NEGOTIATION POLICIES FOR PROVISIONING

OF CLOUD RESOURCES

Salvatore Venticinque

, Viorel Negru

2,3

, Victor Ion Munteanu

2,3

, Calin Sandru

2,3

Rocco Aversa

and Massimiliano Rak

Second University of Naples, Naples, Italy

Research Institute e-Austria (IeAT), Timisoara, Romania

Computer Science Department, West University of Timisoara, Timisoara, Romania

Keywords:

Negotiation, Sky computing, Multi-agent system, Negotiation policies, Cloud resources, Quality of services,

Service-level agreement, Resource provisioning.

Abstract:

Cloud represents today a computing market where users can buy resources according to a pay-per-use business

model. Cloud providers offer different kind of services which can be characterized by different service levels.

Negotiation of the best resource can be very difﬁcult because there is a semantic gap between the different

provider SLAs and the requirements of an user’s application. We address the negotiation issue within the

research activity of the mOSAIC project by designing and developing an agents based service at platform

level for provisioning of cloud resources. In order to allow the execution of the application and fulﬁll the

developer’s requirements, the service uses a policy based approach that is able to choose, at infrastructure

level, the best solution, in terms of a collection of cloud resources from different providers.

1 INTRODUCTION

Cloud Computing has emerged as the reference

paradigm for selling or renting computing resources.

Let’s imagine the Internet as a huge computer mar-

ket where customers look for, and get any kinds of

computer resources. We mean hardware, operating

systems, development platforms, application servers

or complete applications.

In this context, the problems that this paper ad-

dresses are on how to describe one’s own require-

ments, how to buy, pay and get cloud resources,

how to check if the resources are compliant with the

provider’s service level agreement.

The business model of cloud computing is pay per

use. It means that the business model does not tie

customers to one provider and that they could change

it when the application requirements change or when

a more convenient offer is available.

However this opportunity is limited by the fact

that, currently, providers use proprietary technologies

which make it impossible for a transparent migration

of users’ applications and platforms from one cloud

solution to another. This is known as vendor-lock-in

problem. It represents a relevant issue that is adressed

by the scientiﬁc community and founding calls. In

this context, the mOSAIC project (mOSAIC, 2010)

proposes a new, enhanced, programming paradigm

that is able to exploit the cloud computing features,

building applications which are able to adapt them-

selves as much as possible to the available resources

and to acquire new ones when needed. mOSAIC of-

fers together an API for ﬂexible, scalable, fault tol-

erant, provider independent cloud applications and a

framework for enriching them with all needed fea-

tures by programming.

When interoperability and portability of applica-

tions across heterogeneous clouds are supported, au-

tonomous services, which can automatically manage

discovery, negotiation and monitoring of cloud re-

sources, represent an added value for users within the

sky computing paradigm (Aversa et al., 2011).

This article presents the state of art of research ac-

tivities on negotiation and complementary services in

Cloud. Furthermore it focuses on the agent based so-

lution proposed within the mOSAIC project and pro-

vides details about the last design choices and devel-

opments. Section 3 covers negotiation as a service.

The negotiation module, policies, models and algo-

rithms are discussed in section 4. Related work is dis-

347

Venticinque S., Negru V., Ion Munteanu V., Sandru C., Aversa R. and Rak M..

NEGOTIATION POLICIES FOR PROVISIONING OF CLOUD RESOURCES.

DOI: 10.5220/0003747003470350

In Proceedings of the 4th International Conference on Agents and Artiﬁcial Intelligence (ICAART-2012), pages 347-350

ISBN: 978-989-8425-96-6

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

cussed in section 5. Finally, we come to conclusions

in section 6.

2 CLOUD AGENCY

In the current context of the Cloud market, mOSAIC

is expected to offer the freedom of choice at program-

ming level as well as at the resource level. It intends

to create and promote an open-source Cloud applica-

tion programming interface and a platform targeted

for developing multi-Cloud oriented applications.

The Cloud Agency represents the central point for

resource management in the mOSAIC project. As a

multi-agent system, the Cloud Agency has been de-

signed to offer resource negotiation, monitoring and

reconﬁguration services. These services can be ac-

cessed by the user’s application either through mO-

SAIC’s runtime (using mOSAIC’s SDK) or indepen-

dently (using API calls). For the purpose of this pa-

per, the focus will be on the negotiation service. The

process of resource provisioning involves speciﬁca-

tion of parameters, gathering of offers and selecting

the best ones that suit the application needs. Once the

resources have been chosen, the SLA can be ﬁnally

generated. A long study has been made in order to

identify SLAs and QoS requirements that are the in-

puts of a negotiation module. These requirements are

the basis for deﬁning resources and for deﬁning both

negotiation policies and for deﬁning monitoring and

reconﬁguration policies.

3 NEGOTIATION AS A SERVICE

The Cloud Agency’s negotiation service is provided

as an asynchronous API that is designed around an

event driven model. The service calls are divided in

Service Requests, Callbacks and Queries. Access is

enabled through a HTTP REST interface.

• Service Requests are used execute something and

generate events within the agency (ex. to start a

negotiation, to accept to refuse a SLA, to change a

policy, etc). They are asynchronous and the result

comes through a CallBack.

• CallBacks are used to receive and handle events

which follow previous Service Requests.

• Queries are used to get information (ex. get

the list of vendors or resources). They are syn-

chronous and return immediately the response if

it is available, an exception otherwise.

Figure 1: Asynchronous behavior of the Negotiator.

This service is implemented by distributed and spe-

cialized agents, which interact using standard pro-

tocols. A Client Agent acts as access point (proxy

between the HTTP REST interface and the internal

Agent Communication Language). It receives the

user’s requests and cooperates with the Negotiator

Agent in order to broker cloud services and resources.

The Negotiator Agent handles SLA Template (Call

For Proposal) analysis and proposal evaluation. It

works with Mediator Agents in order to broker each

needed resource. The Mediator Agent identiﬁes avail-

able Vendor Agents and contacts each of them for

bids on the needed resources. The Vendor Agents are

wrappers (drivers) that translate the speciﬁc provider

protocols into the uniform protocol used inside the

mOSAIC framework.

The cooperation among agents has been designed

adopting well deﬁned Agents Interaction Protocols of

the FIPA standard (Grama et al., 2000). The Client

Agent implements a standard FIPA Contract Net In-

teraction Protocol with the Negotiator Agent. It sub-

mits a call for proposal using an SLA Template with

the application requirements and gets the SLA Pro-

posal. The Mediator Agent implements the stan-

dard FIPA Brokering Interaction Protocol. The sub-

protocol used to talk with vendors is again the stan-

dard FIPA Contract Net Interaction Protocol.

The behavior of each agent has been designed as

an automata. In Figure 1 the behavior of the Nego-

tiator Agent is shown. All the states can be followed

(from receiving a Call For Proposal to building a SLA

Proposal and, once accepted, booking the resources).

ICAART 2012 - International Conference on Agents and Artificial Intelligence

348

4 NEGOTIATION MODULE:

POLICIES, MODELS AND

ALGORITHMS

The negotiation model was designed around the co-

operation of the agents involved (Client Agent, Ne-

gotiator Agent, Mediator Agents and Vendor Agents.

These agents work together to achieve a ﬁrst-price

sealed-bid auction. The Vendor Agents submit bids

and the Negotiator Agent awards the contract to the

lowest price bidder.

Communication needs (with the client and within

the agency) led to the adoption of WS-Agreement

as a protocol for establishing agreements. The pro-

tocol has been used to describe documents ranging

from Call for Proposal to SLA Proposal. Each WS-

Agreement document has it’s terms deﬁned using

OCCI resource speciﬁcation. Once such a resource

has been negotiated it will contain additional infor-

mation offered by the provider (ex. price, QoS, access

information).

Having all this information enables possible rea-

soning about it. It thus makes the creation of rules for

ﬁltering for the best candidates or for speciﬁc criteria

easy.

Negotiation policies are a set of high level govern-

ing rules which deﬁne assertions or actions to be taken

when certain conditions are met. The language cho-

sen for describing these policies is WS-Policy which

is an XML-based standard that is integrated within the

WS-Agreement. We have extended this language in

order to accommodate for our speciﬁc goals.

When designing the policies, there were several

features sought like the ability to:

• deﬁne the interaction with resource parameters

and provider parameters;

• control the way matching is performed (exact, ap-

proximate etc);

• identify common elements that are relevant to the

negotiation process;

• haveﬂexibility when deﬁning policies: in terms of

resources, providers, negotiation mechanism, etc;

• be able to combine policies easily (as long as they

are not in conﬂict);

All the policies are deﬁned within a < wsp :

Policy > XML tag that is within the WS-Agreement.

They are deﬁned as blocks of < neg : Constraint >.

Each such block is composed of < neg : Targets >

and < neg : Conditions >. The < neg : Targets >

block describes a list of targets that that policy ap-

plies to. The < neg:Conditions> block describes the

Figure 2: Policy example.

rule conditions. Conditions can apply to any of the re-

source attributes both deﬁned through the OCCI spec-

iﬁcations and those added by the provider for QoS

and pricing. If the targets satisfy the conditions then

the whole < neg :Constraint > block is validated. If

all < neg : Constraint > blocks are validated then the

chosen resources are valid for building the SLA Pro-

posal.

Figure 2 expresses the ﬂexibility of the policy sys-

tem. The ﬁgure describes a policy that targets com-

pute elements and forces their vendor to be either

Google or Microsoft.

5 RELATED WORK

Research into cloud market-oriented resource man-

agement and QoS-based resource allocation mecha-

nisms has been done by (Buyya et al., 2009). An

attempt to deﬁne several QoS metrics is presented

in (Cao et al., 2009): response time, availability, re-

liability, cost and reputation are considered. A refer-

ence of SLA model is provided in (Sim, 2010) where

SLA objectives (SLOs) are used to compose an SLA.

A number of service levels and performance metrics

for each resource results in multiple SLOs for ev-

ery service. The work presented in (Kertesz et al.,

2009) represents a ﬁrst proposal to combine SLA-

based resource negotiations with virtualized resources

in terms of on-demand service provision.

Cloud multi-agent management architectures have

been proposed in (Cao et al., 2009) and (You

et al., 2011).The implementation that was presented

in (Sim, 2010) is very interesting because it supports

dynamic negotiations.

NEGOTIATION POLICIES FOR PROVISIONING OF CLOUD RESOURCES

349

Approaches to policy based automatic SLA nego-

tiation have been made by (Zulkernine and Martin,

2011) for Web Services and by (Xiao and Cao, 2010)

for Internet-based Virtual Computing Environments.

Preliminary investigations by the authors on re-

lated topics have been presented in (Venticinque et al.,

2011) and (Aversa et al., 2010). Here we describe last

advances within the scheduled activities of the mO-

SAIC project.

6 CONCLUSIONS

In this paper we described the requirements, the

design and the prototype implementation of Cloud

Agency, an agent based software that provides pro-

visioning capability to Cloud, the focus of the paper

being on the policy based negotiation module. The

Cloud-Agency has been conceived to be used both as

a stand-alone software and as a component of the mO-

SAIC framework. The agency exposes negotiation,

monitoring and autonomic reconﬁguration services.

Asynchronous APIs have been designed and provided

by a REST interface, which follows the OCCI ref-

erence model. To execute services, agents commu-

nicate using standard FIPA messages and according

to FIPA interaction protocols. A demonstrator has

been implemented using the Jade platform. Detection

of critical situations and autonomic reconﬁguration

of cloud resources are based on reconﬁguration poli-

cies. Reconﬁguration policies are event-condition-

action rules that allow to design autonomic behavior

in a ﬂexible and dynamic way. Future works aim at

engineering the complete Cloud Agency also by im-

plementing and testing effective algorithms for nego-

tiation, and real proﬁles for monitoring and recon-

ﬁguration in real case studies. Components of the

mOSAIC framework that allows to automatically in-

tegrate the agent services into the Cloud applications

will be also developed.

ACKNOWLEDGEMENTS

This research is supportedby the European grant FP7-

ICT-2009-5-256910 (mOSAIC) and partially sup-

ported by the grants POSDRU 21/1.5/G/13798, POS-

DRU/88/1.5/S/49516 and by the Romanian Govern-

ment PNII grant nr. 12118/2008 (SCIPA).

REFERENCES

Aversa, R., Di Martino, B., Rak, M., and Venticinque, S.

(2010). Cloud agency: A mobile agent based cloud

system. In Complex, Intelligent and Software Inten-

sive Systems (CISIS), 2010 International Conference

on, pages 132 –137.

Aversa, R., Venticinque, S., Martino, B. D., and Petcu, D.

(2011). Agent based cloud provisioning and manage-

ment, design and prototypal implementation. 1st In-

ternational Conference on Cloud Computing and Ser-

vices Science (CLOSER2011), pages 184–191.

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., 25:599–616.

Cao, B.-Q., Li, B., and Xia, Q.-M. (2009). A service-

oriented qos-assured and multi-agent cloud comput-

ing architecture. In Proceedings of the 1st Interna-

tional Conference on Cloud Computing, CloudCom

’09, pages 644–649, Berlin, Heidelberg. Springer-

Verlag.

Grama, A., Kumar, V., and Sameh, A. (2000).

Foundation for intelligent physical agents.

www.http://www.ﬁpa.org.

Kertesz, A., Kecskemeti, G., and Brandic, I. (2009). An sla-

based resource virtualization approach for on-demand

service provision. In Proceedings of the 3rd inter-

national workshop on Virtualization technologies in

distributed computing, VTDC ’09, pages 27–34, New

York, NY, USA. ACM.

mOSAIC (2010). The mosaic project. http://mosaic-

cloud.eu/.

Sim, K. (2010). Towards complex negotiation for cloud

economy. In Bellavista, P., Chang, R.-S., Chao, H.-C.,

Lin, S.-F., and Sloot, P., editors, Advances in Grid and

Pervasive Computing, volume 6104 of Lecture Notes

in Computer Science, pages 395–406. Springer Berlin

/ Heidelberg.

Venticinque, S., Aversa, R., Di Martino, B., Rak, M., and

Petcu, D. (2011). A cloud agency for sla negotiation

and management. In Guarracino, M., Vivien, F., Trff,

J., Cannatoro, M., Danelutto, M., Hast, A., Perla, F.,

Knpfer, A., Di Martino, B., and Alexander, M., ed-

itors, Euro-Par 2010 Parallel Processing Workshops,

volume 6586 of Lecture Notes in Computer Science,

pages 587–594. Springer Berlin / Heidelberg.

Xiao, Z. and Cao, D. (2010). A policy-based framework

for automated sla negotiation for internet-based vir-

tual computing environment. In Proceedings of the

2010 IEEE 16th International Conference on Paral-

lel and Distributed Systems, ICPADS ’10, pages 694–

699, Washington, DC, USA. IEEE Computer Society.

You, X., Wan, J., Xu, X., Jiang, C., Zhang, W., and Zhang, J.

(2011). Aras-m: Automatic resource allocation strat-

egy based on market mechanism in cloud computing.

Journal of Computers, 6(7).

Zulkernine, F. and Martin, P. (2011). An adaptive and in-

telligent sla negotiation system for web services. Ser-

vices Computing, IEEE Transactions on, 4(1):31 –43.

ICAART 2012 - International Conference on Agents and Artificial Intelligence

350