EU Project SeaClouds

∗

Adaptive Management of Service-based Applications Across Multiple Clouds

Antonio Brogi

, Jos

e Carrasco

, Javier Cubo

, Francesco D’Andria

Ahmad Ibrahim

, Ernesto Pimentel

and Jacopo Soldani

Department of Computer Science, University of Pisa, Pisa, Italy

Department of Computer Science, University of Malaga, Malaga, Spain

ATOS, Barcelona, Spain

Keywords:

Multi-Cloud Deployment, Monitoring, Dynamic Reconﬁguration, Service Orchestration.

Abstract:

The adaptive management of complex applications deployed across multiple heterogeneous PaaS platforms is

one of the problems that have emerged with the cloud revolution. The ongoing EU research project SeaClouds

aims at providing seamless adaptive multi-cloud management of complex applications by supporting the dis-

tribution, monitoring and migration of application modules over multiple heterogeneous PaaS platforms.

In this paper we present context, motivations and objectives of SeaClouds, its relation with other cloud initia-

tives, and its initial architecture.

1 INTRODUCTION

Current cloud technologies suffer from a lack of stan-

dardization, with different providers offering similar

resources in a different manner (M. Armbrust et al.,

2010). As a result, cloud developers are often locked

in a speciﬁc platform environment because it is practi-

cally unfeasible for them, due to high complexity and

cost, to move their applications from one platform to

another (N. Loutas et al, 2011). To take full beneﬁt

of the ﬂexibility provided by the cloud, modules of

a complex application could be deployed on multiple

clouds depending on their characteristics. Thus, in a

scenario where a complex application is distributed

across different cloud service providers, a solution is

needed to manage and orchestrate the distribution of

modules in a sound and adaptive way (Parameswaran

and Chaddha, 2012).

Consider an application consisting of three mod-

ules (A, B, C), each having its own requirements (R

, R

). Suppose that four PaaS platforms are avail-

able (α, β, γ, δ), each providing different offers (O

, O

) and based on different technologies. The

application developer can deploy the application ei-

ther on a single PaaS or over multiple PaaS platforms.

In the latter case, then she has to ﬁgure out the best

offer for each module (e.g., R

: {O

}, R

: {O

, O

: {O

}) and perform the deployment accordingly

∗

Work supported by EU-FP7-ICT-610531 SeaClouds

project (start date: October 2013, end date: March 2016).

(e.g., A → γ, B → β, C → δ). Afterwards, she has to

monitor both the single modules and the whole appli-

cation. In case some requirements are violated (e.g.,

is no longer satisﬁed by O

), she has to manu-

ally reconﬁgure the deployment (e.g., A → γ, B → γ,

C → δ). This implies that to manage multi-cloud ap-

plications, the developer must have knowledge of the

different employed PaaS platforms and has to contin-

uously monitor the application. This is a costly and

cumbersome process and it would be much better to

have some framework doing all the work automati-

cally.

Figure 1: SeaClouds logo.

In this paper, we discuss the ongoing project Sea-

Clouds (Seamless adaptive multi-cloud management

of service-based applications) which focuses on the

problem of deploying and managing complex multi-

component application over heterogeneous clouds in

an efﬁcient and adaptive way. SeaClouds works to-

wards giving organizations the capability of “Agility

After Deployment” (Figure 1) for cloud-based appli-

758

Brogi A., Carrasco J., Cubo J., D’Andria F., Ibrahim A., Pimentel E. and Soldani J..

EU Project SeaClouds - Adaptive Management of Service-based Applications Across Multiple Clouds.

DOI: 10.5220/0004979507580763

In Proceedings of the 4th International Conference on Cloud Computing and Services Science (MultiCloud-2014), pages 758-763

ISBN: 978-989-758-019-2

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

cations. Its approach is based on the concept of ser-

vice orchestration and it is designed to fulﬁll func-

tional and non-functional properties over the whole

application. Applications will be dynamically re-

conﬁgured by changing the orchestration of the ser-

vices when the monitoring detects that such proper-

ties are not respected. So, SeaClouds main objec-

tive is the development of a novel platform which per-

forms a seamless adaptive multi-cloud management

of service-based applications. More speciﬁcally, the

objectives of SeaClouds are:

) Orchestration and adaptation of services dis-

tributed over different cloud providers,

) Monitoring and run-time reconﬁguration of ser-

vices distributed over multiple heterogeneous

cloud providers,

) Providing uniﬁed application management of

services distributed over different cloud providers,

and

) Compliance with major standards for cloud in-

teroperability.

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

tion 2 will position SeaClouds with respect to other

cloud initiatives and it will describe its main chal-

lenges. Section 3 will present the SeaClouds ap-

proach along with its initial architecture. Finally, Sec-

tion 4 will draw some concluding remarks.

2 SeaClouds IN CONTEXT

This section describes how SeaClouds, with the pur-

pose of achieving its main goal, will advance the state

of the art with respect to O

, O

and O

, and will

contribute to obtain O

. Figure 2 illustrates how Sea-

Clouds intends to relate to the previously mentioned

initiatives and standards.

Orchestration and Adaptation in the Cloud. O

will be addressed by developing adaptive orchestra-

tors of cloud-based application modules. Orchestra-

tors are widely used in the service-oriented comput-

ing paradigm (A. Brogi, J. Camara, C. Canal, J, Cubo

and E. Pimentel, 2007; C. Canal and Salaun, 2008;

J. C

amara, J.A. Mart

ın, G. Sala

un, J. Cubo, M. Oued-

erni, C. Canal and E. Pimentel, 2009; H. Nezhad, G.Y.

Xu and B. Benatallah, 2010; J. Cubo and E. Pimentel,

2011), mainly focusing on behavioural and context-

aware adaptation of services, by coordinating the in-

teractions between different services. Several ap-

proaches exist that target formal adaptation of orches-

trated services, but, to the best of our knowledge, only

few (middleware-based) attempts have been proposed

to face challenges such as heterogeneity of cloud plat-

forms and migration to different cloud providers (e.g.,

(Guill

en et al., 2013)).

Challenges in Orchestration and Adaptation for

the Cloud. SeaClouds will address the following chal-

lenges in order to extend service-oriented approaches

to the cloud: (i) Adaptation contracts need to take

into account cloud provider characteristics and Ser-

vice Level Agreements (SLA), (ii) Violations of Qual-

ity of Service (QoS) properties need to be monitored

across different cloud platforms, and (iii) Dynamic

architecture reconﬁguration might involve migrating

some components of the application to other cloud

providers at runtime. The latter two challenges are

discussed further in the following sections.

Monitoring of Multi-cloud Services. The EU

FP7 Cloud4SOA project (www.cloud4soa.eu) pro-

vides an open source interoperable framework for ap-

plication developers and PaaS providers. Cloud4SOA

aims at supporting developers in deploying and mon-

itoring their application with the ultimate objective of

reducing the risk of vendor lock-in. The monitoring

is based on uniﬁed metrics, but Cloud4SOA monitors

each application separately and it is not able to aggre-

gate monitoring results of multi-component applica-

tions.

Several Commercial and Open Source Initia-

tives Target Monitoring of Cloud Applications. Of-

ten these initiatives address only particular platforms,

for example Appsecute (www.appsecute.com) mon-

itors only (open-source) CloudFoundry-based plat-

forms. More platform-independent technologies are

available for the IaaS level, since the latter has

undergone a stronger harmonization effort. Delta-

cloud (deltacloud.apache.org) encapsulates the native

API cloud provider to enable management of re-

sources in different IaaS platforms, such as Ama-

zon EC2. Rightscale (www.rightscale.com) supports

monitoring several public (e.g., Amazon Web Ser-

vices, Google Cloud Platform, Rackspace, Windows

Azure) and private IaaS clouds (e.g., CloudStack, Eu-

calyptus, OpenStack). Truly platform-independent

monitoring solutions exist, the most known being

NewRelic (www.newrelic.com). NewRelic achieves

platform-independency by requiring each provider to

implement a monitoring component and to integrate

it in the offered cloud platform. On the one hand,

this approach yields the best results from a monitoring

point of view. On the other hand, it forces providers

to invest quite some resources in order to implement

the monitoring.

EUProjectSeaClouds-AdaptiveManagementofService-basedApplicationsAcrossMultipleClouds

759

Figure 2: Positioning SeaClouds.

Challenges in Monitoring of Services on Multiple

Clouds. In order to address O

, SeaClouds monitor-

ing will use and enhance existing monitoring func-

tionalities for the PaaS and IaaS levels: (i) With re-

spect to the IaaS level, SeaClouds will simply reuse

what is available (e.g., Deltacloud), and (ii) With

respect to the PaaS level, SeaClouds aims at aug-

menting the set of metrics currently available from

Cloud4SOA (response time and up-time). For both

the IaaS as the PaaS level, SeaClouds aims at coor-

dinating, monitoring and aggregating monitoring in-

formation at the single service level to fulﬁll the pur-

poses of whole orchestration. Thus, SeaClouds aims

at: (i) monitoring each application component, and

(ii) combining and aggregating the above mentioned

data to highlight possible problems and their impact.

Uniﬁed Management of Multi-cloud Applica-

tions. Brooklyn (www.brooklyn.io) is an open

source, policy-driven control plane for distributed ap-

plications delivered by CloudSoft (a member of the

SeaClouds consortium). It enables single-click de-

ployment of applications across machines, locations

and clouds. Then it continuously optimizes running

applications to ensure ongoing compliance with poli-

cies. Brooklyn uses two open source tools to operate

on cloud resources, Apache Whirr (whirr.apache.org)

and jClouds (www.jclouds.org), that support several

IaaS providers. The already mentioned Cloud4SOA

project also offers deployment and lifecycle manage-

ment functionality using a harmonized API layer to

encapsulate the providers’ APIs. Furthermore, the

previously mentioned initiatives (RightScale, Delta-

cloud for the IaaS level, and Appsecute, Cloud-

Foundry for the PaaS level) all offer management

functionality, with the same limitations as discussed

while explaining monitoring.

Challenges in Uniﬁed Application Management of

Services Distributed over Different Cloud Providers.

SeaClouds will use existing management function-

alities. Speciﬁcally, (i) SeaClouds discovery func-

tionality may use and extend existing matchmaking

functionalities to match application requirements with

CLOSER2014-4thInternationalConferenceonCloudComputingandServicesScience

760

PaaS offerings, and (ii) SeaClouds management will

use a REST harmonized API for the deployment,

management and monitoring of simple cloud-based

applications across different and heterogeneous cloud

PaaS offerings. SeaClouds intends to use Brook-

lyn’s policy-driven functionality to integrate support

for IaaS providers. Moreover, Brooklyn’s approach to

policy modeling and enforcing can provide guidance

for SeaClouds orchestration/adaptation and manage-

ment functionality. On the other hand, Brooklyn only

targets the IaaS level and provides no support for or-

chestration. Beyond what Brooklyn provides, Sea-

Clouds will therefore extend policy-driven function-

ality to the PaaS level and also add support for adap-

tation and orchestration. Due to a common partner in

both initiatives (CloudSoft), Brooklyn can also bene-

ﬁt from integrating SeaClouds functionalities, espe-

cially regarding the integration of adaptation tech-

niques in supported policies.

Standards for Cloud Interoperability. CAMP

(Cloud Application Management for Platforms) (OA-

SIS, 2012a) aims at deﬁning harmonized APIs, mod-

els, mechanisms and protocols for the self-service

management (provisioning, monitoring and control)

of applications in a PaaS, independently of the cloud

provider. However, CAMP is only a protocol speciﬁ-

cation, so it needs to be implemented by parties adopt-

ing the protocol.

The OASIS TOSCA (Topology and Orchestra-

tion Speciﬁcation for Cloud Applications) (OASIS,

2012b) aims at enabling the interoperable description

of application and infrastructure cloud services, the

relationships between parts of the service, and the op-

erational behaviour of these services, independently

from the cloud provider. By increasing service and

application portability in a vendor-neutral ecosystem,

TOSCA aims at enabling portable deployment to any

compliant cloud, smoother migration of existing ap-

plications to the cloud, as well as dynamic, multi-

cloud provider applications.

Challenges in Standards for Cloud Interoperabil-

ity. SeaClouds intends to actively contribute to the

standardization effort of CAMP (OASIS, 2012a) both

by exploiting CAMP-compliant interfaces provided

by PaaS providers, and by contributing review pro-

posals that will possibly emerge while specifying

properties of SeaClouds orchestrations, adaptations

and monitoring. SeaClouds will exploit the TOSCA

speciﬁcation to drive the design of the model for spec-

ifying cloud service orchestrations. In doing so, Sea-

Clouds might actively contribute to the standardiza-

tion effort of TOSCA, by contributing review propos-

als that will emerge while trying to devise TOSCA-

compliant instances of the SeaClouds service orches-

tration model. On the other hand, SeaClouds will also

focus on developing functionalities that are deliber-

ately out of the scope of TOSCA (OASIS, 2012b)

to solve issues about policies for the dynamic man-

agement of service orchestrations. Although the

currently available implementations of TOSCA and

CAMP do not yet support the management of com-

plex applications over multiple clouds, SeaClouds

will work towards building such management on top

of them.

3 THE SeaClouds APPROACH

SeaClouds aims at homogenizing the management

of different cloud providers and at supporting sound

and scalable orchestrations of services across them.

Moreover, systems developed with SeaClouds will in-

herently support the evolution of their constituent ser-

vices, so as to easily and dynamically cope up with

needed changes. The development, monitoring and

reconﬁguration via SeaClouds will include a uniﬁed

management service, where services can be deployed,

replicated, and administered by means of standard

and harmonized APIs.

In order to obtain its main goals (viz., the develop-

ment of the platform and the deﬁnition, implementa-

tion and validation of the use cases), SeaClouds will

build a foundation to allow “Agility After Deploy-

ment” by providing necessary tools and a framework

for Modelling, Planning and Controlling cloud appli-

cations.

Figure 3 illustrates SeaClouds initial architecture.

The Planner module will implement SeaClouds plan-

ning policy to orchestrate the multi-cloud deployment

of application modules. The Planner will take the in-

put data provided by SeaClouds users, who will spec-

ify the application modules to be deployed, as well

as desired QoS properties of the whole application

and/or QoS and technology requirements for individ-

ual application modules. The Planner will determine

the best offer for each module (by using the Discov-

ery API) and will determine how to orchestrate the

modules across the selected cloud platforms (viz., or-

chestration speciﬁcation).

The Controller module will implement the multi-

cloud deployment of application modules and the

monitoring policy. In particular, the Multi-Cloud De-

ployer component will input the orchestration spec-

iﬁcation generated by the Planner, and deploy (by

exploiting the Multi-Cloud Deployment API) the

application modules on the speciﬁed clouds. The

Monitor component will be in charge of monitoring

(by exploiting the Monitoring API) that application

EUProjectSeaClouds-AdaptiveManagementofService-basedApplicationsAcrossMultipleClouds

761

Figure 3: SeaClouds initial architecture.

(module) requirements are not violated. If the mon-

itoring component detects (QoS or technological) re-

quirement violations, then it will trigger the Analyzer

component. The latter will generate (if needed) re-

conﬁguration suggestions and pass them to the Plan-

ner in order to trigger the generation of a new orches-

tration plan. Users will exploit the Graphical User

Interface (GUI) module to interact with SeaClouds.

The GUI module will contain a Designer component

to allow users to provide SeaClouds input data (viz.,

application modules, QoS and technological require-

ments). The GUI will also contain a Dashboard com-

ponent to let users easily manage and visualize the

state of the multi-cloud deployed applications.

4 CONCLUSIONS AND

PERSPECTIVES

In this paper, we have presented context, moti-

vations and objectives of the SeaClouds project

(www.seaclouds-project.eu). The project aims at pro-

viding an open source framework to address the prob-

lem of deploying, managing and reconﬁguring com-

plex applications over multiple and heterogeneous

clouds. The SeaClouds approach works towards

achieving “Agility After Deployment” by tackling the

problem from the service orchestration perspective. A

complex application, which consists of modules and

(technological and QoS) requirements, is provided as

input to the SeaClouds planner. The latter gener-

ates an orchestration by assigning (groups of) mod-

ules to different PaaS platforms. The orchestration is

then deployed and monitored according to standard

metrics. If requirements are violated, then the Sea-

CLOSER2014-4thInternationalConferenceonCloudComputingandServicesScience

762

Clouds analyzer will generate reconﬁguration infor-

mation, which leverages the creation of a different or-

chestration of the application.

The expected results of the SeaClouds project can

be summarized as follows:

• Orchestration of Cloud Services. High-level or-

chestration speciﬁcations will be used to describe

the topology and behavior of the application dis-

tributed across different platforms.

• Monitoring Process. SeaClouds will monitor QoS

requirements to determine whether/which mod-

ules should be migrated.

• Traceability of QoS Violations. Monitoring will

have to keep track of possible requirement vio-

lations and their causes. This is needed for the

generation of reconﬁguration suggestions.

• Reconﬁguration Mechanism. SeaClouds will fea-

ture ﬂexible reconﬁguration by supporting the mi-

gration of application modules across different

PaaS platforms.

• Uniﬁed Management API for Cloud Platforms.

SeaClouds will facilitate the access and the ad-

ministration of both public and private cloud

providers by providing multi-cloud management

tools as well as offering cloud providers and con-

sumers a REST-based approach to cloud-based

application management.

• Dashboard for Administration of Services Dis-

tributed across Multiple PaaS Platforms. Sea-

Clouds will provide a graphical interface to visu-

alize and manage the current conﬁguration of the

deployed services.

• Contribute to the Development of Cloud Stan-

dards. SeaClouds will provide feedback to CAMP

by proposing new functionalities and implement-

ing a CAMP-compliant interface. It will also con-

tribute to the standardization effort of TOSCA, by

employing TOSCA as an orchestration modeling

language.

REFERENCES

A. Brogi, J. Camara, C. Canal, J, Cubo and E. Pimentel

(2007). Dynamic contextual adaptation. In Proc. of

Fifth International Workshop on the Foundations of

Coordination Languages and Software Architectures

2006 (FOCLASA’06), vol. 175, no. 2, pages 81–95.

Elsevier.

C. Canal, P. P. and Salaun, G. (2008). Model-Based Adapta-

tion of Behavioural Mismatching Components. IEEE

Transactions on Software Engineering, 34(4):546–

563.

Guill

en, J., Miranda, J., Murillo, J. M., and Canal, C.

(2013). A service-oriented framework for developing

cross cloud migratable software. Journal of Systems

and Software, 86(9):2294–2308.

H. Nezhad, G.Y. Xu and B. Benatallah (2010). Protocol-

aware matching of web service interfaces for adapter

development. In Proc. of 19th International Confer-

ence on World Wide Web 2010 (WWW’10), pages 731–

740. ACM.

J. C

amara, J.A. Mart

ın, G. Sala

un, J. Cubo, M. Ouederni,

C. Canal and E. Pimentel (2009). Itaca: An integrated

toolbox for the automatic composition and adaptation

of web services. In Proc. of International Conference

on Software Engineering 2009 (ICSE’09), pages 627–

630. IEEE Computer Society Press.

J. Cubo and E. Pimentel (2011). Damasco: A framework

for the automatic composition of component-based

and service-oriented architectures. In I. Crnkovic, V.

Gruhn, M. Book (editors), European Conference on

Software Architecture 2011 (ECSA’11), LNCS 6903,

pages 388–404. Springer-Verlag.

M. Armbrust et al. (2010). A view of cloud computing.

Commun. ACM, (4).

N. Loutas et al (2011). D1.1 Requirements Anal-

ysis Report. Cloud4SOA Project Deliverable,

http://www.cloud4soa.eu/.

OASIS (2012a). CAMP 1.0 (Cloud Application Manage-

ment for Platforms), Version 1.0. http://docs.oasis-

open.org/camp/camp-spec/v1.1/camp-spec-

v1.1.html/.

OASIS (2012b). TOSCA 1.0 (Topology and

Orchestration Speciﬁcation for Cloud Ap-

plications), Version 1.0. http://docs.oasis-

open.org/tosca/TOSCA/v1.0/TOSCA-v1.0.pdf.

Parameswaran, A. and Chaddha, A. (2012). Cloud Inter-

operability and Standardization. SETLabs Brieﬁngs -

Infosys, 7(7):19–26.

EUProjectSeaClouds-AdaptiveManagementofService-basedApplicationsAcrossMultipleClouds

763