DEFINING GENERIC ARCHITECTURE FOR CLOUD IAAS
PROVISIONING MODEL
Yuri Demchenko, Cees de Laat
System and Network Engineering Group, University of Amsterdam, Amsterdam, The Netherlands
Aleksej Mavrin
Verizon Nederland B.V., Amsterdam, The Netherlands
Keywords: Cloud computing, Infrastructure as a Service (IaaS), on-Demand Infrastructure Services Provisioning
(ISOD), Composable Services Architecture (CSA), GEANT Multidomain Bus (GEMBus).
Abstract: Infrastructure as a Service (IaaS) is one of the provisioning models for Clouds as defined in the NIST
Clouds definition. Although widely used, current IaaS implementations and solutions doesn’t have common
and well defined architecture model. The paper attempts to define a generic architecture for IaaS based on
current research by authors in developing novel architectural framework for Infrastructure Services On-
Demand (ISOD) provisioning that is originated from the telecommunication and networking area and allows
for combined network and IT resources provisioning. The paper proposes the Composable Services
Architecture (CSA) for dynamically configurable virtualised services. The proposed CSA includes such
important component as the Services Delivery Framework (CSA SDF) that defines the services provisioning
workflow and supporting infrastructure for provisioned services lifecycle management. The CSA SDF
extends existing lifecycle management frameworks with additional stages such as “Registration and
Synchronisation” and “Provisioning Session Binding” that specifically target such scenarios as the
provisioned services recovery or re-planning/migration and provide necessary mechanisms for consistent
security services provisioning as an important component of the provisioned on-demand infrastructure. The
paper also describes the GEMBus (GEANT Multidomain Bus) that is considered as a CSA middleware
platform. The presented architecture is the result of the on-going cooperative effort of the two EU projects
GEANT3 JRA3 Composable Services and GEYSERS.
1 INTRODUCTION
Cloud technologies (NIST, GFD.150) are emerging
as infrastructure services for provisioning computing
and storage resources, and expectedly they will
evolve into the general IT resources. Cloud
Computing can be considered as a natural evolution
of the Grid Computing technologies to more open
infrastructure-based services.
The growing interest and adoption of the Cloud
based service provisioning, operational and business
models are facilitated by the fact that different
groups of users, vendors, providers and operators see
new opportunities related to their own specific needs
and interests: scientist and specialist users like
simplicity of setup and elasticity of
execution/computation environment; computer
scientists and programmers see a new environment
for using and developing new programming models;
providers, and in particular telecommunication
providers, and operators expect to benefit from new
provisioning models and new market for complex
infrastructure services.
The current Cloud services implement 3 basic
provisioning models (as defined in the NIST
document (NIST)): Infrastructure as a Service
(IaaS), Platform as a Service (PaaS), and Software
as a Service (SaaS). IaaS suggests involving
different types of resources, of which one is the
network connectivity with guaranteed Quality of
Service (QoS), and availability of user controlled
infrastructure operation via dynamically created
control and management planes/functions. Most of
currently available Cloud providers are positioned as
79
Demchenko Y., de Laat C. and Mavrin A..
DEFINING GENERIC ARCHITECTURE FOR CLOUD IAAS PROVISIONING MODEL.
DOI: 10.5220/0003392600790085
In Proceedings of the 1st International Conference on Cloud Computing and Services Science (CLOSER-2011), pages 79-85
ISBN: 978-989-8425-52-2
Copyright
c
2011 SCITEPRESS (Science and Technology Publications, Lda.)
PaaS; although some of them may claim to do IaaS
but in practice this turns to be rather PaaS like in
case of Amazon EC2 Cloud or RackSpace Cloud as
in fact they deliver single provider services based on
their private network infrastructure.
Despite a rapid grow, Cloud technologies are
lack of well-defined architectural framework(s) and
operational models. Current industry standardisation
activity is primarily focused on functional
interoperability of deployment platforms and
components supporting basic/core usage scenarios
such as applications and Virtual Machine (VM)
images creation, deployment and management,
services request and execution management.
This paper attempts to define a generic
architecture for the Cloud Infrastructure as a Service
provisioning model which is defined as the
Composable Services Architecture (CSA). The CSA
extends virtualisation and dynamic service
provisioning concepts to complex infrastructure
services that are composed of multiple composable
component services and resources. The CSA is SOA
based and built upon industry adopted Enterprise
Service Bus (ESB).
The presented architecture is the result of the
ongoing cooperative effort of the two EU projects
GEANT3 JRA3 Composable Services (GEANT
Project) and GEYSERS (GEYSERS Project) and
currently considered to be contributed to the Open
Grid Forum (OGF) standardisation activity. Current
development is based on previous works by the
authors in the framework of the EGEE and
Phosphorus projects (EGEE Project), (Phosphorus
Project) that have been resulted in proposing the
general Complex Resource Provisioning (CRP)
model (Demchenko et al., 2009) that includes such
main stages as reservation, deployment, access, and
decommissioning.
The paper is organized as follows. Section 2
analyses the typical infrastructure for e-Science
applications that includes computing, storage,
visualisation and their connection to network
infrastructures. Section 3 presents the proposed
Composable Services Architecture and section 4
describes the proposed Services Delivery
Framework (CSA SDF) that defines the on-demand
service provisioning sequence and workflow.
Section 5 provides information about the ongoing
development of the GEMBus that is considered as a
middleware and enabling technology for the
dynamically provisioned composable services
integration. Section 6 provides general requirements
and suggestions for building consisting security and
refers to other ongoing works by the authors.
2 ON-DEMAND
INFRASTRUCTURE SERVICES
PROVISIONING
Two general use cases for on-demand infrastructure
services provisioning can be considered for defining
basic requirements to Cloud IaaS: large scientific
infrastructure and transport network infrastructure
provisioning. These use cases represent the two
different perspectives in developing infrastructure
services – users and application developers
perspective, on one side, and providers perspective,
on the other side. Users are interested in uniform and
simple access to the resource and the services that
are exposed as Cloud/Grid resources and can be
easily integrated into the scientific or business
workflow. Infrastructure providers are interested in
infrastructure resource pooling and virtualisation to
simplify their on-demand provisioning and extend
their service offering and business model to Virtual
Infrastructure provisioning.
Figure 1 illustrates the typical e-Science
infrastructure that includes Grid and Cloud based
computing and storage resources, instruments,
control and monitoring system, visualization system,
and users represented by user clients. The diagram
also reflects that there may be different types of
connecting network links: high-speed and low-speed
which both can be permanent for the project or
provisioned on-demand.
Figure 1: Components of the typical e-Science
infrastructure involving multidomain and multi-tier Grid
and Cloud resources and network infrastructure.
The figure illustrates possible different relations
between users/actors in Clouds and Grids. Grid
architecture is built around so-called Virtual
Organisations (VO) that are defined as collaborative
associations/federations of the member organisations
CLOSER 2011 - International Conference on Cloud Computing and Services Science
80
to share complex resources which are however
remaining under control and in ownership of the VO
member organisations. Grids natively supports
project oriented collaborative environment with
sharing resources committed/donated by the VO
members. Clouds are considered as a technology for
provisioning complex computing resources,
applications and increasingly infrastructure
resources and services on-demand on the pay-per-
use base.
Typically business relations between provider
and customer are expressed in the Service Level
Agreement (SLA) that defines the services provided
by the provider, including security services that are
provided as a part of the provider Cloud
environment. The offered/provided services are
uniform and cannot be modified or configured by
user what creates problems for their integration into
the existing user infrastructure or building effective
project based collaborative environment. With wider
adoption of the Cloud infrastructure services and
their integration into organisational IT infrastructure
the demand for dynamically configurable/
manageable composable services will be growing.
The solution for mentioned problems can be seen in
provisioning manageable, dynamically configured
services that support all stages of on-demand
infrastructure services provisioning. This problem is
being researched as a part of the GEANT3 JRA3
Composable Services activity (GEANT Project) and
GEYSERS project Logical Infrastructure
Composition Layer (LICL) definition and design
(GEYSERS Project).
3 THE COMPOSABLE SERVICES
ARCHITECTURE (CSA)
The Infrastructure as a Service provisioning involves
dynamic creation of infrastructure consisting of
different types of resources together with necessary
control and management planes, all provisioned on-
demand. The proposed CSA provides a framework
for the design and operation of the
composite/complex services provisioned on-
demand. It is based on the component services
virtualisation, which in its own turn is based on the
logical abstraction of the (physical) component
services and their dynamic composition. Composite
services may also use the Orchestration service
provisioned as a CSA infrastructure service to
operate composite service specific workflow.
Figure 2 shows the major functional components
of the proposed CSA and their interaction. The
central part of the architecture is the CSA
middleware that should ensure smooth service
operation during all stages of the composable
services lifecycle.
Composable Services Middleware (CSA-MW)
provides common interaction environment for both
(physical) component services and
complex/composite services, built with them.
Besides exchanging messages, CSA-MW also
contains/provides a set of basic/general
infrastructure services required to support reliable
and secure (composite) services delivery and
operation:
Service Lifecycle Metadata Service (MD SLC)
that stores the services metadata and in
particular the services state and the provisioning
session context.
Registry service that contains information about
all component services and dynamically created
composite services. The Registry should support
automatic services registration.
Logging service that can be also combined with
the monitoring service.
Middleware Security services that ensure secure
operation of the CSA/middleware.
Note, both logging and security services can be
also provided as component services that can be
composed with other services in a regular way.
The CSA defines also Logical Abstraction Layer
for component services and resources that is
necessary part of creating services pool and
virtualisation. Another functional layer is the
Services Composition layer that allows presentation
of the composed/composite services as regular
services to the consumer.
The Control and Management plane provides
necessary functionality for managing composed
services during their normal operation. It may
include Orchestration service to coordinate
component services operation, in a simple case it
may be standard workflow management system.
CSA defines a special adaptation layer to support
dynamically provisioned Control and Management
plane interaction with the component services which
to be included into the CSA infrastructure must
implement adaptation layer interfaces that are
capable of supporting major CSA provisioning
stages, in particular, service identification, services
configuration and metadata including security
context, and provisioning session management.
The application or service middleware layer can
be defined as an additional upper CSA architecture
layer to provide application – infrastructure interface
DEFINING GENERIC ARCHITECTURE FOR CLOUD IAAS PROVISIONING MODEL
81
ApplicationsandUserTerminals
Composition
Layer
(Reservation
SLA
Negotiation)
LogicalAbstraction LayerforComponent
ServicesandResources
Control&
Management
Plane
(Operation,
Orchestration)
ComposableServicesMiddleware
(GEMBus)
NetworkInfrastructure
Compute
Resources
Storage
Resources
ComponentServices&Resources
AdaptationLayer‐ ComponentServicesandResources
AdaptationLayer–Composite/Virtualised ServicesandResources
MDSLC
Registry Logging Security
User
Client
Datalinks/flows
Controllinks
Figure 2: Composable Service Architecture and main functional components.
with the underlying virtual infrastructure
provisioned by CSA, but this functionality is
typically addressed by the applications provider
themselves.
4 CSA SERVICE DELIVERY
FRAMEWORK
The proposed CSA Service Delivery Framework
(CSA SDF) implements the Service Delivery
Framework defined by the TeleManagement Forum
(TMF) (TMF) and extends it with additional stages
to address secure composable services operation and
integration into the heterogeneous multidomain
Cloud IaaS environment. Figure 3 illustrates the
main service provisioning stages:
Service Request (including SLA negotiation).
The SLA can describe QoS and security
requirements of the negotiated infrastructure service
along with information that facilitates authentication
of service requests from users. This stage also
includes generation of the Global Reservation ID
(GRI) that will serve as a provisioning session
identifier and will bind all other stages and related
security context.
Composition/Reservation that also includes
Reservation Session Binding with GRI what
provides support for complex reservation process in
potentially multidomain multi-provider environ-
ment. This stage may require access control and
SLA/policy enforcement.
Deployment, including services Registration
and Synchronisation. Deployment stage begins
after all component resources have been reserved
and includes distribution of the common composed
service context (including security context) and
binding the reserved resources or services to the GRI
as a common provisioning session ID. The
Registration and Synchronisation stage (that
however can be considered as optional) specifically
targets possible scenarios with the provisioned
services migration or re-planning. In a simple case,
the Registration stage binds the local resource or
hosting platform run-time process ID to the GRI as a
provisioning session ID.
Operation (including Monitoring). This is the
main operational stage of the provisioned on demand
composable services. Monitoring is an important
functionality of this stage to ensure service
availability and secure operation, including SLA
enforcement.
CLOSER 2011 - International Conference on Cloud Computing and Services Science
82
Decommissioning stage ensures that all sessions
are terminated, data are cleaned up and session
security context is recycled. Decommissioning stage
can also provide information to or initiate services
usage accounting.
The two additional (sub-)stages can be initiated
from the Operation stage and/or based on the
running composed service or component services
state, such as their availability or failure:
Re-composition or Re-planning that should
allow incremental infrastructure changes.
Recovery/Migration can be initiated both the
user and the provider. This process can use MD-SLC
to initiate full or partial resources re-
synchronisation, it may also require re-composition.
Figure 3: On-demand Composable Services Provisioning
Workflow.
5 GEMBUS AS A CSA
MDDLEWARE PLATFORM
GEANT Multidomain Bus (GEMBus) is being
developed as a middleware for Composable Services
(GEANT3 Project) in the framework of the
GÉANT3 project aimed for creating a new
generation of the pan-European academic and
research network GÉANT. GEMBus incorporates
the SOA services management paradigm (OASIS) in
on-demand service provisioning. The Composable
Services Architecture will span over different
service interactions, from the infrastructure up to
application elements and will provide functionality
to define, discover, access and combine services in
the GÉANT environment. The goal of GEMBus is to
establish seamless access to the network
infrastructure and the services deployed upon it,
using direct collaboration between network and
applications, and therefore providing more complex
community-oriented services through their
composition.
GEMBus is built upon the industry accepted
Enterprise Service Bus (ESB) (Chappell, 2004) and
extends it with the necessary functional components
and design patterns to support multidomain services
and applications. Figure 4 illustrates the suggested
GEMBus architecture. GEMBus infrastructure
includes three main groups of functionalities:
GEMBus Messaging Infrastructure (GMI) that
includes the messaging backbone and other
message handling services such as routing,
configuration services, secure messaging, and
event handler/interceptors. GMI supports both
SOAP-based and RESTful services conforming
to Representational State Transfer architecture
(REST) (Pautasso et al., 2008).
GEMBus infrastructure services that support
reliable and secure composable services
operation and the whole services provisioning
process. These include such services as service
registries, composition and orchestration,
security and access control, logging and
monitoring, and the also important Lifecycle
Metadata Service.
Component services, although typically
provided by independent parties, will need to
implement special GEMBus adaptors or use
special “plug-in sockets” that allow their
integration into the GEMBus/CSA
infrastructure.
The following functionalities are essential to
enable GEMBus operation in the multidomain
heterogeneous service provisioning environment:
Service registries supporting service registration
and discovery. Registries are considered as an
important component to allow cross-domain
heterogeneous services integration and metadata
management during the whole services
lifecycle.
Security, access control, and logging should
provide consistent services and security context
management during the whole provisioned
services lifecycle.
Service Composition and Orchestration models
and mechanisms should allow integration with
the higher level scientific or business workflow.
The GEMBus and GMI in particular are built on
the top of the standard Apache/Fuse messaging
infrastructure that includes the following
components (Fuse ESB, Apache ServiceMix):
DEFINING GENERIC ARCHITECTURE FOR CLOUD IAAS PROVISIONING MODEL
83
Figure 4: GEMBus infrastructure, including component
services, service template, infrastructure services, and core
message-processing services.
Fuse Message Broker (Apache ActiveMQ)
messaging processor
Fuse Mediation Router (Apache Camel)
normalised message router
The GEMBus services and applications can be
deployed on the standard Fuse or Apache ESB
server. As component services that can be integrated
with the standard OSGi (OSGi) and Spring (Spring
Security) compliant service development
frameworks and platforms such as Fuse Services
Framework/Apache CXF and Fuse ESB/Apache
ServiceMix.
6 CSA SECURITY
INFRASTRUCTURE
Providing consistent security services in CSA and
GEMBus is of primary importance due to potentially
multi-provider and multi-tenant nature of Clouds
IaaS environment. The CSA security infrastructure
(CSA-Security) should address two aspects of the
IaaS operation and dynamic security services
provisioning:
Provide security infrastructure for secure IaaS
operation, including access control and SLA
and policy enforcement for all interacting roles
and components in CSA, secure messaging and
transport services.
Provisioning dynamic security services,
including creation and management of the
dynamic security associations, as a part of the
provisioned complex/composite services or
virtual infrastructures.
The first task is a traditional task in security
engineering, while dynamic provisioning of
managed security services remains a problem and
requires additional research.
An important issue in building consistent
security services for dynamically provisioned virtual
infrastructures is the Security Services Lifecycle
Management (SSLM) that extends the described
above CSA SDF service lifecycle management
model with additional sub-stages and functions to
bind dynamic security context to the general
provisioning session and Cloud virtualisation and
hosting platform in such a way that to ensure all
operations on the virtual infrastructure and user data
to be secured during their whole lifecycle. It is
described in details in earlier authors’ paper
(Demchenko et al., 2010).
CSA-Security and GEMBus should implement
the following basic infrastructure security services to
ensure normal operation of the virtual infrastructure:
Access control (e.g. Authentication,
Authorisation, Identity Management)
Policy and SLA enforcement
Data, messaging and communication security
Additionally, auditing/logging and accounting.
CSA-Security should implement multi-layer
security services including transport, messaging and
application/data security, and additionally network
layer security for distributed VPN based enterprise
domains. Security and security services in the CSA
and GEMBus design are applied at different layers
and can be called from different functional
components using standard/common security
services interface. Security services are governed by
related security policies.
Security services should support the whole
provisioned/composable services lifecycle and
consequently support session-related security
context management.
Security services can be designed as: pluggable
services operating at the messaging layer; OSGi
bundles that can be dynamically added as
composable services to other composable services to
form an instant virtual infrastructure; or exposed as
Web services and be integrated with other services
by means of higher level workflow management
systems.
7 SUMMARY AND FUTURE
DEVELOPMENTS
This paper presents the ongoing research on
developing architecture and framework for
dynamically provisioned and reconfigurable
CLOSER 2011 - International Conference on Cloud Computing and Services Science
84
infrastructure services to support modern e-Science
and high-technology industry applications that
require both high-performance computing resources
(provisioned as Grids or Clouds) and high-speed
dedicated transport network.
The paper proposes the Composable Services
Architecture (CSA) that is intended to provide a
conceptual and methodological framework for
developing dynamically configurable virtualised
infrastructure services.
The proposed CSA is currently being
implemented in the framework of the GEANT3
Project as an architectural component of the GEANT
Multidomain service bus (GEMBus). The GEMBus
extends the industry adopted Enterprise Service Bus
(ESB) technology with the additional functionality
to support multidomain services provisioning. The
GEMBus infrastructure intended to allow dynamic
composition of the infrastructure services to support
collaboration of the distributed groups of
researchers.
The concepts and solutions presented in this
paper are intended to be offered as a contribution to
the Open Grid Forum (OGF) Research Group on
On-Demand Infrastructure Services Provisioning
(ISOD-RG) initiated by the authors and the involved
projects (for details see materials of ISOD BoF at
OGF30 (ISOD BoF) and planned ISOD-RG meeting
at OGF31).
The authors believe that concepts proposed in this
paper will provide a good basis for the further
discussion among researchers about defining an
architecture for dynamically configured virtualised
infrastructure services as a part of the Clouds IaaS
model.
ACKNOWLEDGEMENTS
This work is supported by the FP7 EU funded
project GEANT3 (FP7-ICT-238875), and the FP7
EU funded Integrated project The Generalised
Architecture for Dynamic Infrastructure Services
(GEYSERS, FP7-ICT-248657).
REFERENCES
NIST Definition of Cloud Computing v15. [Online] http://
csrc.nist.gov/groups/SNS/cloud-computing/cloud-def-
v15.doc
GFD.150 Using Clouds to Provide Grids Higher-Levels of
Abstraction and Explicit Support for Usage Modes.
[Online] http://www.ogf.org/documents/ GFD.150.pdf
GEANT Project. http://www.geant.net/pages/ home.aspx
Generalised Architecture for Dynamic Infrastructure
Services (GEYSERS Project)-http://www.geysers.eu/
Enabling Grid from E-sciencE (EGEE Project). [Online].
http://www.eu-egee.org/
Phosphorus Project. [Online]. Available: http:// www.ist-
phosphorus.eu/
Demchenko, Y., M. Cristea, C. de Laat, E. Haleplidis,
Authorisation Infrastructure for On-Demand Grid and
Network Resource Provisioning, Proc. 3rd Intl ICST
Conference on Networks for Grid Applications
(GridNets 2009), Athens, Greece, 8-9 Sept 2009.
ISBN: 978-963-9799-63-9
TMF Service Delivery Framework. http://
www.tmforum.org/servicedeliveryframework/4664/
home.html
Deliverable DJ3.3.1: Composable Network Services use
cases. GEANT3 Project Deliverable. January 11,
2010. http://www.geant.net/Media_Centre/Media_
Library/Media%20Library/GN3-09-198-DJ3_3_1_
Composable_Network_Services_use_cases.pdf
OASIS Reference Architecture Foundation for Service
Oriented Architecture 1.0, Committee Draft 2, Oct. 14,
2009. http://docs.oasis-open.org/soa-rm/soa-ra/v1.0/
soa-ra-cd-02.pdf
Chappell, D., "Enterprise service bus", O’Reilly, June
2004. 247 pp.
Pautasso, C., O. Zimmermann, F. Leymann, "RESTful
Web Services vs. Big Web Services: Making the Right
Architectural Decision", 17th International World
Wide Web Conference (WWW2008), Beijing, China.
Fuse ESB - OSGi based ESB. - http://fusesource.com/
products/enterprise-servicemix/#documentation
Apache ServiceMix an Open Source ESB. - http://
servicemix.apache.org/home.html
OSGi Service Platform Release 4, Version 4.2. -
http://www.osgi.org/Download/Release4V42
Spring Security. Reference Documentation. http://
static.springsource.org/spring-security/site/docs/3.1.x/
reference/springsecurity-single.html
Demchenko, Y., D. R. Lopez, J. A. Garcia Espin, C. de
Laat, "Security Services Lifecycle Management in On-
Demand Infrastructure Services Provisioning",
International Workshop on Cloud Privacy, Security,
Risk and Trust (CPSRT 2010), 2nd IEEE International
Conference on Cloud Computing Technology and
Science (CloudCom2010), 30 November - 3 December
2010, Indianapolis, USA.
On-Demand Infrastructure Services Provisioning BoF
(ISOD BoF), OGF30 meeting, 25 October 2010,
Brussels. http://www.gridforum.org/gf/event_
schedule/index.php?id=1961
DEFINING GENERIC ARCHITECTURE FOR CLOUD IAAS PROVISIONING MODEL
85