Opportunistic Software Composition: Beneﬁts and Requirements

Charles Triboulot

1,2

, Sylvie Trouilhet

2,3

, Jean-Paul Arcangeli

and Fabrice Robert

Sogeti High Tech, Aeropark 3 Ch. Laporte 31300 Toulouse, France

Université de Toulouse, UPS, IRIT, 118 route de Narbonne 31062 Toulouse, France

IUT A, Av. G. Pompidou CS 20258 81104 Castres, France

Keywords:

Opportunism, Software Component, Automatic Bottom-up Composition, Ambient Systems.

Abstract:

Traditional software development relies on building and assembling pieces of software in order to satisfy ex-

plicit requirements. Component-based software engineering simpliﬁes composition and reuse, but software

adaptation to the environment remains a challenge. Opportunistic composition is a new approach for building

and re-building software in open and dynamic contexts. It is based on the ability to compose software com-

ponents in a bottom-up manner, merely because they are available at a point and not because the construction

of a speciﬁc software has been demanded. In this way, software emerges from the environment. This paper

analyzes the advantages of such an approach in terms of ﬂexibility and reuse, along with the requirements that

an infrastructure supporting opportunistic composition should satisfy: it should be decentralized, autonomic,

and dynamically adaptive. The state of the art of automatic software composition shows that few solutions are

actually bottom-up, and that none of them fully satisﬁes the requirements of opportunistic composition.

1 INTRODUCTION

When rationalizing software development, two major

points should be considered: productivity and reuse

on the one hand, capability of evolution on the other

hand. Regarding these points, component-based soft-

ware engineering improves object-based software en-

gineering, mainly due to the explicitation of the re-

quired interfaces at the same level as the provided

ones.

In traditional software engineering, composition

is triggered when a need is made explicit and when the

building of the software is demanded. Components

are identiﬁed, then selected among existing ones or

developed one by one, in order to compose the appli-

cation. This approach is called top-down because the

development is driven by the satisfaction of explicit

requirements. In a general way, meeting requirements

is a constant challenge for software engineers.

Once developed, applications are deployed in so-

cial and technical contexts that are changing. There-

fore requirements change, thus applications must be

ﬂexible enough and adapt. But, even if pieces of so-

lution exist, developers generally struggle to guess

possible evolutions and to design points of variabil-

ity in software. Because stakeholders’ requirements

are considered as a foundation at the early steps of

traditional software engineering and architectural de-

sign processes, considering changes remains a major

challenge in software engineering and software archi-

tecture.

Composition and adaptation issues are particu-

larly crucial for ambient, ubiquitous and mobile sys-

tems. Given the large number of networked devices

that populate ambient environments, as each of them

may host software components, there are possibly nu-

merous applications and services that can be built and

provided to people in order to enrich their environ-

ment. Ambient intelligence can be measured in the

yielded additional value while limiting the active par-

ticipation of users. Ambient systems, with their dy-

namics, must be proactive and intelligent enough to

adapt according to the context in order to react to

some situations, anticipate user needs, or go further

than the predeﬁned behavior.

1.1 Opportunistic Composition

Our work aims to explore a new approach for the

engineering of component-based software applica-

tions. Our approach, called “opportunistic composi-

tion”, differs from the traditional one because com-

position is no longer driven by explicit requirements

but opportunity-driven. Opportunistic composition is

426

Triboulot C., Trouilhet S., Arcangeli J. and Robert F..

Opportunistic Software Composition: Beneﬁts and Requirements.

DOI: 10.5220/0005558404260431

In Proceedings of the 10th International Conference on Software Engineering and Applications (ICSOFT-EA-2015), pages 426-431

ISBN: 978-989-758-114-4

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

the ability to assemble software components merely

because they are available in the surrounding envi-

ronment. In this way, composition is directed by the

execution context and by the opportunity to assemble

components that are ready for composition (but which

have not been developed speciﬁcally for the applica-

tion under construction), and no longer by the neces-

sity of satisfying explicit requirements. Thus, appli-

cations emerge from the environment and can evolve

afterwards by dynamic (re)composition based on new

opportunities.

Opportunism in software construction is a new

idea, which has been little, if at all, studied. A paper

on this subject has been presented at the French con-

ference UBIMOB (Ubiquity and Mobility) in 2011

(Vergoni et al., 2011). In (Conti and Kumar, 2010),

opportunistic computing has been deﬁned as a new

computing paradigm derived from ad hoc and oppor-

tunistic networking tightly coupled with social net-

working, which aims at exploiting available resources

in an environment to provide distributed collabora-

tive computing services and applications; the main

challenges that are stated concern networking: inter-

mittent connectivity, delay tolerance, protocol hetero-

geneity.

Of course, opportunistic composition raises sev-

eral issues. The main ones relate to what can be com-

posed (heterogeneity, composability...), what should

be composed and how combinatorial complexity can

be controled, context-awareness, the semantics of the

resulting application, its quality attributes (security,

reliablility, efﬁciency...), its viability and usefulness,

and how to automatically perform opportunistic com-

position.

1.2 Contents and Plan

This paper presents an analysis of the opportunistic

composition approach. It discusses its advantages in

terms of ﬂexibility and reuse, and focuses on the fea-

sibility and on the requirements that an infrastruc-

ture should satisfy in order to support opportunistic

composition and allow the emergence of component-

based software systems. Finally, the state of the art of

automatic composition is analyzed in relation to these

requirements. However, the paper does not present

any architecture or experimental results. The con-

tribution sets in the analysis of the opportunistic ap-

proach, in the identiﬁcation of the requirements of its

realization, and in the analysis of the state of the art.

The paper is organized as follows. Section 2

presents a real-world illustrating scenario which is

useful for highlighting different situations of compo-

sition. In Section 3, opportunistic composition is de-

ﬁned more formally and the requirements for its re-

alization are exposed. Section 4 analyzes the state of

the art of automatic composition. To conclude, in Sec-

tion 5, we give some directions for the realization of

opportunistic composition.

2 SCENARIO

This section presents a scenario in which opportunis-

tic composition assists users in their life and their job,

and provides applications depending on the context.

Here is Plip, an engineer who works in a secured ex-

periment room of her company. She uses a robotic

arm and a laser to project rays on a ﬂat surface.

The scenario consists of four acts and shows sev-

eral situations of opportunistic composition. It sup-

ports an analysis presented in Section 3. It underlines

how an opportunistic composition system can react to

some situations when they occur in a dynamic envi-

ronment. Only the system’s behavior is described, not

its implementation. It is worth pointing out that com-

positions become possible without the necessity to ex-

plicitly specify user needs. They are realized merely

because the components are available in the environ-

ment and because assembling them may be relevant.

Plip does have requirements, but not necessarily ex-

plicited, and nevertheless the system provides an an-

swer to them in a proactive way.

To carry out this scenario, we assume that the

components are interoperable. Issues, such as secu-

rity or privacy, are set aside in this work.

Act I - Plip is working in her ofﬁce, on a new

experimental model stored on her PC. Once the

model is completed, she runs an application on

her tablet computer that provides estimated results

according to the model. This application detects that

a component implements Plip’s model, thanks to a

Wi-Fi connection between Plip’s PC and the tablet,

and proposes the use of this component. Since it

matches her requirement, she accepts. Then, the

application runs.

Act II - Since the estimated results are satisfactory,

Plip is ready to perform the genuine experiment and

compare the estimated results to the real ones. She

goes to the experiment room with her tablet. The

experiment room has been secured recently: it is

isolated from Wi-Fi signals, but Plip does not know

that. She enters the room with her tablet which is

still connected by Wi-Fi with her PC. As the Wi-Fi

component of the tablet is no longer working, a new

composition is done, which involves a Bluetooth

OpportunisticSoftwareComposition:BenefitsandRequirements

427

component which radio waves cross the walls of the

secured room. This automatic composition preserves

the communication link between the tablet and the

outside PC. Thus, the application is still running and

connectivity issues are imperceptible for Plip.

Act III - Plip starts the experiment. She turns on

the different devices, oscilloscopes and screens, in

the experimental room. She also turns on the central

PC to control the experiment. The different devices

are assembled automatically in order to propose an

adequate experimental environment (see Figure 1).

This is possible because the opportunistic composi-

tion system has learned, from previous experiments,

the context in which such a conﬁguration is relevant.

Furthermore, one of the oscilloscope in the room has

been upgraded, and now proposes a greater quality of

service than the other oscilloscopes. Consequently, it

is used in the assembly.

Act IV - While the experiment is running, the cen-

tral PC suddenly breaks down. Unfortunately, it em-

beds several indispensable components for the exper-

imental assembly, and plays an essential role in the

command of the robotic arm ("Exp Leader" compo-

nent). The PC is also used to collect the results of

the experiment, and then analyze them. To allow the

experiment to go on without waiting for repairs, the

opportunistic composition system searches for other

suitable components. In this case, the system may

have several choices and must efﬁciently select and

perform the most relevant ones (for instance an older

“Exp Leader” component and a “Stock” component

embedded in another device), with the help of context

information.

3 ANALYSIS OF THE APPROACH

From this scenario, we extract three main advantages

of the opportunistic approach compared to top-down

and traditional approaches: proactiveness, ﬂexibility

and genericity. Then, we enumerate the functional

and extrafunctional requirements which seem to be

fundamentalfor an opportunistic composition system.

3.1 Beneﬁts

Proactiveness - While Plip carries on her daily tasks,

she never expresses explicitly her needs. However,

the system proposesand maintains useful applications

according to the current situation without relying on

explicit user needs (explicit request of a component

by the user is possible but not mandatory).

The ﬁnal assembly is characterized as emergent,

because neither the developer, the user, the system

nor the components have a knowledge of it before

its bottom-up fabrication. The key is to select useful

compositions with methods related to learning or con-

text awareness. The development of a proactive sys-

tem, which does not demand requirement formaliza-

tion and proposes relevant compositions, is promis-

ing in an ambient scope: in such undetermined and

unpredictable environments, it is hard to express, or

even identify, what the user needs can be.

Flexibility - Several adaptive reactions of the sys-

tem can be observed in the scenario. Thanks to oppor-

tunistic composition, applications using appropriate

components may replace ineffective and useless ones.

The adaptive reactions are not explicitly expressed,

but are the result of bottom-up decisions. On the one

hand, this approach supports openness (components

may appear or disappear from the local environment).

On the other hand, applications are resilient (they are

continually maintained and enhanced). Here, initial

composition and resilience are supported by the same

opportunistic approach.

Genericity - No hypothesis is made on the pres-

ence of certain components, their business domain,

their type or number. Most of all, the system can

work without the knowledge of the user needs. So, an

opportunistic composition system can be generic and

useful in any (dynamic) environment. The system as-

sisting Plip could also be used in various contexts (e.g.

smart-home, guidance, monitoring...).

However, these beneﬁts are guaranteed only if

several open problems are resolved. How is the rel-

evance of a composition precisely evaluated? How

is the usability of an emergent application guaranteed

without formalization of needs? How is the seman-

tics of the application presented to the user? How

is the combinatorial complexity mastered? The next

sections will identify requirements for the realization

of opportunistic composition. Meeting these require-

ments should bring answers to those questions.

3.2 Functional Requirements

From the different composition situations underlined

in the scenario, ﬁve functional requirements can be

extracted. In the following enumeration, each of them

is presented through the event that triggers the new

composition.

1) Request of a Component (fRC) (Act I). A com-

ponent is requested by a user or the system itself, for

being a part of an application. Thus, this component

tries to compose itself as a priority and properly fulﬁll

its functionality. This is a way to implicitly express a

ICSOFT-EA2015-10thInternationalConferenceonSoftwareEngineeringandApplications

428

Central PC

Labo

Stock

Analysis

Robot Arm

Oscilllo 1.0

Exp Leader

Oscilllo 3.0

Mod

Cmd

Res

Model

Plip's PC

Mod

Tablet

Mod Estim

Mod

Anti-Wi-Fi

Wall

I/O Wi-Fi

I/O BlueTooth

I/O Wi-Fi

Figure 1: Component view of Act III.

user need without formalizing it.

2) Context Evolution (fCE) (Act III and Act IV).

Some external situations may trigger a system reac-

tion. Context awareness allows the system to detect

these situations, while learning capabilities allow to

memorize them. So, the system can properly react to

environment evolution with the proper responses.

3) Appearance of a Component (fAC) (Act III). Due

to the high dynamics of the environment, one or sev-

eral components may be discovered by other compo-

nents. This event can follow the start-up of a device

or a mobility situation. The appearing component can

be included in an existing assembly or trigger a new

composition.

4) Disappearance of a Component (fDC) (Act II

and Act IV). A component may disappear from the

environment, in case of mobility, obsolescence or fail-

ure. The system should detect this situation and main-

tain the applications which were using the missing

component by proposing new efﬁcient compositions.

Note that the disappearance of a component may be

progressive and can be detected and anticipated be-

fore it actually happens, especially during a situation

of mobility.

5) Upgrade (fU)(Act III). Some intrinsic properties

of components may also dynamically evolve. For ex-

ample, quality of service or interface proﬁles may be

modiﬁed at runtime, further to an upgrade. These

modiﬁcations are most likely to trigger new compo-

sitions. An opportunistic composition system should

be able to identify any of these situations and react

accordingly. It must also take into account that they

can occur in any quantity and in any order. Previous

situations are therefore considered as functional re-

quirements for an opportunistic composition system.

3.3 Extrafunctional Requirements

The aim of this section is to identify the extrafunc-

tional requirements for the design of a system able to

perform opportunistic composition. They represent

generic issues the opportunistic system has to deal

with, but do not make any assumptions on the meth-

ods and/or technologies used.

Seven requirements can be extracted and formal-

ized. They can be used for the evaluation of different

composition systems.

1) Decentralization (efDc): Ability to propose a de-

centralized process in which each task is attributed

to an autonomous entity in a synchronous or asyn-

chronous way. Indeed, the system must work at a lo-

cal level and consider the local neighborhood of com-

ponents but not the whole environment. Furthermore,

a local view is the natural way to perform bottom-up

composition and obtain emergent results. The rele-

vance of decentralized design is conﬁrmed by the in-

ability for a central entity to handle high dynamics,

bottlenecks and/or failures.

2) Dynamic Adaptation (efDA): Ability to handle

undetermined and unpredictable environments. The

opportunistic composition system runs in highly dy-

namic and unpredictable environment: components

may appear or disappear at runtime, even context and

users’ actions can change. Thus, it should be open

and able to generic and efﬁcient adaptation.

3) Combinatorial Optimization (efCO): Ability to

handle a fair amount of possible compositions among

available components. Ignoring this point may lead to

efﬁciency and speed issues for the composition pro-

cess. Thus, the system must handle this problem and

efﬁciently select the most useful compositions using

discriminating strategies.

4) Recomposition (efRc): Ability to dynamically

maintain and/or enhance existing compositions. In

order to support the resilience of applications, com-

ponents may be added or replaced, or the entire as-

sembly may be challenged.

5) Learning and Context Awareness (efLC): Abil-

ity to take into account past activities and context to

perform relevant compositions.

6) Utility of the Result (efUR): Ability to guaran-

tee a useful application as a result of the composition.

The emergent result must be useful and satisfy im-

plicit needs. Its utility can be automatically veriﬁed,

OpportunisticSoftwareComposition:BenefitsandRequirements

429

Table 1: Functional and extra-functional requirements.

fRC fCE fAC fDC fU efDc efDA efCO efRc efLC efUR efSUN

Vallée 2005 x x + ++ + - + + - -

Grondin 2006 x x x - - + - ++ + ++ - -

Desnos 2007 x x - - + + ++ - - - +

Bartelt 2008 x x ++ - - - - - - - +

Rouvoy 2009 x x x - ++ - - + - + -

Guidec 2010 x x x ++ ++ - - - - ++ - -

Sykes 2011 x x x + + - ++ - - -

Bonjean 2013 x x x x x ++ + ++ + + + -

in local or global scope, a posteriori by feedback on

the assembly, or a priori by identiﬁcation of promis-

ing compositions.

7) Silence of User Needs (efSUN): Ability of the sys-

tem to operate without the preliminary expression of

user needs. This requirement is a main one in order

to let the system propose emergent solutions without

explicit requirements or user needs.

Next section confronts the state of the art with the

functional and extrafunctional requirements.

4 STATE OF THE ART

We analyzed eight research works on component au-

tomatic composition in order to determine if and how

they could answer the requirements of the opportunis-

tic approach (Vallée et al., 2005; Grondin et al., 2006;

Desnos et al., 2007; Bartelt et al., 2008; Rouvoy et al.,

2009; Guidec et al., 2010; Sykes et al., 2011; Bon-

jean et al., 2013). Table 1 indicates both if the solu-

tion could handle the functional requirements (if the

system could detect and then properly handle the sit-

uation, the corresponding box contains a cross) and

the extrafunctional requirements (evaluated from - -

to ++).

None of the solutions has (x) or (++) for all the

requirements. The fDC (Disappearance of a Compo-

nent) requirement is globally met. This is due to the

systematic self-adaptation aspects developed in the

reviewed works. Likewise, some extrafunctional re-

quirements, such as efDc (Decentralization) or efDA

(Dynamic Adaptation), are admitted as essential for

automatic composition and a lot of solutions exists.

However, efCO (Combinatorial Optimization) and

efLC (Learning Context), although they are extremely

important in dynamic environments, have yet to be

considered. A similar assessment can be made for

functional requirements; situations such as fU (Up-

grade) and fCE (Context Evolution) also seem to be

hardly anticipated. The last issue is the status of the

expression of user needs, many systems demanding

their formalization before making any composition

decision. This expression is not mandatory, but it can

be noted that efUR (Utility of the Result) and efSUN

(Silence of User Needs) are not satisﬁed together (see

Table 1).

5 CONCLUSION AND

PERSPECTIVES

In this paper, we have deﬁned opportunism as an ap-

proach for building and re-building software by com-

posing available software components. We have il-

lustrated the advantages of this approach: ﬂexibility

and adaptiveness. In our opinion, opportunistic soft-

ware composition is not limited to ambient systems:

for example, it could assist software engineers, which

have to compose reusable components as part of large

libraries or to adapt component-based software, by

proposing them relevant compositions.

Opportunistic composition is a promising generic

approach, well-adapted to highly dynamic and open

environments, which disregards formalization of user

needs. However, the link between needs and compo-

nent assemblies should be considered through learn-

ing, evaluation of utility and user’s feedback, in or-

der to control the emergence. Thus, opportunistic

composition may be a basis for design of intelligent

systems, which proactively provide emergent applica-

tions adapted to (possibly unforeseen) situations, and

anticipate user needs.

Systems that support opportunistic software com-

position should meet several requirements, in partic-

ular those related to combinatorial complexity and

relevance of the emergent applications. In order to

face these challenges, we currently develop a solution

based on multi-agent systems (MAS) which offer sev-

eral advantages from an architectural point of view

(Arcangeli et al., 2014), and precisely on cooperative

MAS (Georgé et al., 2011), which support the emer-

gence of functions through local interactions between

cooperative agents. In order to select relevant compo-

ICSOFT-EA2015-10thInternationalConferenceonSoftwareEngineeringandApplications

430

sitions, agents have learning capabilities and are able

to consider user’s feedback.

REFERENCES

Arcangeli, J.-P., Noel, V., and Migeon, F. (2014). Software

Architectures and Multi-Agent Systems. In Oussalah,

M. C., editor, Software Architectures, volume 2, chap-

ter 5, pages 171–208. Wiley-ISTE.

Bartelt, C., Fischer, B., and Rausch, A. (2008). To-

wards a Decentralized Middleware for Composition

of Resource- Limited Components to Realize Dis-

tributed Applications. In 3rd Int. Conf. on Pervasive

and Embedded Computing and Communication Sys-

tems (PECCS 2013), pages 245–251.

Bonjean, N., Gleizes, M.-P., Maurel, C., and Migeon, F.

(2013). Forward Self-combined Method Fragments.

In Agent-Oriented Software Engineering XIII, number

7852 in LNCS, pages 168–178. Springer.

Conti, M. and Kumar, M. (2010). Opportunities in oppor-

tunistic computing. Computer, 43(1):42–50.

Desnos, N., Huchard, M., Urtado, C., and Vauttier, S.

(2007). Automated and Unanticipated Flexible Com-

ponent Substitution. In Proc. of 10th Int. Symp. on

Component-Based Software Engineering.

Georgé, J.-P., Gleizes, M.-P., and Camps, V. (2011). Co-

operation. In Di Marzo Serugendo, G., Gleizes,

M.-P., and Karageogos, A., editors, Self-organising

Software, Natural Computing Series, pages 7–32.

Springer.

Grondin, G., Bouraqadi, N., and Vercouter, L. (2006).

MaDcAr: An Abstract Model for Dynamic and Au-

tomatic (Re-)Assembling of Component-Based Ap-

plications. In Component-Based Software Engineer-

ing, number 4063 in LNCS, pages 360–367. Springer-

Verlag.

Guidec, F., Sommer, N. L., and Mahéo, Y. (2010). Oppor-

tunistic Software Deployment in Disconnected Mo-

bile Ad Hoc Networks. Int. Journal of Handheld Com-

puting Research, 1:24–42.

Rouvoy, R., Barone, P., Ding, Y., Eliassen, F., Hallsteinsen,

S., Lorenzo, J., Mamelli, A., and Scholz, U. (2009).

Music: Middleware support for self-adaptation in

ubiquitous and service-oriented environments. In Soft-

ware Engineering for Self-Adaptive Systems, pages

164–182. Springer-Verlag.

Sykes, D., Magee, J., and Kramer, J. (2011). FlashMob:

Distributed Adaptive Self-Assembly. In Proc. of the

6th Int. Symp. on Software Engineering for Adaptive

and Self-Managing Systems, pages 100–109.

Vallée, M., Ramparany, F., and Vercouter, L. (2005). A

Multi-Agent System for Dynamic Service Composi-

tion in Ambient Intelligence Environments. In The

3rd International Conference on Pervasive Comput-

ing (PERVASIVE 2005), pages 175–182.

Vergoni, C., Tigli, J.-Y., Rey, G., and Lavirotte, S. (2011).

Construction Bottom-up d’applications ambiantes en

environnements partiellement connus a priori. In

7èmes Journées Francophones Mobilité et Ubiquité

(UBIMOB 2011).

OpportunisticSoftwareComposition:BenefitsandRequirements

431