Applications Deployment in Multiple PaaS Environments:

Requirements, Challenges and Solutions

Rami Sellami

, Mehdi Ahmed-Nacer

2,3

and St

ephane Mouton

Software and Services Technologies Department, CETIC, Charleroi, Belgium

Computer Science Departement, Institut Mines-Telecom, Telecom SudParis, CNRS UMR Samovar,

University of Paris-Saclay, Evry, France

University of Sciences and Technology Houari Boumediene, Algiers, Algeria

Keywords:

Cloud Computing, Plateform-as-a-Service (PaaS), Multiple PaaS Deployment, Cloud Services Discovery,

Cloud Services Selection, Applications Migration.

Abstract:

Cloud computing has recently attracted full attention of many organizations due to its economic, business

and technical beneﬁts. Indeed, we observe that the proliferation of offers by cloud providers raises several

challenges. One of these innovative challenges is applications deployment in multiple PaaS providers. In fact,

developers need to provision components of the same application across multiple PaaS depending on their

related requirements and PaaS capabilities. They will not only have to deploy their applications, but they

will also have to consider migrating services from one PaaS to another, and to manage distributed applica-

tions spanning multiple environments. In this paper, we present and discuss the requirements of applications

deployment in multiple PaaS providers and we analyze current state of the art.

1 INTRODUCTION

Cloud Computing, a relatively recent term, has be-

come nowadays a buzzword in the Web applications

world. It is deﬁned by the National Institute of

Standards and Technology (Mell and Grance, 2009)

as a model for enabling ubiquitous, convenient, on-

demand network access to a shared pool of conﬁg-

urable computing resources (e.g., networks, servers,

storage, applications, etc.) that can be quickly provi-

sioned in an elastic manner that allows to automati-

cally scale up or down in line with demand. In addi-

tion, these resources are released with minimal man-

agement effort or service provider interaction. Cloud

Computing is characterized by its economic model re-

ferred to as ”pay-as-you-go” model. This latter allows

users to consume computing resources as needed.

Services in the Cloud Computing are basically de-

livered under three well discussed layers namely the

Infrastructure as a Service (IaaS), the Platform as a

Service (PaaS) and the Software as a Service (SaaS).

We present these layers below:

• IaaS: Consumers are able to access Cloud re-

sources on demand. These resources can be of

type virtual machines, storage and networks. The

provider is responsible for installing, transpar-

ently managing and maintaining these resources.

• PaaS: Consumers are able to develop, deploy

and manage their applications into the Cloud us-

ing libraries, editors and services offered by the

provider. It supports provisioning, managing and

maintaining the Infrastructure resources.

• SaaS: Consumers are able to use running applica-

tions on an IaaS or a PaaS through an interface.

They are not responsible for managing or main-

taining the provisioned Cloud resources.

Over the past years, a plethora of commercial so-

lutions and research projects have been sought to im-

prove security, scalability, elasticity and interoper-

ability in the IaaS level. Nevertheless, very few works

deal with the PaaS and SaaS levels even if they are as

important as the IaaS level. Hence, we propose to

focus in this document on the PaaS level. A PaaS

provider is supposed to support the whole require-

ments of an application during its life cycle while

maintaining Cloud features (e.g. scalability, elastic-

ity, high availability, etc.). However, applications re-

quirements change frequently. Thus, it seems illu-

sory to ﬁnd a single PaaS provider that efﬁciently

supports various applications with different require-

ments in terms of applications life cycle management.

636

Sellami, R., Ahmed-Nacer, M. and Mouton, S.

Applications Deployment in Multiple PaaS Environments: Requirements, Challenges and Solutions.

DOI: 10.5220/0006370906640671

In Proceedings of the 7th International Conference on Cloud Computing and Services Science (CLOSER 2017), pages 636-643

ISBN: 978-989-758-243-1

This forces an application to migrate to another PaaS

provider. Moreover, we can ﬁnd an application with

very speciﬁc requirements that a single PaaS provider

does not have the necessary capabilities to support it.

To circumvent this obstacle, an application may

be deployed in multiple PaaS providers and beneﬁts

from various PaaS providers capabilities at the same

time. For instance, an application can use the Re-

dis database service from DEIS and the middleware

Tomcat of Microsoft Azure. In this paper, we focus

on existing solutions of the state-of-the-art supporting

applications deployment in multiple PaaS providers.

More precisely, our contributions are (i) describing

different scenarios related to the way applications

are deployed in a single or multiple PaaS providers,

(ii) deﬁning a set of requirements in terms of ap-

plications deployment in a single or multiple PaaS

provider(s), and (iii) analyzing and classifying the re-

viewed works.

The rest of this paper is organized as follows. Sec-

tion 2 presents the multiple PaaS deployment require-

ments. In sections 3, 5, and 4, we analyze the state of

the art and we present the list of the considered crite-

ria and a synthesis of this analysis. Section 6 provides

a conclusion and some open issues.

2 MULTIPLE PaaS

DEPLOYMENT

REQUIREMENTS

The PaaS is the development, deployment and execu-

tion environment of an application in the Cloud. It

is devised to provide utmost beneﬁts to the develop-

ers and to get rid them of the onerous task of man-

aging applications life cycle. Indeed, it enables high

scalability and on demand cloud services provision-

ing by automating its allocation/ deallocation to opti-

mally support the requirements of an application. In

addition, it ensures a high level of availability and re-

liability thanks to the load balancing and the failover

of services capabilities with respect to the application

requirements. It is also characterized by the automa-

tion in order to ensure an end-to-end support for the

applications lifecycle. Last but not least, it allows de-

velopers to use multiple operating systems, multiple

run-time and frameworks, and multiple databases to

collaborate with other developers. Against this back-

ground, we can emphasize a ﬁrst multiple PaaS de-

ployment requirement which is:

• R

: Ensuring scalability, elasticity and portability

In order to introduce the other requirements, we

present two possible use cases of the PaaS level that

a developer may encounter. In Figure 1, we showcase

the ﬁrst use case which is applications deployment in

a single PaaS. In this case, the application can only

allocate the services of this PaaS. In this ﬁgure, we il-

lustrate three applications deployed in a single Cloud

environment and they are provisioning with three ser-

vices. Based on this, we stress the problem which is

to discover and to decide if the capabilities of a given

service are convenient for an application requirements

or not. In addition, applications are very dynamic and

frequently change its requirements in a Cloud envi-

ronment. Hence, they are obliged to allocate new ser-

vices. In most of the cases, an application is forced to

migrate to a new Cloud environment. In this context,

we identify the following requirements:

• R

: Choosing a PaaS provider based on Applica-

tions requirements

– R

: Exposing the application requirements

– R

: Capturing PaaS providers capabilities

– R

: Matching between application require-

ments and PaaS providers capabilities

• R

: Migrating an application from one PaaS to

another

Figure 1: Scenario 1 - Application deployment in a single

PaaS provider.

Whereas the second use case matches the appli-

cations deployment in multiple PaaS providers (see

Figure 2). For instance, the application App1 uses

two services that are service 1 from the Cloud envi-

ronment where it is deployed and service 2 from an-

other Cloud environment. We do not deny that this

approach is very interesting since it enables to op-

timally support applications requirements. It allows

taking beneﬁt from the variety of the PaaS offerings.

It also provides the developers with more alternatives

to satisfy their needs and optimize their applications

operation. In addition, it offers a PaaS-independent

mechanisms which considerably harmonize and sim-

plify the provisioning procedures of application com-

ponents across several PaaS providers. Nevertheless,

deploying an application in multiple PaaS providers

is a cumbersome and requires a painstaking work.

Based on this, we add the following requirement:

Applications Deployment in Multiple PaaS Environments: Requirements, Challenges and Solutions

637

• R

: Easy deployment to single or multiple PaaS

Figure 2: Scenario 2 - Application deployment in multiple

PaaS providers.

In the rest of the paper, we will analyze the cur-

rent state of the art regarding these four requirements.

We will mainly focus on the requirements R

with its

three sub-requirements and R

. We do not specify a

section to the requirement R

since we suppose that is

implicitly ensured through the other requirements.

3 CHOOSING A PaaS PROVIDER

BASED ON APPLICATIONS

REQUIREMENTS

Several works deal with the problem of the discov-

ery and selection of services in Cloud environments.

Generally, these contributions enable developers to

describe their applications requirements using a very

speciﬁc model (e.g. a manifest, a SLA-based model,

a metric-based model, etc.). Then, they propose a set

of mechanisms enabling the selection of the most ap-

propriate PaaS provider to deploy the application.

Sellami et al. (Sellami et al., 2015) propose an

approach to automatically discover Cloud resources

in a PaaS environment. Indeed, developers express

their requirements in an abstract application mani-

fest. Then, their matching algorithm discovers the

PaaS providers capabilities that are exposed in an of-

fer manifest. It computes the distance between the

two manifests in order to elect the most appropri-

ate PaaS provider to the application requirements. In

their solution, authors mainly focus on the discovery

of data services and interest to single PaaS discovery.

Redl et al. (Redl et al., 2012) present an auto-

matic approach to check if the elements of the SLA

are valid or not. Doing so, they map each PaaS SLA

to an ontology and they deﬁne a matching algorithm

to compare the SLAs. Results of this matching are

analyzed in order to select the most appropriate ser-

vices provider. Although the idea of selecting Cloud

services based on the SLA is interesting, this ap-

proach does not enable application requirements de-

scription in terms of deployment in one or multiple

PaaS providers.

Garg et al. (Garg et al., 2011) present SMICloud

which is a framework enabling the comparison be-

tween two Cloud providers based on users require-

ments. In fact, SMICloud computes a set of metrics

referred to as Service Measurement Indexes (SMI).

The results of these SMIs are used by users in order

(1) to compare the different offers of Cloud environ-

ments, (2) to classify them using a predeﬁned order,

(3) to evaluate their Quality of Service (QoS), and (4)

to select the most appropriate one to their require-

ments. Inspite of the completeness of SMICloud, it

is intended to work in the IaaS level.

Wittern et al. (Wittern et al., 2012) propose a

model based solution to select services in a Cloud en-

vironment in order to express users requirements and

services capabilities. For this purpose, they deﬁne an

algorithm that select services using a set of predeﬁned

criteria. Nevertheless, this approach does not support

the multiple PaaS discovery.

In the SeaClouds project (Athanasopoulos et al.,

), authors propose an open source platform to support

applications in a multiple clouds environments. In-

deed, one of its key component is the SeaClouds Dis-

coverer which enables to discover available capabili-

ties and add-ons offered by available cloud providers.

It allows to declaratively select multiple Cloud ser-

vices based on the QoS expressed by the user. A

matching algorithm is implemented in order to select

the Cloud provider corresponding to the QoS required

by the user.

Quinton et al. (Quinton et al., 2014; Quinton et al.,

2016) propose SALOON which is a platform for au-

tomatically selecting and conﬁguring cloud environ-

ments. Their solution is based on the functional and

non-functional application requirements in order to

select the most appropriate cloud environment to an

application. The key ingredient of their solution is the

use of the Software Product Lines and the Features

Models to represent the Cloud variability.

Li et al. (Li et al., 2010; Ang Li, 2010) deﬁne an

automatic solution to select the most appropriate PaaS

provider to an application requirements. This solu-

tion is referred to as CloudCmp. It compares the per-

formance and the cost of cloud providers in terms of

computing, storage, and networking resources using

benchmark tasks. It aims to elect the cloud provider

that has the best performance and the less cost. It

supports four cloud providers that namely are Ama-

zon AWS, Microsoft Azure, Google AppEngine, and

Rackspace CloudServers.

Kang et al. (Kang and Sim, 2011; Kang and Sim,

2016) propose an agent-based solution to discover

CLOSER 2017 - 7th International Conference on Cloud Computing and Services Science

638

cloud resources using ontologies. These ontologies

are referred to as CO-1 and CO-2, and enable to se-

mantically describe resources and the relationship be-

tween each others. The user requirements are deﬁned

in the form of classAds. Authors introduce the dis-

covery process using four stages: the selection, the

evaluation, the ﬁltering, and the recommendation.

Qu et al. (Qu et al., 2013) propose a model for

Cloud services selection by aggregating both (1) the

feedback issued from cloud users and (2) the objec-

tive performance benchmark testing from a trusted

third party. Their solution integrates four key compo-

nents: The Cloud Selection Service takes into charge

a request for services selection reception and pre-

processing. The Benchmark Testing Service which

designs a set of testing scenarios to evaluate the per-

formances metrics. The User Feedback Manage-

ment Service that is in charge to collect and manage

users feedback. The Assessment Aggregation Service

which aggregates the user feedback and the resulting

performance values in order to evaluate cloud services

scores and weight.

Han et al. (Han et al., 2009) propose a cloud ser-

vices selection framework based on a recommender

system with respect to their requirements. Indeed, the

recommender system deﬁne ranks to the different ser-

vices and the user should select the appropriate ones.

The rank is evaluated using the QoS and the analysis

of resources provided by cloud providers (using qual-

ity of virtualization, allocation cost, users feedback,

etc.). It is noteworthy that authors propose to select

multiple services from multiple cloud providers.

Table 1: Synthesis of the related works about multiple PaaS

providers discovery.

Studied Solutions

Criteria

Multiple PaaS

Declarative

Automatic

Sellami et al. (Sellami et al., 2015) + + + − + +

Redl et al. (Redl et al., 2012) − + + − + +

SMICloud (Garg et al., 2011) + + + − + +

Wittern et al. (Wittern et al., 2012) + + + − + +

SeaClouds + + + + + −

(Athanasopoulos et al., )

SALOON (Quinton et al., 2016) + + + + + +

(Quinton et al., 2014)

Li et al. (Li et al., 2010), + + + − + +

(Ang Li, 2010)

Kang et al. (Kang and Sim, 2011), + + + − + +

(Kang and Sim, 2016)

Qu et al. (Qu et al., 2013) + + + − + +

Han et al. (Han et al., 2009) + + + + + +

Synthesis

In this section, we present a synthesis about the pre-

sented works above. To do so, we ﬁx a set of six cri-

teria on which we rely to evaluate these works. We

check whether these works propose a solution to (1)

describe application requirements, (2) expose PaaS

providers capabilities and (3) apply matching tech-

niques to elect the most suitable PaaS provider to an

application. We also verify if these works use a model

to automatically and declaratively dicover and select

PaaS providers. Finally, we point out that we target

applications in multiple PaaS providers. In Table 1,

we check for each work, whether it responds or not

to our criteria. We use + to denote that the proposed

work treats a given criteria. While, we use − to show

that the related work does not propose a solution for

a given criteria. It is worth noting that we use this no-

tation throughout the upcoming sections to synthesis

the studied works.

Against this analysis, we conclude that these

works support the requirement R

, R

, and R

In addition, each solution has its own modeling in

order to ensure applications requirements and PaaS

providers capabilities description in a declarative way.

However, apart from some works, the rest does not

support multiple PaaS providers discovery and selec-

tion.

4 EASY DEPLOYMENT TO

SINGLE OR MULTIPLE PaaS

PROVIDERS

Today, we ﬁnd several solutions enabling application

provisioning and deployment in Cloud environments.

However, the majority of these works focus on the

IaaS level. For instance, we can cite mOSAIC

, Per-

fCloud (Mancini et al., 2009), and PaaSage

. In-

deed, mOSAIC is a european project and proposes

a solution enabling the portability and the interop-

erability of data, services, and applications in differ-

ent IaaS providers. This project is based on a set of

negotiation mechanisms, contracts and ontologies al-

lowing to ﬁnd Cloud services supporting applications

requirements. We also introduce the PaaSage euro-

pean project which proposes a model driven engineer-

ing based solution to model applications requirements

and IaaS capabilities and to deploy applications in

multiple IaaS providers. Although the completeness

and the performance of these solutions, it does not

http://www.mosaic-cloud.eu/

http://www.paasage.eu/

Applications Deployment in Multiple PaaS Environments: Requirements, Challenges and Solutions

639

support the PaaS level. Nevertheless, we ﬁnd some

approaches to deploy applications in the PaaS. In fact,

we cite Cloud4SOA (D’Andria et al., 2012; Kama-

teri et al., 2013), PaaSHopper (Paraiso et al., 2012;

Walraven et al., 2015), ANEKA (Wei et al., 2011),

COAPS (Sellami et al., 2013), SeaClouds (Athana-

sopoulos et al., ), Nucleus (Kolb and R

ock, 2016;

ock and Kolb, 2016), PaaS Manager (Cunha et al.,

2014) and Spinnaker

Cloud4SOA (D’Andria et al., 2012; Kamateri

et al., 2013) is a FP7 european project enabling to

interconnect heterogeneous PaaS providers that share

the same technologies. To do so, they use an ontology

to cover the heterogeneity between a set of software

components and models. Hence they ease the task

of developers and enhance platform providers with a

number of core capabilities. It is noteworthy that this

solution only supports SOA-based applications.

PaaSHopper (Paraiso et al., 2012; Walraven et al.,

2015) is a middleware platform enabling to develop

and deploy SaaS-based applications in multiple PaaS

providers. In fact, developers declaratively describe

their application features and requirements in terms of

PaaS. Then, PaaSHopper selects the appropriate PaaS

providers to execute the application and store its data.

ANEKA (Wei et al., 2011) is a framework for

building applications based on users requirements and

deploying them on multiple PaaS providers. It sup-

ports private and public providers. It includes four

key components. Indeed, the Aneka Master and the

Aneka Worker taking into charge the deployment pro-

cess. Whereas the Aneka Management Console and

the Aneka Client Libraries take into charge the appli-

cation development.

The Compatible One Application and Platform

Service API (COAPS API) (Sellami et al., 2013)

allows human and/or software agents to provision

and manage PaaS applications. This API provides a

unique layer to interact with any Cloud provider based

on manifests. It manages two kinds of resources: (1)

the environment which is a set of settings required

by developers to host and run their applications, and

(2) applications which are computer software or pro-

gram deployable in a Cloud environment. Recently,

COAPS API has been extended to support multiple

PaaS providers deployment.

The SeaClouds project (Athanasopoulos et al., )

enables to deploy applications in multiple and het-

erogeneous Clouds (i.e. IaaS and PaaS). Indeed, it

is based on two main components. The ﬁrst one is re-

ferred to as the SeaClouds Planner which generates an

orchestration plan with respect to the application re-

quirements and topology. Whereas the second one is

http://www.spinnaker.io/

the SeaClouds Deployer that executes the deployment

plan in one or multiple Clouds provider.

Nucleus (Kolb and R

ock, 2016; R

ock and Kolb,

2016) is a uniﬁed interface enabling applications de-

ployment and management in PaaS providers. It is

based on the REST architecture. This interface covers

the heterogeneity between the proprietary APIs and

models of each PaaS provider.

PaaS Manager (Cunha et al., 2014) is a frame-

work to integrate and aggregate various public PaaS

providers based on their similarities. Doing so, this

framework includes a unique API in order to reduce

the vendor lock-in between PaaS providers and ap-

plications developers. Indeed, it enables to create,

deploy, start, stop, delete and monitor an applica-

tion. It includes a REST API to automatically de-

ploy an application in a PaaS provider using Git. It is

noteworthy that this solution supports only four PaaS

providers and enables single PaaS deployment.

Spinnaker (Spinnaker, 2017) is an open source so-

lution enabling to prepare environment based on mul-

tiple clouds and to continuously deploy and manage

applications. It involves two key component. The

management component to manage resources in the

cloud and the deployment management component to

construct and manage continuous delivery workﬂows.

Synthesis

In this section, we present a synthesis of the works

that we have analyzed above (see Table 2). Indeed,

we investigate whether these solutions enable an easy

deployment in multiple PaaS providers. This will al-

low taking beneﬁt from the variety of the PaaS offer-

ings. It also provides the developers with more alter-

natives to satisfy their needs and optimize their appli-

cations operation. In addition, we check if these so-

lutions are declarative. Indeed, this criteria is ensured

using a uniﬁed model and generic provisioning mech-

anisms. PaaS-independent mechanisms will consider-

ably harmonize and simplify the provisioning proce-

dures of applications components across several PaaS.

Finally, we are interested in the works that enable

automatic provisioning of applications across several

PaaS providers. This simpliﬁes the deployment pro-

cess and alleviates the burden of this task imposed on

developers.

Against this analysis, we conclude that the ma-

jority of the studied works support multiple PaaS.

However, Cloud4SOA and SeaClouds partially meets

the second requirement related to the automatic

provisioning of applications across several PaaS.

Cloud4SOA supports only provisioning of applica-

tions that are designed according to SOA speciﬁ-

CLOSER 2017 - 7th International Conference on Cloud Computing and Services Science

640

Table 2: Synthesis of the related works about easy deploy-

ment to multiple PaaS providers

Studied Solutions

Criteria

Declarative

Automatic

Multiple PaaS

Cloud4SOA + +/− +

(Kamateri et al., 2013; D’Andria et al., 2012)

PaaSHopper − + +

(Walraven et al., 2015; Paraiso et al., 2012)

ANEKA (Wei et al., 2011) − − +

COAPS API (Sellami et al., 2013) + + +

SeaClouds (Athanasopoulos et al., ) + +/− +

Nucleus − + +

(Kolb and R

ock, 2016; R

ock and Kolb, 2016)

PaaS Manager (Cunha et al., 2014) − + −

Spinnaker + + +

cations and SeaClouds for TOSCA speciﬁcations.

PaaSHopper, Nucleus and PaaS Manager are de-

signed for supporting multi-PaaS and they meet the

automatic deployment through an API and they do

not provide any declarative model and/or provisioning

operations. Hence, hey do not meet the ﬁrst require-

ment. COAPS API and Spinnaker provide an API for

application deployment across several PaaS offerings.

Both works meets requirements related to the need for

declarative model and automatic deployment.

5 MIGRATING AN APPLICATION

FROM ONE PaaS TO ANOTHER

Applications migration consists in moving already

deployed applications from one PaaS provider to an-

other. It is done in two main steps: (1) moving the

application data and (2) moving and re-deploying the

application itself to the new PaaS provider. In the lit-

erature, several studies on cloud migration have been

achieved. The majority of existing cloud migration

studies are focusing on the provisioning or on the mi-

gration of legacy software to the Cloud (e.g. MODA-

Clouds (Ardagna et al., 2012), CloudMIG (Frey and

Hasselbring, 2010),etc.). None of these are actually

dealing with the migration between PaaS providers.

In the rest of this section, we present few works deal-

ing with the PaaS level.

Beslic et al. (Beslic et al., 2013) propose a Model-

Driven Architecture and Refactoring based approach

to migrate an application between PaaS providers.

Their contribution includes three steps: the vendor

discovery, the application transformation, and the de-

ployment. However, authors remain only to a high-

level concept of the migration process and do not give

any details about the implementation level of their ap-

proach.

Cloud Motion Framework (CMotion) (Binz et al.,

2011) ensures applications migration into and be-

tween clouds. It enables to reduce the vendor lock-

in for applications already hosted in a given cloud.

Based on the application’s topology, authors show

how adapt existing software to use different PaaS on

the Cloud services. Each of the components is mi-

grated while taking into account the relations and de-

pendencies of the components to the others.

ConPaaS (Pierre and Stratan, 2012) is a run-time

environment for elastic applications hosting and de-

ployment on multiple PaaS providers. The services

are offered through abstractions that hide the real

implementations and can be instantiated on multi-

ple clouds. Applications in ConPaaS are composed

of services programmed through a common generic

Python interface for the service management and

a Javascript part to extend the front-end GUI with

service-speciﬁc information and control.

Migration Assessment Tool (MAT) (Sharma et al.,

2013) enables to assess a software application for mi-

gration to different popular cloud platforms. This ap-

proach is based on (1) analyzing the dependencies of

the source code of an application taking into account

their technical services and capabilities and (2) eval-

uating the support for those services and capabilities

in the target PaaS platform(s). These capabilities and

services are gathered in a set of repositories.

Other existing works use the containerization and

virtualization technologies (e.g. Docker, Linux-

VServer, OpenVZ, LXC, and Rock) to enable dis-

tributed applications deployment across several PaaS

(Dua et al., 2014) (Kolb et al., 2015). These ap-

proaches consists on packaging the applications com-

ponents dynamically in a generated service contain-

ers. The generated service containers implement the

applications components requirements. Then, the

containers are pushed to the cloud as standalone ap-

plications. Today, various PaaS solutions are based on

the container technology and they mainly use Docker

coupled with Kubernates. For instance, we can cite

Flynn, DEIS, AppFog solutions. Applications in

these PaaS solutions are based on the same technolo-

gies and standards. Hence, the migration process will

be easier.

Synthesis

In this section, we elaborate a synthesis about the

discussed works above. Doing so, we ﬁx our crite-

ria based on the classiﬁcation of Andrikopoulos et al.

(Andrikopoulos et al., 2013) that identify four possi-

Applications Deployment in Multiple PaaS Environments: Requirements, Challenges and Solutions

641

ble types of applications migration. Indeed, we will

check what is the type of application migration sup-

ported by the considered approaches. First, we have

the type I that consists on replacing the application

components with new PaaS services. Then, we have

the type II which represents the case of partially mi-

gration. Indeed, some of the application components

are migrated to another PaaS provider. Afterward, we

have the type III that consists on migrating the whole

software stack of the application to a PaaS provider.

This is the classic example of migration to the Cloud,

where for example the application is encapsulated in

VMs and ran on the Cloud. Finally, we have the

type IV which is a complete migration of the appli-

cation. The application functionality is implemented

as a composition of services running on the Cloud.

Based on this classiﬁcation, we showcase in Table 3

the migration methods mentioned above and we asso-

ciate each one to the appropriate migration type.

Table 3: Synthesis of the related works about migrating an

application from one PaaS to another.

Studied Solutions

Criteria

Type I

Type II

Type III

Type IV

Beslic et al. (Beslic et al., 2013) − + + −

CMotion (Binz et al., 2011) − + + −

ConPaaS (Pierre and Stratan, 2012) + − + −

MAT (Sharma et al., 2013) − + + −

Dua et al. (Dua et al., 2014) − + + +

org et al.(Kolb et al., 2015) − + + +

To sum up, apart from ConPaaS, all the studied

works ensure the Types II and III of application mi-

gration in PaaS levels. However, only ConPaaS take

into account the Type I since they can replace com-

ponent by Cloud services. Indeed, it proposes six ser-

vices for hosting the applications (web and database).

Whereas the works published in (Dua et al., 2014)

and (Kolb et al., 2015) are considered of Type IV

since they ensure the migration of the whole appli-

cation functionality by using the containerization.

6 CONCLUSION

In this paper, we have deﬁned and classiﬁed require-

ments of application deployment in multiple PaaS

providers through two possible scenarios. We have

identiﬁed four requirements (i.e. R

,andR

Regarding each requirement, we have analyzed the

current state of the art and stressed the strenghts and

the limitations of each solution. As a result, we sum-

marized the different solutions in three tables. In fact,

we have deﬁned for each table a set of criteria taking

into account the existing lock-in in the PaaS market

and developers expectations (i.e. to easily and seam-

lessly deploy and manage their applications in multi-

ple PaaS providers).

We are currently working to address part of these

requirements. Indeed, we are focusing on proposing

an automatic and declarative solution (1) to discover

and to select services, and (2) to deploy applications

in multiple PaaS providers.

ACKNOWLEDGMENT

This work has been funded by the Belgian research

project MoDePaaS identiﬁed by the convention num-

ber 1610385.

REFERENCES

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

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

ronment. Computing, 95(6):493–535.

Ang Li, Xiaowei Yang, S. K. M. Z. (2010). Cloud-

cmp: Shopping for a cloud made easy. In 2nd

USENIX Workshop on Hot Topics in Cloud Comput-

ing. USENIX.

Ardagna, D. et al. (2012). Modaclouds: A model-driven

approach for the design and execution of applications

on multiple clouds. In Proceedings of the 4th interna-

tional workshop on modeling in software engineering,

pages 50–56.

Athanasopoulos, D. et al. Seaclouds: Agile management of

complex applications across multiple heterogeneous

clouds. In STAF Projects Showcase, pages 54–61.

Beslic, A., Bendraou, R., Sopenal, J., and Rigolet, J. (2013).

Towards a solution avoiding vendor lock-in to enable

migration between cloud platforms. In Proceedings

of the 2nd International Workshop on Model-Driven

Engineering for High Performance and CLoud com-

puting co-located with 16th International Conference

on Model Driven Engineering Languages and Systems

MODELS’13, Miami, Florida, USA, September 29,

pages 5–14.

Binz, T., Leymann, F., and Schumm, D. (2011). Cmo-

tion: A framework for migration of applications into

and between clouds. In IEEE International Confer-

ence on Service-Oriented Computing and Applica-

tions, SOCA’11, Irvine, CA, USA, December 12-14,

pages 1–4.

Cunha, D., Neves, P., and de Sousa, P. N. M. (2014). Paas

manager: A platform-as-a-service aggregation frame-

work. Comput. Sci. Inf. Syst., 11(4):1209–1228.

D’Andria, F., Bocconi, S., Cruz, J. G., Ahtes, J., and

Zeginis, D. (2012). Cloud4soa: Multi-cloud appli-

cation management across paas offerings. In 14th

International Symposium on Symbolic and Numeric

Algorithms for Scientiﬁc Computing, SYNASC’12,

CLOSER 2017 - 7th International Conference on Cloud Computing and Services Science

642

Timisoara, Romania, September 26-29, pages 407–

414.

Dua, R., Raja, A. R., and Kakadia, D. (2014). Virtualization

vs containerization to support paas. In The IEEE Inter-

national Conference on Cloud Engineering, IC2E’14,

pages 610–614.

Frey, S. and Hasselbring, W. (2010). Model-based

migration of legacy software systems to scalable

and resource-efﬁcient cloud-based applications: The

cloudmig approach.

Garg, S. K., Versteeg, S., and Buyya, R. (2011). Smicloud:

A framework for comparing and ranking cloud ser-

vices. In IEEE 4th International Conference on Util-

ity and Cloud Computing, UCC’11, Melbourne, Aus-

tralia, December 5-8, pages 210–218.

Han, S.-M., Hassan, M. M., Yoon, C.-W., and Huh, E.-N.

(2009). Efﬁcient service recommendation system for

cloud computing market. In Proceedings of the 2Nd

International Conference on Interaction Sciences: In-

formation Technology, Culture and Human, ICIS ’09,

pages 839–845. ACM.

Kamateri, E. et al. (2013). Cloud4soa: A semantic-

interoperability paas solution for multi-cloud platform

management and portability. In Service-Oriented and

Cloud Computing - Second European Conference, ES-

OCC, M

alaga, Spain, September 11-13, pages 64–78.

Kang, J. and Sim, K. M. (2011). Towards agents and ontol-

ogy for cloud service discovery. In 2011 International

Conference on Cyber-Enabled Distributed Comput-

ing and Knowledge Discovery, CyberC’11, Beijing,

China, October 10-12, pages 483–490.

Kang, J. and Sim, K. M. (2016). Ontology-enhanced agent-

based cloud service discovery. IJCC, 5(1/2):144–171.

Kolb, S., Lenhard, J., and Wirtz, G. (2015). Applica-

tion migration effort in the cloud - the case of cloud

platforms. In 8th IEEE International Conference on

Cloud Computing, CLOUD’15, New York City, NY,

USA, June 27 - July 2, pages 41–48.

Kolb, S. and R

ock, C. (2016). Uniﬁed cloud application

management. In IEEE World Congress on Services,

SERVICES’16, San Francisco, CA, USA, June 27 -

July 2, pages 1–8.

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

Cloudcmp: comparing public cloud providers. In Pro-

ceedings of the 10th ACM SIGCOMM Internet Mea-

surement Conference, IMC’10, Melbourne, Australia

- November 1-3, pages 1–14.

Mancini, E. P., Rak, M., and Villano, U. (2009). Perf-

cloud: GRID services for performance-oriented de-

velopment of cloud computing applications. In 18th

IEEE International Workshops on Enabling Technolo-

gies: Infrastructures for Collaborative Enterprises,

WETICE’09, Groningen, The Netherlands, 29 June -

1 July, pages 201–206.

Mell, P. and Grance, T. (2009). The NIST deﬁnition of

cloud computing. National Institute of Standards and

Technology, 53(6):50.

Paraiso, F., Haderer, N., Merle, P., Rouvoy, R., and Sein-

turier, L. (2012). A federated multi-cloud paas infras-

tructure. In 2012 IEEE Fifth International Conference

on Cloud Computing, Honolulu, HI, USA, June 24-29,

pages 392–399.

Pierre, G. and Stratan, C. (2012). Conpaas: a platform

for hosting elastic cloud applications. IEEE Internet

Computing, 16(5):88–92.

Qu, L., Wang, Y., and Orgun, M. A. (2013). Cloud ser-

vice selection based on the aggregation of user feed-

back and quantitative performance assessment. In The

IEEE International Conference on Services Comput-

ing, Santa Clara, CA, USA, June 28 - July 3, pages

152–159.

Quinton, C., Romero, D., and Duchien, L. (2014). Auto-

mated selection and conﬁguration of cloud environ-

ments using software product lines principles. In The

IEEE 7th International Conference on Cloud Comput-

ing, Anchorage, AK, USA, June 27 - July 2, pages 144–

151.

Quinton, C., Romero, D., and Duchien, L. (2016). SA-

LOON: a platform for selecting and conﬁguring cloud

environments. Softw., Pract. Exper., 46(1):55–78.

Redl, C., Breskovic, I., Brandic, I., and Dustdar, S.

(2012). Automatic SLA matching and provider se-

lection in grid and cloud computing markets. In 13th

ACM/IEEE International Conference on Grid Com-

puting, GRID’12, Beijing, China, September 20-23,

pages 85–94.

ock, C. and Kolb, S. (2016). Nucleus – uniﬁed deployment

and management for platform as a service. Technical

report, University of Bamberg.

Sellami, M., Yangui, S., Mohamed, M., and Tata, S. (2013).

Paas-independent provisioning and management of

applications in the cloud. In 2013 IEEE Sixth Interna-

tional Conference on Cloud Computing, Santa Clara,

CA, USA, June 28 - July 3, pages 693–700.

Sellami, R., Vedrine, M., Bhiri, S., and Defude, B. (2015).

Automating resources discovery for multiple data

stores cloud applications. In CLOSER’15 - Proceed-

ings of the 5th International Conference on Cloud

Computing and Services Science, Lisbon, Portugal,

20-22 May, pages 397–405.

Sharma, V. S., Sengupta, S., and Nagasamudram, S. (2013).

Mat: A migration assessment toolkit for paas clouds.

In The IEEE Sixth International Conference on Cloud

Computing, Santa Clara, CA, USA, June 28 - July 3,

pages 794–801.

Spinnaker (2017). Spinnaker website.

http://www.spinnaker.io/.

Walraven, S., Landuyt, D. V., Raﬁque, A., Lagaisse, B., and

Joosen, W. (2015). Paashopper: Policy-driven middle-

ware for multi-paas environments. J. Internet Services

and Applications, 6(1):2:1–2:14.

Wei, Y., Sukumar, K., Vecchiola, C., Karunamoorthy, D.,

and Buyya, R. (2011). Aneka cloud application plat-

form and its integration with windows azure. CoRR,

abs/1103.2590.

Wittern, E., Kuhlenkamp, J., and Menzel, M. (2012).

Cloud service selection based on variability modeling.

In Service-Oriented Computing - 10th International

Conference, ICSOC’12, Shanghai, China, November

12-15, pages 127–141.

Applications Deployment in Multiple PaaS Environments: Requirements, Challenges and Solutions

643