal scalability on the other

hand mostly depends on the application components

and the application as a whole to support it as an op-

tion. Scaling a stateful component like a database, for

example, requires additionally the choice of an ap-

propriate data replication mechanism (Andrikopoulos

et al., 2013).

Different elasticity strategies can be implemented

depending on the decisions to be made on a) what

part of the application to scale, b) how much it is nec-

essary to scale taking also into account issues related

to licensing costs for multiple VMs, c) which scaling

type to use and, d) when and how fast is it necessary

to scale, considering the observed effect of scaling la-

tency (Brebner, 2012) for starting new images in the

Cloud. Furthermore, the choice of the Cloud deploy-

ment model also imposes limitations to the beneﬁts of

elasticity, as identiﬁed by (Badger et al., 2012). Pri-

vate Cloud solutions, for example, impose essentially

the same limitations in maximum capacity as in the

case of traditional data centers.

Research Challenges. How to decide which elasticity

strategy to implement for a given application? What

is the effect of this decision to the operational costs of

the application under different work loads?

Multi-tenancy. The multi-tenant model of serving

multiple consumers from a common pool of comput-

ing resources, including storage, processing, mem-

ory and network bandwidth, is one of the essential

characteristics of Cloud computing (Mell and Grance,

2011). There are two fundamental aspects of multi-

tenancy awareness (Strauch et al., 2012a): commu-

nication, i.e. supporting message exchanges isolated

per tenant, and administration and management, i.e.

allowing each tenant to conﬁgure and manage individ-

ually their communication endpoints at application or

service level. Tenant isolation is further decomposed

into data and performance isolation between tenants

of the same system. Existing approaches on enabling

multi-tenancy typically focus on different types of

isolation in multi-tenant applications for the SaaS de-

livery model, see for example (Guo et al., 2007); most

of the work however, especially for performance iso-

lation, remains an open research challenge (see for ex-

ample (Alexandrov et al., 2012; Krebs et al., 2012)).

Furthermore, as discussed in (Andrikopoulos

et al., 2013), the three Cloud service models differ

signiﬁcantly in the granularity of the functionality

provided to the consumer, and the required capabil-

ity of the consumer to manage and control the under-

lying Cloud infrastructure. The result is a trade-off

between the effort required by the consumer to enable

multi-tenancy on application level (especially in the

IaaS model), and the responsibility of the provider to

offer the suitable mechanisms for this purpose (as ex-

empliﬁed in the SaaS model).

Research Challenges. Which Cloud service model

(IaaS, PaaS, SaaS) to choose when considering the

multi-tenancy needs of the application? What is the

effort required to implement multi-tenancy for the

chosen service model?

QoS. Migrating an application to the Cloud entails a

loss of control over the QoS characteristics of the ap-

plication due to the reliance on the QoS levels offered

by the service provider. As a result, the QoS charac-

teristics offered by a Cloud service provider appear to

have a greater importance to application stakeholders

than hosting the application in a non-Cloud environ-

ment. The most obvious case of this dependence is

the effect of service provider outages to businesses

hosting their applications with them. As discussed

in (Badger et al., 2012), availability is usually cal-

culated in practice on the basis of the billing cycle,

which may vary from minutes to hours, and not on

the total up-time of the service. Furthermore, failure

to provide this availability is typically compensated

with service credit in future use of the services. In

any case, the overall availability of a migrated ap-

plication can actually vary signiﬁcantly with respect

to the availability of the used services, depending on

the actual behavior of the application. Safely predict-

ing the availability of such an application is largely an

open issue.

Beyond availability, application QoS levels in the

Cloud are also affected by two major factors: the

performance variability of the Cloud providers, and

the network latency between the Cloud service con-

sumers and the service (Andrikopoulos et al., 2013).

While performance variability may be (partially) ad-

dressed by benchmarking Cloud service providers to

select an appropriate one in the manner of (Li et al.,

2010), network latency is in the general case beyond

the control of the application provider. SLAs with ex-

isting application consumers are therefore difﬁcult to

DecisionSupportforApplicationMigrationtotheCloud-ChallengesandVision

151

be enforced in this environment.

Research Challenges. What is the acceptable level

of Cloud service availability and performance vari-

ability for an application with respect to existing and

future SLAs? How to factor and address network la-

tency in the performance analysis? How does the

choice of service provider affects this analysis?

Migration & Operation Cost. As discussed in

Section 2, the focus of existing work on decision sup-

port systems is on selecting the best provider in terms

of costs for a given application (Hajjat et al., 2010;

Khajeh-Hosseini et al., 2012; Menzel and Ranjan,

2012), and under different work loads (Li et al., 2010;

Brebner and Liu, 2011). Beyond provider selection

however, cost analysis should also consider a num-

ber of other aspects. Firstly, different providers offer

different pricing models, see for example (Suleiman

et al., 2011), and combinations of these models, e.g.

subscription to a service like Windows Azure offer-

ing a discount on the per-use charging for given us-

age quotas, beyond which regular per-use rates ap-

ply. More complicated models have also starting aris-

ing, with Amazon Web Services for example allowing

customers to bid for unused EC2 capacity by means

of Spot Instances (Amazon Web Services, 2012).

Furthermore, software licensing is often based on

the number of CPUs which does not ﬁt the dynam-

icity of the solutions offered by the Cloud (Reese,

2009). For example, scaling a system may easily re-

sult in unintended license agreement violations. In

practice, some Cloud providers offer different licens-

ing options to their consumers. In Amazon RDS

for example, consumers can bring their own license

for MySQL, Oracle or IBM DB2, get charged for

a per-hour license using Oracle DB, or pay a one-

time charge per Amazon RDS instance to get reduced

hourly charging rates (Suleiman et al., 2011). Some

companies include licensing fees for free with each

account.

The cost analysis for the migration of an applica-

tion to the Cloud requires also to compare the costs

of operating in the Cloud versus operating in a local

data center. The discussion in (Armbrust et al., 2009)

illustrates the factors that allow Cloud computing to

be more proﬁtable than on-premises hosting: pay per

use only, avoid data center power, cooling and build-

ing costs, out-sourcing of the operational costs to the

provider. Existing works identify a relation between

the type and work load of the application, and the ben-

eﬁts by migrating the operation to the Cloud. In (Tak

et al., 2011) for example, the authors conclude that

complete application migration is beneﬁciary only for

small or stable organizations, while partially migrat-

ing some of the application components to the Cloud

is too expensive due to the high costs of data transfer.

Reese posits that, in principle, signiﬁcant cost savings

due to the transfer from capital expenses (CAPEX) to

operating expenses (OPEX) incur only as the variance

increases between peak, average and low capacity of

the system (Reese, 2009).

Finally, cost analysis should additionally consider

the effort required for the adaptation of the applica-

tion in order to operate in the Cloud. In (Andrikopou-

los et al., 2013), the authors identify four types of ap-

plication migration to the Cloud: replacement of ex-

isting application components with Cloud offerings,

partial migration of some of the application function-

ality to the Cloud, migration of the complete software

stack, and cloudiﬁcation of the application by replac-

ing all of its components with Cloud services. De-

pending on the type of the application, and the ar-

chitectural layer that is affected by the migration, the

authors conclude that different degrees of adaptation

effort may be required.

Research Challenges. How to decide on the best

provider in terms of costs in the presence of different

pricing models? What is the trade-off between cost

and performance? How to decide which deployment

option (completely locally, fully on the Cloud, combi-

nation of the above) ﬁts better the application work

load proﬁle in question? What is the additional effort

required for adapting the application to operate in the

Cloud?

Security & Data Conﬁdentiality. Security and

conﬁdentiality of data are major concerns and ob-

stacles for many enterprises to migrate to the

Cloud (Armbrust et al., 2009) and for this reason they

need to be addressed separately from the other QoS

concerns. Security entails both the communication

and data aspects, but also the physical/digital one, i.e.

the risk of losing or compromising data due to data

center failures or other physical attacks. In princi-

ple, security is a shared concern between Cloud ser-

vice providers and consumers (Andrikopoulos et al.,

2013). On one hand, service providers are required

to offer security-enabling mechanisms like encrypted

communications. On the other hand however, appli-

cation developers are the ones responsible for adapt-

ing the migrated application accordingly to use these

mechanisms and further conﬁgure it appropriately.

Ultimately, it is also the responsibility of the appli-

cation users to interact with it in a secure manner

following for example the security recommendations

discussed in (Badger et al., 2012).

In addition to security, conﬁdentiality of Cloud-

migrated data is an important concern for all kinds of

CLOSER2013-3rdInternationalConferenceonCloudComputingandServicesScience

152

organizations. Conﬁdentiality refers to keeping criti-

cal data private and secure. Critical data may be busi-

ness secrets, personal data, health care data and so on.

(Strauch et al., 2012b; Andrikopoulos et al., 2013)

provide a number of techniques like pseudonymiza-

tion and anonymization that can be used to ensure

the conﬁdentiality of migrated data. However as they

point out, these techniques are not sufﬁcient to avoid

data breaches due to e.g. attacks from inside the Cloud

environment, and the leaking of information while

processing these data in the Cloud. Data retention

policies beyond contract terms are also not explicitly

deﬁned for most of the existing providers, creating

possible sources of data breaches.

Research Challenges: What is the effort to

(re-)engineer the application to incorporate security

mechanisms like encryption of data? Which part of

the application data can be characterized as critical,

and under which conditions can it be migrated to the

Cloud securely?

4 DECISION SUPPORT FOR

CLOUD MIGRATION

What becomes obvious from the previous discussion

is that migrating an application to the Cloud requires

making a number of decisions related to the how the

application is to be refactored for the Cloud, how it

is supposed to scale, an so on. In this sense, cost

analysis and performance prediction, which has been

the focus of the SotA as discussed in Section 2 pro-

vide only partial support for these purposes. Address-

ing these concerns, Fig. 1 provides an overview of

the concepts required to implement a decision support

system to support the migration of applications to the

Cloud. This conceptual view deﬁnes our vision for

what constitutes a complete solution for application

developers and stakeholders alike that are consider-

ing whether and how to migrate their application to

the Cloud.

More speciﬁcally, two types of concepts are iden-

tiﬁed: decisions that need to be made (and therefore

are the focus of the system), and tasks that need to be

performed in order to support these decisions — and

that in turn affect their outcome. The key decision

that the system supports refer to:

1. How to distribute the application across service

providers, and between the Cloud and the lo-

cal data center. Different application topolo-

gies based on the various identiﬁed migration

types (Andrikopoulos et al., 2013) may be con-

sidered.

2. How to select a (Cloud) service provider and of-

fering that ﬁts the application needs in terms of

cost, expected performance, compliance require-

ments and security concerns.

3. What are the requirements of the application in

terms of multi-tenancy, i.e. to what extent the

existing application is required to support multi-

tenancy, to what degree it is designed for this pur-

pose, and how it should be (re-)engineered to sup-

port multi-tenancy.

4. Which is the elasticity strategy that the applica-

tion needs to implement in order to cope with its

demand in the face of SLAs and expectations of

its users.

Each of these decisions has a direct or unintentional

inﬂuence on the others, illustrated with transparent ar-

rows in Fig. 1. Deciding on the appropriate elastic-

ity strategy for example is meaningless without the

Cloud service provider supporting it. The selection of

service provider includes the decision which service

delivery model to use. As a result, both the options

in distributing the application and how multi-tenancy

can be implemented is directly inﬂuenced by it.

In addition to the decisions and their relationships,

the following tasks are also identiﬁed:

• Work Load Proﬁling: Deﬁning or estimating the

expected work load proﬁle of the application is

prerequisite for both other tasks to be performed

(performance calculation and cost analysis), as

well as input for any decision on how to distribute

the application, and which elasticity strategy ﬁts

this proﬁle.

• Compliance Assurance. Ensuring the compli-

ance to regulations regarding, e.g., privacy of per-

sonal data, affects directly both the selection of

the provider (especially in terms of location of the

data service), as well as how the application is to

be distributed so that e.g. personal data may need

to be retained on premises.

• Identiﬁcation of Security Concerns. Deﬁning

which data and communications are critical to be

protected drives the selection of an appropriate

service provider that fulﬁlls these security con-

straints. Delivering the application in a multi-

tenant manner imposes further constraints with re-

spect to data isolation that need to be considered

for this task.

• Identiﬁcation of Acceptable QoS Levels. Based

on existing and planned SLAs, acceptable levels

for QoS characteristics like availability of the ser-

vice provider can be inferred. Beyond service

provider selection, this task also drives the deﬁni-

tion of an appropriate elasticity strategy to ensure

DecisionSupportforApplicationMigrationtotheCloud-ChallengesandVision

153

Distribute

Application

Select Service

Provider/Offering

Define

Multi-tenancy

Requirements

Define

Elasticity

Strategy

Work load

profiling

Performance

prediction

Cost

analysis

Effort

estimation

Identification

of security

concerns

Compliance

assurance

Identification

of acceptable

QoS levels

Decision

Task

Affects

Influences

Figure 1: Decision Support System for Cloud Migration: conceptual view.

these QoS levels (for the given work load proﬁle)

and constrains the options for application multi-

tenancy.

• Performance Prediction. Depending on how the

application is distributed across different service

providers, which providers where selected for this

purpose, and which elasticity strategy was de-

ﬁned, this task creates projections about the non-

functional behavior of the application after it is

migrated to the Cloud. These projections can in

turn be used to change any of the previous deci-

sions, starting a new feedback loop between them.

• Cost Analysis. Similarly to performance predic-

tion, cost analysis both affects and is affected by

the application distribution, elasticity strategy and

service provider selection decisions. Furthermore

however, this task should also consider the effort

estimation for adapting the application to operate

in the Cloud.

• Effort Estimation. This task focuses on provid-

ing an estimate related to the amount of work re-

quired to adapt the application depending on the

type of migration that is considered, and the ar-

chitectural layer that is affected. As a result, this

task requires input from the application distribu-

tion, service provider selection and multi-tenancy

requirements decisions, and may result in changes

to them in turn.

As shown in Fig. 1, decisions and tasks form a

network of relationships, providing multiple points of

entry to the decision making process of migrating an

application to the Cloud. A Cloud migration deci-

sion support system should therefore support this pro-

cess not in a structured hierarchical manner, but rather

in a networked fashion, allowing application design-

ers to provide different answers for each decision that

needs to be made, ask them for the appropriate input

required to perform the identiﬁed tasks, and demon-

strate the effects of their decisions to concrete goals

like the cost and performance of the application. Such

an approach is missing from the current State of the

Art and constitutes our future work in the area.

5 CONCLUSIONS

The position we took in this work is that support-

ing the decision making process in migrating the ap-

plication to the Cloud is a multi-dimensional prob-

lem with multiple decisions to be taken: how to seg-

ment and distribute the application, which Cloud ser-

vice provider and offering to choose, which elastic-

ity strategy ﬁts better the application, and what are

CLOSER2013-3rdInternationalConferenceonCloudComputingandServicesScience

154

the requirements in implementing application multi-

tenancy. These decisions inﬂuence each other, and

depend on a number of tasks like cost analysis and

performance prediction that need to be performed in

order to support them. Related to these decisions and

tasks are a series of research challenges stemming

from the need to adapt the application to operate in

the Cloud environment that also need to be addressed.

Our vision for a decision support system for migration

of applications to the Cloud as discussed here consid-

ers all of the above points in order to advance the State

of the Art in the ﬁeld.

ACKNOWLEDGEMENTS

The research leading to these results has re-

ceived funding from the 4CaaSt project (http://

www.4caast.eu) part of the European Union’s Seventh

Framework Programme (FP7/2007-2013) under grant

agreement no. 258862.

REFERENCES

Alexandrov, A., Folkerts, E., Sachs, K., Iosup, A., Markl,

V., and Tosun, C. (2012). Benchmarking in the Cloud:

What it Should, Can, and Cannot Be. In 4th TPC

Technology Conference on Performance Evaluation &

Benchmarking (TPCTC), VLDB 2012.

Amazon Web Services (2012). How AWS Pricing Works.

http://aws.amazon.com/whitepapers/.

Andrikopoulos, V., Binz, T., Leymann, F., and Strauch, S.

(2013). How to adapt applications for the cloud en-

vironment. Computing (to appear). http://dx.doi.org/

10.1007/s00607-012-0248-2.

Armbrust, M. et al. (2009). Above the Clouds: A Berke-

ley View of Cloud Computing. Technical Report

UCB/EECS-2009-28, EECS Department, University

of California, Berkeley.

Badger, L., Grance, T., R., P.-C., and Voas, J. (2012). Cloud

computing synopsis and recommendations - recom-

mendations of the national institute of standards and

technology. NIST Special Publication 800-146.

Brebner, P. (2012). Is your cloud elastic enough?: perfor-

mance modelling the elasticity of infrastructure as a

service (iaas) cloud applications. In Proceedings of

ICPE’12, pages 263–266. ACM.

Brebner, P. and Liu, A. (2011). Performance and cost as-

sessment of cloud services. Service-Oriented Com-

puting, pages 39–50.

Guo, C., Sun, W., Huang, Y., Wang, Z., and Gao, B. (2007).

A Framework for Native Multi-Tenancy Application

Development and Management. In Proceedings of

CEC/EEE’07, pages 551–558. IEEE.

Hajjat, M., Sun, X., Sung, Y., Maltz, D., Rao, S., Sripanid-

kulchai, K., and Tawarmalani, M. (2010). Cloudward

bound: planning for beneﬁcial migration of enterprise

applications to the cloud. In ACM SIGCOMM Com-

puter Communication Review, volume 40, pages 243–

254. ACM.

Khajeh-Hosseini, A., Greenwood, D., Smith, J. W., and

Sommerville, I. (2012). The cloud adoption toolkit:

supporting cloud adoption decisions in the enterprise.

Software: Practice and Experience, 42(4):447–465.

Khajeh-Hosseini, A., Sommerville, I., Bogaerts, J., and

Teregowda, P. (2011). Decision support tools for

cloud migration in the enterprise. In 2011 IEEE Inter-

national Conference on Cloud Computing (CLOUD),

pages 541–548. IEEE.

Krebs, R., Momm, C., and Kounev, S. (2012). Metrics

and Techniques for Quantifying Performance Isola-

tion in Cloud Environments. In Buhnova, B. and Val-

lecillo, A., editors, Proceedings of the 8th Interna-

tional ACM SIGSOFT Conference on the Quality of

Software Architectures, CBSE’12, pages 91–100, New

York, USA. ACM Press.

Li, A., Yang, X., Kandula, S., and Zhang, M. (2010). Cloud-

cmp: comparing public cloud providers. In Proceed-

ings of the 10th annual conference on Internet mea-

surement, IMC ’10, pages 1–14, New York, NY, USA.

ACM.

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

computing (draft). NIST special publication, 800:145.

Menzel, M. and Ranjan, R. (2012). Cloudgenius: decision

support for web server cloud migration. In Proceed-

ings of WWW ’12, pages 979–988, New York, NY,

USA. ACM.

Menzel, M., Sch

onherr, M., and Tai, S. (2011). (mc2) 2: cri-

teria, requirements and a software prototype for cloud

infrastructure decisions. Software: Practice and Ex-

perience.

Reese, G. (2009). Cloud application architectures. O’Reilly

Media, Inc.

Schad, J., Dittrich, J., and Quian

e-Ruiz, J. (2010). Runtime

measurements in the cloud: observing, analyzing, and

reducing variance. Proceedings of the VLDB Endow-

ment, 3(1-2):460–471.

Strauch, S., Andrikopoulos, V., G

omez S

aez, S., Ley-

mann, F., and Muhler, D. (2012a). Enabling Tenant-

Aware Administration and Management for JBI Envi-

ronments. In Proceedings of SOCA’12. IEEE Com-

puter Society Conference Publishing Services.

Strauch, S., Breitenb

ucher, U., Kopp, O., Leymann, F., and

Unger, T. (2012b). Cloud Data Patterns for Conﬁden-

tiality. In Proceedings of CLOSER’12, pages 387–

394. SciTePress.

Suleiman, B., Sakr, S., Jeffery, R., and Liu, A. (2011). On

understanding the economics and elasticity challenges

of deploying business applications on public cloud in-

frastructure. Journal of Internet Services and Appli-

cations, pages 1–21.

Tak, B., Urgaonkar, B., and Sivasubramaniam, A. (2011).

To move or not to move: The economics of cloud

computing. In Third USENIX Workshop on Hot Topics

in Cloud Computing (HOTCLOUD 2011).

Vaquero, L., Rodero-Merino, L., and Buyya, R. (2011). Dy-

namically scaling applications in the cloud. ACM SIG-

COMM Computer Communication Review, 41(1):45–

52.

All links were last followed on March 12, 2013.

DecisionSupportforApplicationMigrationtotheCloud-ChallengesandVision

155