A CLOUD PLATFORM FOR REAL-TIME INTERACTIVE

APPLICATIONS

Andreas Menychtas, Dimosthenis Kyriazis, Spyridon Gogouvitis

National Technical University of Athens, Iroon Polytechniou 9, Zografou Campus, 15773 Athens, Greece

Karsten Oberle, Thomas Voith

Alcatel-Lucent Bell Labs, 70435 Stuttgart, Germany

Georgina Galizo, Soeren Berger

HLRS - High Performance Computing Center Stuttgart, Nobelstrasse 19, 70569 Stuttgart, Germany

Eduardo Oliveros

Telefónica I+D, Emilio Vargas 6, 28043 Madrid, Spain

Michael Boniface

IT Innovation Centre, University of Southampton, 2 Venture Road, SO16 7NP, Southampton, U.K.

Keywords: Cloud computing, Cloud architecture, Cloud platform, Software as a service, Platform as a service,

Infrastructure as a service, Real-time, Service oriented.

Abstract: Cloud Computing is considered nowadays as the future of ICT systems leveraging new methodologies for

developing, providing and consuming services. Even though many people believe that “Cloud” is just

another buzzword for utility computing, this new computing paradigm is not only changing the design of

modern computing platforms in technical level, but it also impels, from the market perspective, the creation

of new value chains and business models. Beyond the great advantages of cloud technologies for scalability,

elasticity and low operational cost, there are still many technical complexities and limitations on

provisioning and management of applications with high QoS demands that disallow the wide adoption of

cloud solutions. In this paper we present a novel cloud platform capable to support real-time interactive

applications considering their full lifecycle including service engineering, SLA negotiation, provisioning

and monitoring. This platform has been designed and implemented consolidating management and control

of the infrastructure and services at all points in the value chain to support real-time interaction focusing on

its business and commercial orientation.

1 INTRODUCTION

Although cloud computing (Buyya, 2009) as another

distributed computing paradigm is not something

new, nowadays seems that the number of people and

organizations exploiting the cloud computing

capabilities is increasing. The main IT players such

as Google and Microsoft have already developed

platforms to offer cloud services hosted in their

datacenters and at the same time hundreds of new

companies worldwide are involved in the service

delivery value chain either by using their owned

infrastructures or by providing added value services.

The new cloud ecosystems are changing the way the

computing, storage and networking resources are

purchased and consumed creating new business

paradigms and value networks for the service

397

Menychtas A., Kyriazis D., Gogouvitis S., Oberle K., Voith T., Galizo G., Berger S., Oliveros E. and Boniface M..

A CLOUD PLATFORM FOR REAL-TIME INTERACTIVE APPLICATIONS.

DOI: 10.5220/0003387003970403

In Proceedings of the 1st International Conference on Cloud Computing and Services Science (CLOSER-2011), pages 397-403

ISBN: 978-989-8425-52-2

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

delivery. In contrast with the proprietary software

where the license schemas are rather simple, the

cloud based services -exploiting the advantages of

the cloud for scalability, elasticity, multi-tenancy

and reliability- are strongly related with the business

aspects of the application and platform influencing

all processes of the service lifecycle.

From the technical perspective and based on the

SPI (Service - Platform - Infrastructure) model

(NIST, 2009), the cloud solutions can be categorized

in three main classes:

 Infrastructure as a Service (IaaS)

 Platform as a Service (PaaS)

 Software as a Service (SaaS)

In the paper we present a novel cloud platform,

which was developed in the frame of the EU-funded

project IRMOS (see References section) targeting

soft real-time applications that have stringent timing

and performance requirements. This platform

combines Service Oriented Infrastructures - SOIs

(Erl, 2005) with virtualisation technologies to

manage and provision computational, storage and

networking resources as well as to communicate

with legacy systems such as wifi locators. Service-

oriented design has several advantages that the

proposed solution exploits in order to dynamically

connect people, processes, information and services

that are spanning across different domains or layers

of the platform architecture. The platform

specification advances existing service-oriented

approaches by providing methodologies, tools and

mechanisms in order to efficiently operate, manage

and reconfigure services and resources under real-

time constraints. The constraints are expressed as

Quality of Service (QoS) terms in Service Level

Agreements (SLAs) that are dynamically negotiated

and define commitments between the different

stakeholders in the value chain.

Provisioning applications in virtualised

infrastructures with guaranteed QoS is a non-trivial

task. At the core is the need to intelligently allocate

and adapt resource provisioning policies based on

knowledge of the application, customer and

infrastructure while during the operational phase the

ability to monitor the actual performance triggering

mitigating actions in real-time to overcome from

exceptional behaviour of services and resources. The

platform supports these requirements through a set

of Framework Services - FS that implement a QoS-

oriented Service Engineering methodology

(Kyriazis, 2010) linking the lifecycle processes with

novel modelling tools and autonomic management

services. In that sense, the platform is not only a set

of interacting services but also a real-time enabled

system in which instances of services with real-time

capabilities are deployed in the virtual environments

supervising the application lifecycle and

guaranteeing the agreed QoS level.

Figure 1: Platform Layers based on SPI Cloud Model.

Following the SPI cloud model (Figure 1) the

overall management processes of the platform as

well as the individual services are designed,

developed and distributed in order to support real-

time interactivity not only for the applications but

also for the infrastructure itself. The platform

provides semantic representations of systems in

order to efficiently allow for scalability,

interoperability and adaptation and therefore solves

several problems in reference to QoS provisioning

such as the real-time scheduling of the service

execution and mapping of the application workflows

and requirements to low level resource parameters.

To achieve the guaranteed real-time end-to-end

performance, the platform continuously exchanges

management information between its two main

building blocks the Framework Services - FS (PaaS

layer) and the Intelligent Service Oriented Network

Infrastructure - ISONI (IaaS layer) that are described

in detail in next sections.

The remainder of the paper is structured as

follows: Section 2 describes the methodology and

the principles that were followed in the design of the

platform architecture to achieve real-time QoS

provision while section 3 introduces the control

loops concept that allows the infrastructure to

provide real-time QoS guarantees and the overall

platform architecture design as well as the

specifications of the FS and ISONI are detailed in

section 4. The paper concludes with a discussion on

future research and potentials for the current study.

ServiceEngineering

ServiceManagement,

WorkflowManagement,SLA

Management,Monitoring

VirtualizedResources,

ExecutionEnvironment

ApplicationAdaptation

SaaS

PaaS

IaaS

CLOSER 2011 - International Conference on Cloud Computing and Services Science

398

2 REAL-TIME QOS PROVISION

IN CLOUDS

The primary objective of the proposed platform is to

develop a cloud solution capable to support QoS

guarantees through all layers of the system for

interactive real-time multimedia applications. The

architecture considers the full service lifecycle of

both service-based systems and legacy applications

deployed on cloud resources including service

engineering, service level agreement design and

resource management and monitoring. QoS

parameters at application, platform and

infrastructure levels are given specific attention as

the basis for dynamic QoS provisioning in real-time.

The research work around QoS provisioning and

SLA enforcement in clouds is not something new.

State of the Art cloud providers such as Amazon

EC2 have already implemented mechanisms for

monitoring the guarantees for QoS which are mainly

expressed as “Annual Uptime Percentage” enriched

with penalty models (Amazon EC2-SLA). This

methodology is generic and static without taking

into consideration the particular QoS requirements

of each application and cannot be adapted during

runtime. Emeakaroha et al. (Emeakoraha, 2010) in

the LoM2HiS approach and Stantchev et al.

(Stantchev, 2009) propose solutions for mapping

the high level application requirements to the

monitored metrics focusing mainly on SLA

monitoring and SLA violation detection for

achieving QoS provisioning. Nevertheless these

solutions do not exploit any application or platform

reconfiguration techniques so as to maintain the

overall system stability and acceptable operational

level as required by real-time interactive

applications. On the other hand the real-time clouds

approaches have been discussed from Liu et al. (Liu,

2010) and Sarathy et al. (Sarathy, 2010) focusing

only on resource management at infrastructure level.

The proposed platform architecture supporting

real-time service, human and resources interactivity

implements the following key features:

 Real-Time QoS Specification

 Event Prediction

 Dynamic SLA (Re)Negotiation

 On-Demand Resource Provisioning

 QoS aware Event Monitoring

Service-oriented design principles are considered as

an important aspect of the architecture throughout all

cloud platform layers. The platform adopts a

service-oriented approach to allow services interact

dynamically and continuously even though they

span between different domains, from the

application layer to the layer of network resources

and execution environment. The challenge is to

carefully design and synchronize this rich set of

services so as to efficiently operate, manage and

reconfigure all resources under real-time conditions,

providing to the end users and to the associated

applications the appropriate and required Quality of

Service. All QoS terms are dynamically negotiated

and agreed in SLAs between the various actors of

the value chain (Gallizo, 2009) taking into

consideration the QoS guarantees from both

application and resource perspectives. All platform

and infrastructure capabilities are offered as on-

demand services, although the architecture design of

the media applications may vary from traditional n-

tier enterprise applications to service-oriented

workflows. In that frame, the service orchestrations

and processes are developed in a way so as to

preserve the real-time attributes throughout the

whole infrastructure layers.

A major challenge for SaaS providers wanting to

exploit the benefits of cloud computing is to manage

QoS commitments to customers throughout the

lifecycle of a service. The PaaS offers to SaaS

providers services tools for estimating resource

needs in advance of execution and mechanisms for

negotiating QoS with service providers and

provisioning virtualised resources. This also

includes assessing the probable technical and

economic outcomes of provisioning policies and

management actions even if the application or

resources do not perform as expected or need to be

adjusted. The proposed approach considers analysis

and decision support within temporal and business

constraints to determine which actions are triggered

offline (i.e. pre-execution) or online (i.e. during

execution). Because faults are inevitably going to

occur, strong fault detection and recovery

mechanisms are implemented. This can have a great

impact on the real-time capabilities of the platform,

since intelligent fault recovery mechanisms allow

timing constraints to still be met in case of a failure.

The performance of the monitoring and control

loops between cloud layers, described in section 3, is

as essential factor in ensuring that QoS guarantees

are maintained.

At the IaaS layer, real-time functionality is

supported by the Intelligent Networking and the

Execution Environment infrastructures applying

virtualization techniques for several types of

resources such as networking, storage and

computational. The IaaS layer instantiates, manages

and monitors the various resources according to the

A CLOUD PLATFORM FOR REAL-TIME INTERACTIVE APPLICATIONS

399

set of services that are deployed. To this direction,

Execution Environment considers multitasking,

threads with priorities and an appropriate number of

interrupt levels to achieve QoS objectives.

Another essential element of cloud computing,

especially of PaaS layer, is the ability to deliver on-

demand services with minimal manual

configuration. In that sense, all platform subsystems

can be self-managed and reconfigured in order to

achieve management efficiencies, to react to QoS

failures (such as for instance an SLA violation or

Network link failure) in a timely way and avoid the

escalation of interlayer problems.

Cloud utilisation involves several processes that

span in different cloud layers and stakeholders. For

example, the platform supports application

developers in engineering their applications for the

cloud implementing standard specifications and

methodologies, while other processes support

application provisioning and execution through the

innovative virtualised execution environment and

networking infrastructure. Therefore, the platform

does not only provide a set of services but also cross

layer workflow mechanisms that consider the

control channels and information exchanges which

are required to support real-time management of

interactive applications throughout the full lifecycle.

3 CONTROL LOOPS

In order to provide QoS guarantees for interactive

real-time multimedia applications, platform provides

a set of services and cross layer workflows that

consider the control channels and information

exchanges which are required to support real-time

management throughout the full lifecycle. All

subsystems are self-managed and reconfigured in

order to achieve management efficiencies, and to

react on QoS failures (such as an SLA violation or

network link failure) in a timely way. To achieve

this, we introduce three control loops at

infrastructure level providing the necessary

functionality in order to maintain QoS metrics across

the architectural levels. The IRMOS Control Loops

are the following and are depicted in Figure 2:

 Application Control:

It deals with the

relationship between users and applications required

to guarantee the application QoS. This control loop

is managed by the application itself in response to

either user events or platform events. It is

implemented with the use of models, workflows and

tools that produce artifacts capturing the

applications’ behavior and estimating resource needs

in advance of execution. During runtime it refers to

application monitoring that may for example trigger

events or require for changes in the provided

resources.

 Environment Control:

It deals with the

relationship between applications and virtual

resources in order to guarantee the platform QoS, as

agreed in the SLAs. This control loop is managed by

the platform services in response to application and

virtualisation events. It is implemented by the

framework services that support and manage the

applications at run-time (e.g. actions triggered if

either the application or resources do not perform as

expected or need to be adjusted).

 Virtualization Control:

It deals with the

relationship between virtual and physical resources

in order to guarantee the infrastructure QoS. This

control loop is managed within IaaS layer called

ISONI in response to platform or physical events. It

is implemented by intelligent networking

mechanisms as well as by the real-time enabled

execution environment for computational and data

storage services.

Figure 2: Platform Control Loops.

We identified five (5) main processes / channels

implementing the control loops:

 Service Engineering

 Negotiation / Renegotiation

 Reservation

 Monitoring and Evaluation

The actual implementation of the control loops

refers to tools and services used on different levels

in order to monitor the applications’ execution,

communicate possible events and take corrective

actions if needed. These mechanisms are analyzed in

the following section as part of the platform

architecture.

IRMOSplatform

Applicationprovider=IRMOScustomer

Infrastructure Resources

(computing,storage,network)

Software-as-a-Service

Platform-as-a-Service

Infrastructure-as-a-Service

Consumer

PlatformControl

InfrastructureControl

ResourceControl

Service Engineering

Negotiation

Reservation

Monitoring & Evaluation

ISONI

CLOSER 2011 - International Conference on Cloud Computing and Services Science

400

4 PLATFORM ARCHITECTURE

In this section we describe the overall architecture of

the platform and its main subsystems. The high-level

view of the platform architecture is shown in the

following figure:

Figure 3: Platform Overall Architecture.

As indicated in Figure 3, the platform has the

two main building blocks: PaaS (Framework

Services) and IaaS (ISONI). During the architecture

design and specification we followed an innovative

approach on how these blocks will interact, and in

that sense their relation is considerably different of

the conventional SOA or Cloud platforms because of

its real-time orientation, the virtualization

capabilities and the way the management

information is shared between platform and

infrastructure layers to ensure end-to-end QoS.

Initially, the Framework Services provide service

engineering tools for the application developer and

provisioning services for the IRMOS provider as the

entity responsible for offering applications.

Each VSN is instantiated by ISONI and includes

particular technical requirements defined by the

application developer at design time and specific

QoS customisation defined by the customer at

runtime. These requirements are relayed to ISONI

during the SLA negotiation through the Framework

Services. ISONI cannot be accessed directly by end

users (Customers or Consumers) and their access

privileges are limited to application service

components.

As already mentioned, the Framework Services

communicate continuously and in various ways with

ISONI. Each ISONI provider advertises its

capabilities to Framework Services so as to be

discovered later and as second step negotiate SLAs

for an application. Additionally ISONI provides

monitoring data and notification events for each

VSN (at the SC level) to Framework Services that

are used for both runtime (control) and design time

(development and modelling). The fact that real-time

functionality is required on some components of the

Framework Services layer demands that instances of

these components will be deployed and run in the

VSN where the real-time QoS is guaranteed. As

presented in Figure 3, the core platform services in

PaaS interact with service instances running in

VSNs for controlling and monitoring the application

during application execution.

The Execution Environment and the Intelligent

Networking subsystems are architecturally close and

are expected to communicate continuously during all

the processes of the platform (Oberle, 2010). These

subsystems are wrapped in the ISONI infrastructure.

The main objective of this layer is to virtualize

resources, provision of application services and

monitor the resources without the need for

knowledge on the application itself. The Execution

Environment subsystem, considered as an enhanced

virtualization platform, includes the storage systems

and is implemented so as to address the QoS and

especially the real-time requirements of the

application services. The network resources,

provided through a VPN like approach, are

classified and advertised to the Framework Services

in QoS classes.

4.1 Framework Services

The Framework Services – FS is the layer between

applications and virtualized resources offered by

IaaS providers. This layer corresponds to the PaaS

layer of SPI cloud model and the architecture is

shown in Figure 4. The architecture consists of two

main elements, Service Engineering and Service

Management.

The Framework Services layer aims to provision

and manage the execution of real-time services on

request of the Application Layer using virtualized

resources. These resources are offered by the IaaS

providers conforming to the real-time constraints as

determined in the application SLAs. Apart from the

execution of the services that are provided to

customers, Framework Services support service

engineering, fully automated SLA (re-)negotiation,

mapping of high level performance parameters to

low level resource parameters, discovery and

reservation of the virtualized resources needed for

the execution of an application. In the execution

phase of the application, FS monitor continuously

and manage the application components and the

resources either directly, through the application

IRMOSPlatform

Execution

&Storage

Virtualisation

Network

Virtualisation

ISONI

Provider

Virtualised

ServiceNetwork

Application

Resource

Management

SLA

Management

&Discovery

Service

Engineering

IRMOS

Provider‐

Framework

Services

Application

control

Information

Services

Application

Services

Application

Developer

Consumer

CCSC SC

ApplicationProvider

(IRMOScustomer)

IRMOSvaluechain

IRMOS

Provider

ISONI

Provider

Application

Management

Development Negotiation Monitoring

Discovery

Negotiation

Resource

monitoring

Access

Application

monitoring

A CLOUD PLATFORM FOR REAL-TIME INTERACTIVE APPLICATIONS

401

wrappers based on predefined application specific

policies, or relaying the management requests to

ISONI layer based on operational policies of the

platform. It should be noted that instances of the

Framework Services such as Workflow Enactor and

Monitoring Services are deployed within the

application VSN so as to benefit from the QoS

provisions the IaaS can offer.

Figure 4: Framework Services General Architecture.

For the communication with the user of the

platform, the IRMOS Portal component has been

implemented, which provides the necessary interface

to enable the end-user of the application to request

the SLAs templates, invoke the negotiation process

and the reservation of virtualized resources on the

IaaS layer. In addition, FS functionality includes

starting and stopping of an application execution

relaying the requests, through the Service

Management system, to the appropriate application

service components running in a VSN.

4.2 Intelligent Service Oriented

Network Infrastructure

ISONI (Intelligent Service Oriented Network

Infrastructure) is an IaaS environment, consisting of

a network of resources (e.g. CPU, storage, software,

etc) managed and controlled by a middleware, which

allows resource sharing among multiple services

(ISONI White Paper, 2010), (Oberle, 2009), (Oberle,

2010). The general idea is to provide QoS capable

infrastructure resources on demand for dynamically

deployed services. As already described in a

previous section, a service is usually composed out

of several smaller and simpler services, in the

following called Service Components (SC). ISONI

is agnostic to services, thus the decomposition of

services into SC is not its responsibility, and is

accomplished by the Framework Services layer. The

objective of ISONI is to provide these SCs with the

best resources (Execution Environments and

network links). Figure 5 depicts the main

components of the IaaS layer.

ISONI exposes its virtualized infrastructure in

the form of VSNs, which can be seen as a graph

whose vertexes are the SCs and whose edges are the

Virtual Links. In the proposed platform the

Framework Services will state their infrastructure

requirements using a VSN description.

It is the role of the Framework Services layer,

which is application aware, to decompose its

Services and Applications into Services

Components, build the VSN description and request

resources from ISONI which is application unaware.

The VSN description is transferred to the IaaS layer

with the request to instantiate the service. Then, the

ISONI has to automatically and autonomously map

the highly abstracted resource request in form of the

VSN description onto the network of real resources,

to deploy the components in tailored execution

environments on suitable resources, and to interlink

them while observing QoS requirements. This

instantiated VSN builds an independent layer 3

overlay network, i.e., there is no limitation on the L3

protocol stack used by the SCs.

The ISONI architecture is composed of

functional blocks, where each takes on a different

task for the management of the ISONI resources and

deployed VSNs.

Figure 5: ISONI General Architecture.

Figure 5 shows the functional building blocks of

the ISONI management. The interfaces to the

Framework Services and tools are indicated on top.

The functional blocks Resource Manager, Storage

Manager and Path Manager would be deployed in a

two-level management architecture based on the

composite structure, Domain level and Node level.

The resource responsibility lies with the Node level,

i.e. the Node control and resource reservations are

maintained by the middleware functional blocks

PortalGUI

WorkflowManager

Information &

Discovery

FS(PaaS)

Mapping

Service

Monitoring&

Evaluation

SLAManagement

System

IRMOSPortal

Application

ModellingTool s

Performance

ModellingTool s

ServiceEngineering ServiceManagement

ISONI(IaaS)

Wor k f low Enactor Monitoring

ASC

Wrapper

VSN

Wrapper

ISONI

ASC

Wor k flow 

Enactor

Monitoring

VSN

Wra pper

Wrapper

Application

Developer

Application

Provider

SaaS

Customer

ServiceManagement

FS(PaaS)

IRMOSPortal

ServiceEngineering

ISONIDomain(IaaS)

ISONI

ASC

Wor k f low

Enactor

Monitoring

VSN

Wrapper

Application

Developer

Application

Provider

SaaS

Customer

ISONIGateway

ResourceManager

Deployment

Manager

Repository

ISONISLA

Manager

Registrar

ISONI

info

EE/VMU

factory

PathManager

Interworking

DomainManager

Repository

Manager

CLOSER 2011 - International Conference on Cloud Computing and Services Science

402

running at Node level, whereas the Domain level

instances coordinate the ISONI Nodes. This

approach guarantees efficient management of the

VSNs as well as the resource scalability, a key

requirement in cloud environments.

5 CONCLUSIONS

The paper presented a novel cloud platform capable

to support the full lifecycle of applications with real-

time QoS requirements. In addition to the platform

design and specification we described

methodologies and best practices that have been

followed in the platform to effectively provision and

manage application services and infrastructure

resources during runtime. The proposed platform

promises to significantly advance the state-of-the-art

in provisioning applications with guaranteed QoS on

virtualised infrastructures and has been validated by

three different application scenarios, which were

also the basis for the requirements identification

during the design process. The evaluation results

were impressive in all scenarios with the system

capable to reconfigure in an acceptable time frame.

In cases of live migration of virtual machines the

reconfiguration time (from the user perspective) is

close to one second, depending on the application,

while in cases of user or system driven SLA

renegotiation the platform scaling is completed in

less than a minute. The final prototype supporting

service resilience and events evaluation processes

advancing further the QoS provisioning and real-

time management capabilities of the platform was

released on January 2011 and is available on project

website.

ACKNOWLEDGEMENTS

The research leading to these results has been

performed in the context of the project “Interactive

Real-time Multimedia Applications on Service

Oriented Infrastructures” (IRMOS). The project has

received funding from the EC Seventh Framework

Programme FP7/2007-2011 under grant agreement

n° 214777.

REFERENCES

Amazon EC2 Service Level Agreement:

http://aws.amazon.com/ec2-sla.

Buyya, R., Yeo, C. S., Venugopal, S., Broberg, J., and

Brandic, I. 2009. Cloud computing and emerging IT

platforms: Vision, hype, and reality for delivering

computing as the 5th utility. Future Gener. Comput.

Syst. 25, 6 (Jun. 2009), 599-616.

Emeakaroha, Vincent C.; Brandic, Ivona; Maurer,

Michael; Dustdar, Schahram; , "Low level Metrics to

High level SLAs - LoM2HiS framework: Bridging the

gap between monitored metrics and SLA parameters

in cloud environments," High Performance Computing

and Simulation (HPCS), 2010 International

Conference on , vol., no., pp.48-54, June 28 2010-July

2 2010.

EU IST IRMOS Project, http://www.irmosproject.eu

G. Gallizo et al., “A Service Level Agreement

Management Framework for Real-time Applications

in Cloud Computing Environments”, CloudComp

2010, Barcelona, Spain, 25.10-28.10.2010.

Georgina Gallizo et al., “Service Level Agreements in

Virtualised Service Platforms”, eChallenges, Warsaw,

21-23 October 2009.

IRMOS Project ISONI Whitepaper V2.0, ALUD and

USTUTT, July 2010. Available on-line at

http://www.irmosproject.eu

Kyriazis D, Einhorn R, Furst L, Braitmaier M, Lamp D,

Konstanteli K, Kousiouris G, Menychtas A, Oliveros

E, Loughran N, Nasser B, “A Methodology for

engineering real-time interactive multimedia

applications on Service Oriented Infrastructures”,

IADIS Applied Computing 2010, Timisora, Romania,

2010.

NIST Definition of Cloud Computing, Peter Mell and Tim

Grance, Version 15, 2009, available online here:

http://csrc.nist.gov/groups/SNS/cloud-computing

Oberle, K., Voith, T., Stein, M., Gallizo, G., Kübert, R.,

“The Network Aspect of Infrastructure-as-a-Service”,

ICIN2010, Berlin, October 2010.

Oberle, L., Kessler, M., Voith, T., Stein, M., Lamp, D.,

Berger, S., “Network Virtualization: The missing

piece”, ICIN2009, Bordeaux 26-29.10.09.

Sarathy, V.; Narayan, P.; Mikkilineni, R.; "Next

Generation Cloud Computing Architecture: Enabling

Real-Time Dynamism for Shared Distributed Physical

Infrastructure," Enabling Technologies: Infrastructures

for Collaborative Enterprises (WETICE), 2010 19th

IEEE International Workshop on , vol., no., pp.48-53,

28-30 June 2010.

Shuo Liu, Gang Quan, Shangping Ren, "On-Line

Scheduling of Real-Time Services for Cloud

Computing," services, pp.459-464, 2010 6th World

Congress on Services, 2010.

T. Erl, “Service-oriented Architecture: Concepts,

Technology, and Design”, Upper Saddle River:

Prentice Hall PTR, ISBN 0-13-185858-0, 2005.

Vladimir Stantchev and Christian Schriipfer, Negotiating

and Enforcing QoS and SLAs in Grid and Cloud

Computing," Proceedings of the 4th International

Conference on Advances in Grid and Pervasive

Computing, pp. 25-35, April 2009.

A CLOUD PLATFORM FOR REAL-TIME INTERACTIVE APPLICATIONS

403