TOWARDS A UNIFIED MARKETPLACE FOR

FUNCTIONALITY-BASED CLOUD SERVICE DISCOVERY

∗

Mohamed Sellami, Walid Gaaloul, Bruno Defude and Samir Tata

TELECOM SudParis, UMR CNRS 5157 Samovar, Evry, France

Keywords:

Cloud Marketplace, Service Publication and Discovery, Cloud Registry, Cloud Communities.

Abstract:

Cloud Computing brings an unparalleled shift to utility computing by providing unprecedented scalability

and ﬂexibility to IT services. Indeed, the concept of Cloud computing is linked intimately those of IaaS,

PaaS, SaaS and collectively *aaS (Everything as a Service or Cloud services). This implies a service-oriented

architecture where registries play an important role. Registries, which can be as many as the large number of

Cloud services provider, compose the market where Cloud consumers and providers go to search and advertise

Web services. With the proliferation of Cloud consumers and providers, ﬁnding an adequate Cloud service

has become a complex task for a Cloud service consumer. In this paper we propose a functionality-driven

clustering approach for Cloud service registries. This clustering will be helpful for selecting an adequate

registry for service requesters. Since registries and communities are dynamic by nature, we also propose an

approach for managing registries and communities to reconcile conﬂicts resulting from the dynamic change

aspect of service registries. Experimental evaluation shows that our approach is usable in realistic situations.

1 INTRODUCTION

Cloud Computing is a new supplement, consumption,

and delivery model for IT services based on Inter-

net protocols. More and more companies are adopt-

ing the new economic model offered by Cloud Com-

puting. For instance, in a Mckinsey Quarterly sur-

vey (The Mckinsey Quarterly, 2010) conducted in

2010 on 332 companies, 75% believe that the use of

Cloud computing could drive value at their compa-

nies. Among these companies, 68% says that they

are currently adopting the Cloud to set up electronic

collaboration and 82% are planning to do it in the 18

coming months.

In the Cloud, services are the basis for achieving

electronic B2B transactions. For worries of compet-

itiveness and rentability, more and more companies

are using Web services to adapt to new demands and

expectations (Cliquet, 2011) by modeling resources

as Cloud services. In our work, we are interested in

Cloud service discovery. Finding a speciﬁc service in

a multi-Cloud service providers environment can be a

cumbersome task (for a Cloud service consumer), due

to the lack of a standardized model for service discov-

ery in this context. In fact, supposing that each Cloud

∗

This work was partially funded by the French ANR

project PAIRSE.

provider owns its repositories of offered services, and

considering that each Cloud has its own access pol-

icy to these repositories, screening all these services

repositories, to discover a service, will be inconceiv-

able for a Cloud consumer and would dramatically

slow down the discovery process.

To face this issue, we propose in this paper to build

an overlay marketplace over the existing physical ser-

vices registries (i.e. repositories of Cloud services)

belonging to different Cloud providers. To address

the likely large number of service registries in such

a marketplace and their poorly organized network,

we propose to organize registries into communities

according to the functionalities of the services they

advertise. This organization allows an efﬁcient ser-

vices requester’s query routing to potential registries

offering services with the needed functionality. We

use communities as a means for a functionality-driven

organization of a distributed Cloud registry environ-

ment. So, existing registries belonging to different

Cloud providers can be structured into communities

and each registry belongs to at least one community

with a certain extent. The registries organization in-

side this marketplace is built automatically, based on

the functional descriptions of the services offered by

a registry, and do not require any explicit intervention

from Cloud providers.

In addition, registries operate within a very dy-

252

Sellami M., Gaaloul W., Defude B. and Tata S..

TOWARDS A UNIFIED MARKETPLACE FOR FUNCTIONALITY-BASED CLOUD SERVICE DISCOVERY.

DOI: 10.5220/0003931402520257

In Proceedings of the 2nd International Conference on Cloud Computing and Services Science (CLOSER-2012), pages 252-257

ISBN: 978-989-8565-05-1

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

namic environment where changes are frequent. In

fact, a new registry can join a Cloud and others can

leave at any time. To ensure the evolution of the inter-

nal organization of our marketplace, we deﬁne man-

agement operations to handle (i) new registry emer-

gence, and (ii) existing registry disappearing.

The paper is organized as follows. Section 2 intro-

duces the proposed architecture for our marketplace.

In Section 3, we present our automatic and implicit

approach for organizing Cloud services registries as

communities. Section 4 presents the deﬁned man-

agement operations for handling the evolution of the

created organization. The implementation efforts are

shown in Section 5. Related works are discussed in

Section 6. We conclude this paper in Section 7.

2 OVERVIEW OF THE

PROPOSED MARKETPLACE

As deﬁned by NIST (the National Institute of Stan-

dards and Technology), Cloud Computing is a model

for enabling convenient, on-demand network access

to a shared pool of conﬁgurable computing resources

deﬁned, advertised and consumed as services that

can be rapidly provisioned, released, discovered with

minimal management effort or service provider or

consumer interaction. Indeed, the concept of Cloud

computing is linked intimately those of IaaS, PaaS,

SaaS and collectively *aaS (Everything as a Service

or Cloud services

). This implies a service-oriented

architecture where registries play an important role.

Cloud computing has particular characteristics that

distinguish it from classical service oriented environ-

ments: (1) currently the major Cloud providers had

already invested in large scale infrastructure ;(2) how-

ever little attention is given to Cloud service organi-

zation and discovery to exploit them; (3) as a conse-

quence the Cloud service registries are poorly orga-

nized heterogeneous, and without agreed and feder-

ated interfaces; (4) and there are concerns over service

availability and provision.

In this paper we describe a Cloud marketplace,

which represents a portal for existing registries in dif-

ferent Cloud network, in order to handle the above

challenges. As depicted in Figure 1, the marketplace

contains an overlay network (a P2P network for ex-

ample) over the physical network. Each node of the

overlay will act as a proxy for a Cloud service reg-

istry in the physical network. This overlay is orga-

nized as communities and each community represents

a set of registries offering services providing similar

We confuse the terms *aaS and Cloud service

functionalities. The general architecture of a uniﬁed

Cloud marketplace contains two main components:

the Marketplace Builder and Manager and the Cloud

services Discovery System. Just like countries’ elec-

tric networks are interconnected to facilitate market

trading across infrastructuresand thus reduce the risks

of shortage and increase efﬁciency, the service reg-

istry providers network would beneﬁt from a feder-

ated organization.

Figure 1: Overview of the proposed marketplace.

The Marketplace Builder and Manager is used

to organize registries as communities, to deploy the

overlay network and manage the organizational evo-

lution of the overlay network. In our work, we pro-

pose to ensure this organization through a dynamic

clustering technique (see Section 3). So, the organi-

zation of the different registries will be done automat-

ically and does not require any intervention from the

Cloud registry providers. The Marketplace Builder

and Manager component also ensures the evolution

of the registries organization in the view of the differ-

ent possible changes in the network (see Section 4).

The Cloud services Discovery System component im-

plements the Cloud service discovery process. In our

work, we consider Cloud service discovery as a two

successive steps process: (i) discovering the adequate

Cloud service registries and (ii) discovering the ade-

quate Cloud service services inside these registries.

In the rest of this paper, we will focus on the ﬁrst

component, the Marketplace Builder and Manager,

of our proposed marketplace due to space limitations.

Details on our work on the Cloud services Discovery

System will be given in further papers.

3 ORGANIZING REGISTRIES IN

THE MARKETPLACE

To organize registries into communities according to

their functionalities

, we associate a WSRD descrip-

Functionalities of a services registry and functionalities

of the services it advertise are interchangeably used.

TOWARDSAUNIFIEDMARKETPLACEFORFUNCTIONALITY-BASEDCLOUDSERVICEDISCOVERY

253

tion to each registry and we use these descriptions

as criterion for the organization. A registry’s WSRD

description is automatically computed based on

the services functional descriptions advertised by a

registry and do not require any explicit intervention

from a registry provider. Hence a WSRD description

gives an overview of the functional properties of

the services offered by a registry and his syntax

is inspired by the WSDL format. As we provide

a description of a registry (not of a service), we

are only interested in the abstract section of the

services description. In this work, we choose to

use semantic Web service descriptions written in

SAWSDL (Lausen and Farrell, 2007) to describe

Cloud services. Other semantic languages, such as

OWL-S (Martin et al., 2004) or WSMO (De Bruijn

et al., 2005) can be also adopted. Details about

the WSRD computing process can be found in our

previous paper (Sellami et al., 2011).

A community of service registries will bring

together registries offering similar functionalities.

Since a service registry generally offer services

proposing different functionalities, it is difﬁcult to

properly deﬁne in advance classes categorizing the

functionalities of the different registries. To organize

service registries into communities, we use a cluster-

ing technique (where the different communities will

be deduced from the registry descriptions) rather than

a classiﬁcation technique (where the different com-

munities have to be deﬁned in advance). When using

a dynamic clustering technique, the different clusters

(i.e. the service registries communities) will be iden-

tiﬁed from the given input data (i.e. the WSRD de-

scriptions) and each data point (i.e. services registry)

will be assigned to one community.

Since a registry can belong to different communi-

ties at the same time, we use an overlapping cluster-

ing method, also known as fuzzy clustering, to orga-

nize our distributed registries into communities. Con-

cretely, we proceed in three steps to build registry

communities:

1. Step 1: WSRD descriptions are processed to map

to the term-document structure.

2. Step 2: we propose a method to measure the dis-

tance between two registry vectors as needed by

the fuzzy C-means Algorithm.

3. Step 3: using the results of the two previous steps

we apply a fuzzy clustering technique to build reg-

istry communities.

The reader is referred to (Sellami et al., 2011) for

details on our clustering approach. After these three

steps, we getC communities represented by their cen-

troids. These centroids represent the mean function-

ality f of these communities. In addition, to each

community c

are associated N membership vectors

), i = 1. . . N indicating the degrees of member-

ship of the N registries to the community. These data

allow us to infer the membership degrees MEM of

the N registries to organize into communities. How-

ever, using this technique, the membership degree of

some registries to certain communities may be very

low. For this, we deﬁne a threshold

for member-

ship degrees. If the membership degree of a registry

to some community is lowerthan this threshold, it will

not be considered as a member.

Formally, we deﬁne the obtained communities as

couples c = ( f, M) where: f is a vector representing

the mean functionality of the community c and M is

the set of community members (i.e. registries).

Registries inside a community are deﬁned as cou-

ples r = ( f, MEM) where: f is a vector representing

functionalities offered by the advertised Web services

within the registry and MEM represents the registry

membership degrees to the different communities in

the network. MEM is represented as a binary relation

deﬁned as follows. MEM = {(c, d)|c ∈ C, d ∈ [0, 1]}

where: C is the community set and d is the member-

ship degree of the registry r to the community c.

4 MANAGING THE REGISTRIES

ORGANIZATION IN THE

MARKETPLACE

The obtained organization in the previous section will

be used to deploy the overlay network (a P2P network

for example) of our marketplace. Each node of the

overlay will act as a proxy for a Cloud services reg-

istry and the nodes will be organized in communities

according to the results given by our clustering tech-

nique. Technical details are given in Section 5.

However, in Cloud environments it is normal that

new registries may emerge and existing registries may

disappear for some reason. To maintain the organiza-

tion of the different registries according to their func-

tionalities and to guarantee its evolution after these

changes, we deﬁne operations for handling the fol-

lowing situations: Cloud services registries joining

the marketplace (Section 4.1) and Cloud services reg-

istries leaving the marketplace (Section 4.2).

4.1 Cloud Services Registries Joining

the Marketplace

When a new registry r is identiﬁed, ﬁrst its WSRD

CLOSER2012-2ndInternationalConferenceonCloudComputingandServicesScience

254

description is computed. After that, the registry will

be guided to the adequate communities according

to its set of membership degrees MEM. MEM is

computed by the SelectCommunity algorithm (Algo-

rithm 1). This algorithm takes as input the current reg-

istry’s WSRD description. The membership degree

is computed using the function U

. This algorithm

outputs the set of membership MEM containing the

membership degrees of the current registry to the dif-

ferent communities in the marketplace. Taking into

account that the membership degree must be greater

than a certain threshold

(line 4), the registry will

be guided to the adequate communities ((line 5)).

Algorithm 1: SelectCommunity.

Require: r :registry

Ensure: r.MEM

1: for each community c ∈ C do

2: m ←− U

(r)

3: r.MEM ←− r.MEM∪ {(c, m)}

4: if m >

then

5: M ←− M ∪ {r}

6: end if

7: end for

When the membership degrees of a registry r to all

the existing communities are lower than the thresh-

old

, it provides necessarily a new functionality

inexistent in the network. In this case, a new com-

munity c

new

is established automatically. The com-

munity mean functionality c

new

. f will be the same

as the functionality r. f proposed by the registry that

triggered the community creation. Afterwards, some

members are recruited for the new community. In this

aim, the membership degrees of existing registries to

the new community are computed. Every registry

having a membership degree higher than the thresh-

old

will be a member of c

new

. The community cre-

ation as well as the necessary precondition and effects

are displayed in Algorithm 2.

Algorithm 2: CreateCommunity.

Require: r: the registry that triggers the community creation

Ensure: c

new

: the new community

1: precondition: ∀d ∈ ran(r.MEM), d <

2: Community c

new

;

3: List members;

4: c

new

. f = r. f

5: for all communities c ∈ C do

6: for all registries r

∈ c.M do

7: m ←− U

new

)

8: r

.MEM ←− r

.MEM ∪ (c

new

, m)

9: if m >

then

10: members ←− members∪ {r

}

11: end if

12: end for

13: end for

14: effect: c

new

∈ C ∧ c

new

. f = r. f ∧ c

new

.M = members

15: return c

new

The same function U is also used in our community

building approach (Sellami et al., 2011)

Also, when a registry r is assigned to a commu-

nity c, c may become sparse and thus may need to be

split into two or more sub-communities. The com-

munity sparsity describes a non density in the center

vicinity and a dispersion between members. In this

case, the Mcut method (Ding et al., 2001; Nie et al.,

2010) can be applied for splitting a community into

several sub-communities. Concretely, this method

allows to minimize the similarity between two sub-

communities, while maximizing the similarity within

each sub-community (details on how to split commu-

nities can be found in (Bouchaala et al., 2011)).

4.2 Cloud Services Registries Leaving

the Marketplace

When an existing registry r leaves the Cloud or is un-

available for some reason, its representing node in the

overlay network is to be removed from the market-

place. In this case, if a community c contains this reg-

istry, r has to be removed from it member set c.M.

If c does not contain any registry after removing r

(c.M = {}), c will be automatically dismantled. Oth-

erwise, we check whether c can be merged with an-

other community in the overlay network. If c cannot

be merged, the mean functionality f of c will be up-

dated. Finally, the membership degree of each reg-

istry in the different communities is recomputed.

Algorithm 3: MergeCommunity.

Require: c: a community

Ensure: true or false depending on whether c is merged with an-

other community or not

1: isMerged ← false

2: for all c

oth

∈ C \ {c} do

3: if c.M * c

oth

.M then

4: continue

5: end if

6: isMerged ← true

7: C ← C \ {c}

8: break

9: end for

10: return isMerged

Next we introduce the MergeCommunity algo-

rithm. c represents the community to be merged. As

presented in Algorithm 3, if the members of c rep-

resent a subset of the members of another commu-

nity (i.e., c

oth

), c is removed from the overlay network

(lines 6-12).

5 IMPLEMENTATIONS

We have prototyped our approach in java. Cloud ser-

vices are described in a formatlike SAWSDL. We cre-

ated a test collection of semantic Web service descrip-

TOWARDSAUNIFIEDMARKETPLACEFORFUNCTIONALITY-BASEDCLOUDSERVICEDISCOVERY

255

tions written in SAWSDL. These service descriptions

are annotated using concepts from an ontology de-

scribed in OWL. For our experiments, we generated

a collection of 1400 SAWSDL descriptions. We split

up the generated SAWSDL descriptions into 7 Web

service registries r

, j = 1. . . 7. To organize these reg-

istries into communities, we calculate for each reg-

istry its WSRD description. We developed a tool,

WSRDCreator, to create WSRD descriptions from a

set of SAWSDL descriptions. WSRDCreator gener-

ates a vector representation of each registry’s created

WSRD description. This vector r

. f represents the

functionality f offered by a registry r

and serves

as input for our community building approach. We

use these WSRD vector representations to build the

communities of registries. For this, we developed

CommunityBuilder a tool implementing our commu-

nity building approach. Using CommunityBuilder, we

construct 4 communities for the 7 registries.

Based on this organization, we now deploy the

overlay network of our marketplace. The JXTA P2P

platform

is suitable to simulate an overlay network

organized as communities above the physical reg-

istries network that we have put in place at the pre-

vious steps. A registry community is then viewed as a

Peer Group

where each service registry is mandated

by a JXTA peer. We developed a CommunityConsole

to deploy our test-bed on the JXTA platform.

To test the feasibility of our registries organization

management approach, we implemented a Commu-

nity Manager. This tool graphically simulates a net-

work of registries organized as communities and im-

plements our algorithms for managing the registries

organization in the marketplace. We used this tool to

simulate a network of communities of registries and

introduce changes to test our managing operations.

6 RELATED WORKS

In this paper, we presented our approach for building

a uniﬁed marketplace for Cloud services by virtually

organizing Cloud service registries as registry com-

munities. We also speciﬁed the needed management

mechanisms to guarantee the evolution of a commu-

nity organization. As far as we know, this aspect has

not been studied before in Cloud environments. Thus

we provide in this section a state of the art study of

Web services registries organization in service ori-

ented environments. Several Web services discovery

systems rely on a distributed registry environment to

http://jxta.kenai.com/

Peer Group is a JXTA concept deﬁned as a set of peers

offering a speciﬁc service.

overcome the problems related to single centralized

registry based discovery (bottlenecks, single point of

failure ...). Our literature review yielded a good num-

ber of research projects that used distributed registries

such as (Sivashanmugam et al., 2004; Pilioura and

Tsalgatidou, 2009; Ayorak and Bener, 2007; Du et al.,

2005), but as far as we know, none of them have pro-

moted the idea of using a functionalitybased semantic

description model to organize Web service registries.

MWSDI (Sivashanmugam et al., 2004) propose

a distributed Web services registry infrastructure for

Web services discovery. In this work, the registry net-

work is structured into federations using a specialized

ontology called the Registries Ontology. This ontol-

ogy allows grouping different Web service registries

based on their business domain as the authors map

each registry to one or several nodes in the Registries

Ontology. Associating with each registry a speciﬁc

domain or a group of domains from the Registries On-

tology enhances Web services discovery. In (Pilioura

and Tsalgatidou, 2009), the authors also propose a

distributed registry infrastructure called PYRAMID-

S, where Web service registries are categorized by

concepts token from a specialized ontology.

In (Ayorak and Bener, 2007), an hybrid architec-

ture for Web services discovery in distributed reg-

istry environment is proposed. Unlike previous ap-

proaches, the structure of the network is build based

on the business domain of the services advertised by

the registries: services operating in similar domains

are placed in the same group.

Authors in (Du et al., 2005) present Ad-UDDI; an

active and distributed infrastructure for organizing a

network of UDDI registries. The proposed infrastruc-

ture is formed by three layers: The root registry layer

contains an Ad-UDDI registry containing information

about the registries forming the network of registries,

the business service registry layer contains the net-

work of registries and the service layer contains the

services advertised by these registries. Authors orga-

nize statically their business service registry layer by

using a taxonomy (GICS

) classifying different real

industrial sectors, industries, sub-industries and in-

dustry groups.

To sum up, our literature review yielded a good

number of research projects that used distributed reg-

istries, but as far as we know, none of them have pro-

moted the idea of using a functionality-based seman-

tic description model to organize Web service reg-

istries. Compared to the above literature review, our

contributions can be summarized in three main points:

1. Our registry description and organization is based

only on implicit knowledge using existing adver-

GICS : Global Industry Classiﬁcation Standard.

CLOSER2012-2ndInternationalConferenceonCloudComputingandServicesScience

256

tised service descriptions. Thus, our approach is

self-contained within the Cloud service discovery

process, independent from any explicit or human

centric or error prone knowledge.

2. Our approach proposes a functionally-based or-

ganization of service registries into communities.

Such an organization enhances the services dis-

covery process since the registries will be grouped

according to the functionalities proposed by their

advertised services, and thus a service requester’s

query can be guided to the adequate registry clus-

ter for his needs.

3. By fuzzy clustering service registries according to

their service descriptions, our approach has intrin-

sically the means for a dynamic, ﬂexible and au-

tomatic management of service evolutions.

7 CONCLUSIONS

We presented an approach for building and manag-

ing a marketplace for Cloud services. In the pro-

posed marketplace, the Cloud services registries are

organized as communities. These communities are

implicitly and automatically created using the reg-

istries WSRD descriptions. Our registries organi-

zation approach uses a functionality-driven cluster-

ing and organizes registries according to the func-

tionalities of the Cloud services they advertise. This

functionality-driven organization of the marketplace

registries enhances services discovery. In fact, a ser-

vice requester’s search space can be reduced to the

registry community advertising services offering the

needed functionalities. We deﬁned the required man-

agement operations to ensure the evolution of the reg-

istries organization. We also demonstrated and de-

ployed the proposed approach on top of a JXTA-based

distributed registries environment and implemented

our management operationto test the feasibility of our

management approach.

Our future research work will focus on *aaS ser-

vices speciﬁcation and discovery. In this work, we

have used traditional Web service description in or-

der to specify the WSRD description of a Cloud ser-

vices registry. However, the Open Cloud Computing

Interface (OCCI) (Metsch et al., 2010) has emerged as

standard to describe *aaS services speciﬁc features.

Originally initiated to create a remote management

API for IaaS model based services, allowing for the

development of interoperable tools for common tasks

including deployment, autonomic scaling and moni-

toring, it now can be used to severe other *aaS ser-

vices as well. As part of our perspectives, we aim

in our future work to extend our approach in order to

take into account the OCCI speciﬁcation.

We also plan to deﬁne the Cloud services discov-

ery system component to test the efﬁciency of our

functionality driven organization of the Cloud ser-

vices registries for services discovery. This discovery

system will use the functional requirements of a ser-

vices requester to reduce the search space by selecting

adequate communities and in a second time will com-

bine his functional and non-functional requirements

for an efﬁcient services selection.

REFERENCES

Ayorak, E. and Bener, A. B. (2007). Super peer web service

discovery architecture. In ICDE, April 15-20, 2007,

Istanbul, Turkey, pages 1360–1364.

Bouchaala, O., Sellami, M., Gaaloul, W., Tata, S., and

Jmaiel, M. (2011). Modeling and managing commu-

nities of web service registries. In WEBIST, Noordwi-

jkerhout, The Netherlands, May 6-9, 2011.

Cliquet, G. (2011). Method of innovation in the era of Web

2.0. Phd thesis, Paris Institute of Technology.

De Bruijn, J. et al. (2005). Web service modeling ontology

(wsmo). Technical report.

Ding, C. H. Q., He, X., Zha, H., Gu, M., and Simon, H. D.

(2001). A min-max cut algorithm for graph partition-

ing and data clustering. In ICDM’01, San Jose, Cali-

fornia, USA.

Du, Z., Huai, J., and Liu, Y. (2005). Ad-uddi: An active and

distributed service registry. In TES, Trondheim, Nor-

way, September 2-3, 2005, Revised Selected Papers.

Lausen, H. and Farrell, J. (2007). Semantic annota-

tions for WSDL and XML schema. W3C recom-

mendation, W3C. http://www.w3.org/TR/2007/REC-

sawsdl-20070828/.

Martin, D. et al. (2004). Owl-s: Semantic markup for web

services. Technical report.

Metsch, T., Edmonds, A., and Bayon, V. (2010). Using

cloud standards for interoperability of cloud frame-

works. Technical report.

Nie, F., Ding, C., Luo, D., and Huang, H. (2010). Im-

proved minmax cut graph clustering with nonnegative

relaxation. In ECML PKDD 2010, Barcelona, Spain,

September 20-24, 2010.

Pilioura, T. and Tsalgatidou, A. (2009). Uniﬁed publication

and discovery of semantic web services. ACM Trans-

actions on the Web (TWEB), 3(3).

Sellami, M., Gaaloul, W., and Tata, S. (2011). An Implicit

Approach for Building Communities of Web Service

Registries. In iiWAS’2011, 5-7 December, 2011, Ho

Chi Minh City, Vietnam.

Sivashanmugam, K., Verma, K., and Sheth, A. P. (2004).

Discovery of web services in a federated registry en-

vironment. In ICWS’04, San Diego, California, USA.

The Mckinsey Quarterly (2010). How IT is managing new

demands: McKinsey Global Survey results.

TOWARDSAUNIFIEDMARKETPLACEFORFUNCTIONALITY-BASEDCLOUDSERVICEDISCOVERY

257