=Paper=
{{Paper
|id=Vol-1400/paper12
|storemode=property
|title=SeaClouds: Agile management of complex applications across multiple heterogeneous clouds
|pdfUrl=https://ceur-ws.org/Vol-1400/paper12.pdf
|volume=Vol-1400
|dblpUrl=https://dblp.org/rec/conf/staf/AthanasopoulosB15
}}
==SeaClouds: Agile management of complex applications across multiple heterogeneous clouds==
https://ceur-ws.org/Vol-1400/paper12.pdf
SeaClouds: Agile management of complex applications
across multiple heterogeneous clouds
D. Athanasopoulos(1), M. Barrientos(2), L. Bartoloni(3), A. Brogi(3),
M. Buccarella(3), J. Carrasco(2), J. Cubo(2), F. D'Andria(4),
E. Di Nitto(1), A. Nieto(2), M. Oriol(3), E. Pimentel(2), S. Zenzaro(3)
(1) Dipartimento di Elettronica, Informazione e Bioingegneria,
Politecnico di Milano, Italy
(2) Departamento de Lenguajes y Ciencias de la Computación,
Universidad de Málaga, Spain
(3) Department of Computer Science, University of Pisa, Italy
(4) ATOS, Spain
Abstract. SeaClouds is a European FP7 research project, whose goal
is to develop a novel open solution to provide developers with the ca-
pabilities to design, configure, deploy, and manage complex applica-
tions across multiple heterogeneous clouds in an efficient and adaptive
way.
1 General data of the project
Name Seamless adaptive multi-cloud management of service-based applications
Acronym SeaClouds
Project Id EU-FP7-ICT-610531
Funding source European Union
Duration From October 1, 2013 to March 31, 2016
Total funding Budget: 2.99 MEuros - EU contribution: 2.19 MEuros
Consortium ATOS (Spain), University of Malaga (Spain), University of Pisa (Italy),
Politecnico di Milano (Italy), Cloudsoft (UK), Nurogames (Germany)
Web site http://www.seaclouds-project.eu/
2 Brief description of the project
How to deploy and manage, in an efficient and adaptive way, complex applications
across multiple heterogeneous cloud platforms is one of the problems that have
emerged with the cloud revolution. SeaClouds is a European FP7 research project
which aims at enabling a seamless adaptive multi-cloud management of complex
applications by supporting the distribution, monitoring and migration of application
modules over multiple heterogeneous cloud platforms. In short:
SeaClouds is a software platform which supports the “design, development,
planning and management” of complex business applications distributed on
multi-cloud environments.
SeaClouds provides an enterprise capability with continuous software delivery
that enables independent application vendors to mitigate risks and reduce time
and cost to market.
SeaClouds orchestrates services, platforms and infrastructures to ensure that
they dynamically meet the needs of cloud applications.
SeaClouds provides an integrated standards-based multi-cloud application
management system that follows the DevOps approach with various basic ca-
pabilities delivered to the developer via an innovative Graphical User Interface
(GUI).
3 Objectives of the project
The SeaClouds project aims to develop a new open source platform featuring a seam-
less adaptive multi-cloud management of service-based applications. The platform
consists of an application management system over IaaS and PaaS clouds, and it im-
plements a DevOps approach for continuous software delivery.
SeaClouds permits developers to design, deploy, manage and configure complex
applications across multiple and heterogeneous clouds, something unfeasible hitherto.
The specific objectives of SeaClouds are:
Orchestration and adaptation of services distributed over different cloud
providers. SeaClouds aims at providing the assisted design, synthesis, and
simulation of service orchestrations on different cloud providers, by distrib-
uting modules of cloud-based applications over multiple heterogeneous clouds.
Unified application management of services distributed over different clouds.
SeaClouds will be able to deploy, manage, scale and monitor services over
technologically diverse cloud providers. Such operations will be performed by
taking into account application requirements and by providing developers with
support beyond the handling of single services.
Monitoring and run-time reconfiguration operations of services distributed
over multiple heterogeneous cloud providers. Monitoring will be in charge of
detecting the possible need of redistributing services across cloud providers.
Dynamic reconfiguration will let orchestrations evolve so as to realize all the
required changes. Reconfiguration ranges from dynamically replacing mal-
functioning services to migrating them to different cloud providers.
Compliance with major standards for cloud interoperability. SeaClouds will
manage applications deployed on technologically diverse cloud platforms, uni-
fying operations such as monitoring and lifecycle management, promoting the
adoption of OASIS standards for cloud interoperability, in particular TOSCA 1
and CAMP2.
1 https://www.oasis-open.org/committees/tosca/
2 https://www.oasis-open.org/committees/camp/
2
4 Final expected outcomes of the project
In order to achieve the aforementioned objectives, the following capabilities are ex-
pected as outcomes of SeaClouds:
Matchmaking. To allow querying or browsing available (IaaS and PaaS) cloud
offerings and select them based on application requirements.
Deployment optimizer. To optimize the deployment topology of an application
across multiple clouds to address non-functional requirements.
Application management. To support efficient deployment and multi-cloud
governance of complex applications on various (IaaS and PaaS) cloud offer-
ings by leveraging cloud harmonized APIs and platform-specific adapters.
Monitoring and SLA enforcement. To feature monitoring and independent
metrics to allow users to monitor the health and performance of applications.
Repairing. To scale both horizontally and vertically cloud resources to maxim-
ize the performance of each module of an application.
Replanning and application migration. To provide a seamless migration of
different application modules between dissimilar clouds.
Database migration and data synchronization. To enable the portability of
data among databases on different clouds.
5 Current status of the project
We now describe the platform architecture, SeaClouds main functionalities, the case
studies used to validate the platform, and relations and collaborations with other pro-
jects, initiatives and standards.
5.1 Platform architecture
Dashboard / SeaClouds API
Business SLA info
Abstract Deployable Confirmed Deployable Agreement
Application Application Application Model generation
Model Model
Trigger
SLA Service alerts
Deployer Subscription
Planner
to rules
Monitoring
Repairing setup
Replanning actions
Cloud
Providers request Monitor
Deploys
Discoverer
Monitoring
Application Application event
module module
Data Data
Collector Collector
Figure 1. SeaClouds architecture.
3
Figure 1 illustrates the architecture and main components of SeaClouds. On top of the
figure, there is the Dashboard / SeaClouds API, which provides a harmonized API
and a dashboard to manage cloud applications with SeaClouds. This component or-
chestrates the different modules of the SeaClouds platform:
SeaClouds Discoverer. It is in charge of discovering available capabilities and
add-ons offered by available cloud providers.
SeaClouds Planner. It is in charge of generating an orchestration plan taking
into account the application topology and requirements.
SeaClouds Deployer. It is in charge of executing deployment plans generated
by the Planner.
SeaClouds Monitor. It is in charge of monitoring that the Quality of Services
(QoS) properties of the application are not violated by the clouds in which the
application modules were deployed, and of determining -together with the De-
ployer- the reconfiguration strategies to trigger repairing or replanning actions.
SeaClouds SLA Service. It is in charge of mapping low level information
gathered from the Monitor into Quality of Business (QoB) information about
the fulfilment of the SLA defined.
A distinguishing aspect of the SeaClouds architecture is that it builds on top of two
OASIS standards initiatives: TOSCA, to specify the topologies and deployment plans,
and CAMP, to interoperate with heterogeneous PaaS providers.
5.2 Design-time functionalities
The supported design time functionalities are:
Matchmaking, implemented by the Planner component,
Deployment optimizer, implemented by the Planner component,
Discovery, implemented by the Discoverer component.
At design time, the designer specifies the Abstract Application Model (AAM),
which is a TOSCA YAML model that defines the topology and requirements of the
application to be deployed. Once the AAM has been specified, SeaClouds starts the
Planning stage. The Planner acts with two sub-processes: Matchmaking and Optimiz-
er (as described in D3.13).
The matchmaking aims at finding, for each deployable module in the AAM, the
list of candidate cloud offerings that fulfil the requirements for that module. To gen-
erate such list of cloud offerings, the planner interacts with the Discoverer. Then, the
Optimizer selects one cloud offering from the list of candidates using metaheuristic
algorithms. As a result, the Planner generates a Deployable Application Model
(DAM), which specifies the concrete cloud services used to distribute the application.
3 SeaClouds Deliverable D3.1 - Discovery, design and orchestration functionalities: First speci-
fication and prototype. Available from http://www.seaclouds-project.eu/deliverables.html
4
5.3 Run-time functionalities
The supported run-time functionalities are:
Application management, implemented by the Deployer component.
Monitoring and SLA enforcement, jointly implemented by the Deployer,
Monitoring and SLA Service components.
Repairing, jointly implemented by the Deployer and Monitor components.
Replanning and application migration, jointly implemented by the Deployer,
Monitor, and Planner components.
Database migration and data synchronization, integrated from the
MODAClouds EU project4 and is integrated with the other elements of Sea-
Clouds so that they can activate it.
The Deployer component, detailed in Deliverable D4.15, executes the confirmed
DAM, while the Monitor is configured with the monitoring rules taken from the user
requirements. Also, the SLA service subscribes to rules or alerts and in connection
with the Monitor, it enforces the policies of the agreements. The Deployer permits the
deployment of the application’s modules over heterogeneous IaaS and PaaS, and it
tracks the dynamic evolution of the deployment and management of the application
modules themselves, by means of a Live Application Model (LAM), as described in
Deliverable D4.1.
Once the application is deployed, the Deployer manages it and instruments the
Monitor. In particular, the Deployer installs DataCollectors in the cloud machine(s) in
which the application modules have been deployed. A DataCollector component
gathers raw monitoring data and pushes them to the Monitor. The latter component
interacts with the SLA service to manage possible violations of QoS and QoB proper-
ties.
When a violation issue occurs, and it can be fixed without replanning, the Monitor
and the Deployer interact to repair the issue. Otherwise, the Monitor interacts with the
Planner to trigger the generation of a new plan to be executed by the Deployer.
5.4 Case studies
5.4.1 ATOS case study
The ATOS case study is a complex software application that aims at developing an
innovative and integrated solution for the use of social inclusion tools by elderly peo-
ple and for the general management (self-management included) of their medical
problems. The currently existing application will be decomposed into a set of modules
that will be deployed, monitored and administered in one or more cloud (IaaS and
PaaS) cloud providers with the support featured by the SeaClouds platform.
From the re-engineering our health application, ready to be ported on the cloud
through the SeaClouds platform, we expect three main benefits:
4 http://www.modaclouds.eu
5 SeaClouds Deliverable D4.1 - Definition of the multi-deployment and monitoring strategies.
Available from http://www.seaclouds-project.eu/deliverables.html
5
To reduce the vendor lock-in problem, by having the opportunity to choose
and change competitive providers according to the changing resource require-
ments during the lifetime of the application,
To execute the application on fully managed cloud infrastructure running in
enterprise-class data centers with secure, reliable cloud services to support and
deliver hosted software solutions, and
To deliver flexible resources to accommodate fluctuations in customers’ pro-
duction workloads as well as on-demand capacity for training, demonstration,
proof-of-concept or test/development.
5.4.2 NURO case study
NURO cloud game case study is based on a Nurogames Engine, which is responsible
for data consistency and cheating protection. NURO will modify its monolithic server
approach to get a cloud ready version.
The goal of the case study is to keep the functionalities of the currently available
version, by taking advantage of cloud flexibility and of SeaClouds tools to optimize
and to save administrative time and effort. Reliability, stability and security of the
game server application should be improved by the SeaClouds solution. Saving re-
sources in burst scenarios and a flexible pricing of real used resources will be possi-
ble.
The main benefits that we expect from porting NURO application to the cloud via
SeaClouds are:
Stable QoS. The flexible use of resources should guarantee the QoS on pick
phases. Maintenance time should be optimized on rollout of new versions.
Simpler administration and maintenance. Designers and administrators can fo-
cus on the application. They just have to configure the modules according to
their needs and invoke deployment, installation and updates of the application.
Administration of hardware and operating system is to be done by the Sea-
Clouds system or the cloud provider.
Scalability over time. If the game is accepted and running for a long time, in-
creasing number of players will cause the server applications to consume more
resources. A cloud solution should make it easy to scale the system to the us-
ers’ needs. Moreover, on pick phases the QoS limits will be met and on burst
phases wasting of resources will be avoided.
Flexibility and control on payment. SeaClouds could considerably reduce
costs, only paying for resources which are actually needed. Consuming and re-
leasing resources depends on the needs to satisfy the QoS rules. We expect to
set also cost limits to avoid exceeding budgets.
6
5.5 Relations and collaborations with other projects and standards
As discussed earlier, SeaClouds aims at covering different needs of application
deployment and management over multiple IaaS and PaaS cloud providers.
As far as IaaS environments are concerned, the SeaClouds platform takes into ac-
count the main IaaS APIs and technologies, such as OpenStack 6 and CloudStack7. To
solve the problem of deploying multi-cloud in IaaS environments, SeaClouds is using
Apache jclouds8 via Apache Brooklyn9. Recently, SeaClouds members have been
contributing on adding the support for Docker and Microsoft Azure 10 compute to
Apache jclouds.
As far as PaaS environments are concerned, cloud developers are often locked into
a specific cloud environment. To mitigate the vendor lock-in issues, several research
projects try to wrap services offered by different providers in order to make them
compatible. One of the most related projects in this regard is Cloud4SOA11, which
allows the management of PaaS services of different providers. Also, Cloud4SOA
provides a set of cloud-independent monitoring mechanisms and a matchmaking
methodology. SeaClouds takes advantage of Cloud4SOA and moves beyond it by
defining its own matchmaking mechanisms for selecting target services that better
adapt to the application components.
Another related project is CloudWave12. With CloudWave, service providers will
be able to rapidly design and deliver innovative, sustainable digital services for con-
sumers at low cost and high quality. SeaClouds converges with CloudWave in the
used distributed algorithms and data models that enable cloud infrastructures and
applications to take actions in response to the dynamic changes in their environment.
On the other hand, SeaClouds focuses on tackling the problem of deploying and man-
aging complex multi-services applications over technologically heterogeneous clouds
environments. It is worth mentioning SeaClouds and CloudWave organize a joint
scientific workshop at ESOCC 201513.
Also, SeaClouds relates to MODAClouds14, which aims at providing quality as-
surance during the application’s life-cycle, mitigating part of the portability vendor
lock-in issues. To achieve its purposes, MODAClouds offers a monitoring methodol-
ogy independent of any provider API. This monitoring methodology is going to be
extended by SeaClouds project.
The peculiarity of SeaClouds is in the development of an almost market-ready
product, adopting and extending Brooklyn (whose main developers are part of the
6 https://www.openstack.org
7 https://cloudstack.apache.org
8 https://jclouds.apache.org
9 https://brooklyn.incubator.apache.org
10 http://azure.microsoft.com
11 http://www.cloud4soa.com
12 http://cloudwave-fp7.eu
13 http://esocc2015.unime.it
14 http://www.modaclouds.eu
7
project) and employing OASIS TOSCA15 and CAMP16, as main standards for the
deployment and management of cloud applications. SeaClouds members have been
involved with CAMP and TOSCA technical committees, suggesting improvements to
the specifications based on our activity. Also, SeaClouds members attend conference
calls being held within the TOSCA monitoring subgroup, which focuses on extending
TOSCA specification to describe monitoring aspects of an application.
Last, but not least, a cross-fertilization effort between Alien4Cloud17 and Sea-
Clouds is currently ongoing to assess the possibility for Alien4Cloud to use Sea-
Clouds deployer, and for SeaClouds to use Alien4Cloud TOSCA YAML parser.
A more thorough discussion of related projects is provided in deliverables
D2.3.118 and D2.3.219.
6 Impact and Exploitation
SeaClouds adopted from the very beginning of the project an open source strategy,
which includes transparent development, Apache 2.0 licensing, and developer en-
gagement. This strategy also serves as enabler for three main impact paths:
Apache Brooklyn Integration and Extension for Application Deployment.
The Apache Brooklyn application management framework will not only serve
as part of SeaClouds’ Deployer component, but also as a vehicle for Sea-
Clouds’ post-project value proposition by supporting TOSCA and extending
deployment capabilities to PaaS layer and multi-cloud scenarios. The Sea-
Clouds-extended Apache Brooklyn will offer a more robust multi-cloud appli-
cation deployment for European service providers to accompany their IaaS and
PaaS service, strengthening a solution already supported in the market.
Contribution to Vendor-Driven Standards. SeaClouds is actively contrib-
uting to two leading standards in cloud computing PaaS segment: OASIS
CAMP and TOSCA. By extending these standards, SeaClouds is tapping a
channel directly to Europe’s cloud supply (vendors) to help empower multi-
cloud scenarios for demand (their customers) via improved topology, orches-
tration, management and interoperability.
Consortium Exploitation Opportunities. Project partners will be able to ex-
ploit SeaClouds’ open-source solution to strengthen their commercial and re-
search portfolios. Industry partners such as Cloudsoft and ATOS can integrate
SeaClouds assets into their own selected commercial offerings, benefiting as
primary contributors to development. Similarly, academic partners can propel
their research in the multi-cloud domain.
15 https://www.oasis-open.org/committees/tosca
16 https://www.oasis-open.org/committees/camp
17 http://alien4cloud.github.io
18 SeaClouds Deliverable D2.3.1 - Periodic Standardization report.
Available from http://www.seaclouds-project.eu/deliverables.html
19 SeaClouds Deliverable D2.3.2 - Periodic Standardization report.
Available from http://www.seaclouds-project.eu/deliverables.html
8