Systems

Volker Gruhn,Clemens Sch¨a fer

University of Leipzig

Faculty of Mathematics and Computer Science

Chair for Applied Telematics/e-Business∗

Klostergasse3,04109Leipzig,Germany

{gruhn,schaefer}@ebus.informatik.uni-leipzig.de

January12,2004

Abstract

Mobile software applications have to meet new re-quirements directly arising from mobility issues. To address these requirements at an early stage in development,an architecture description language (ADL)is proposed,which allows to manage issues like availability requirements,mobile code,security, and replication processes.Aspects of this ADL, Con Moto,are exemplified with a case study from the insurance sector.

1Motivation

We can observe in our daily life,that existing and emerging communication technologies and the growing availability of powerful mobile communi-cation and computing devices are the enablers for software systems with mobile parts.Regardless whether we look atfleet management systems or point-of-sale systems,all applications with mobile parts have in common that the point where the in-formation technology is used is moved towards the point where the business process is really executed. Although these mobile applications bring new requirements with them,we usually apply devel-opment methods and modeling methods used for ‘conventional’applications.In this work,we will ∗The chair for Applied Telematics/e-Business is en-dowed by Deutsche Telekom AG.

show how the development of mobile applications can benefit from an approach explicitly addressing mobility issues.

The remainder of this paper is organized as fol-lows.In section2we motivate the necessity of an architecture description language(ADL)for mobile systems.Section3contains an application of this mobile ADL to a case study from the insurance sec-tor.Finally,the conclusion in section4rounds out the paper.

2Con Moto as Mobile ADL As discussed in the previous section,more and more software systems are determined by some sort of mobility.In the following we call a software system part which is used at different locations(without these locations being known in advance)a“mobile component”.We call a business process mobile,if its execution depends on a mobile component.For a more precise definition of the term“mobile”we refer to[RPM00].

If a software system contains mobile components, its management has to deal with a number of tasks and challenges which do not occur in the manage-ment of systems without mobile components.Ex-amples of such tasks and challenges are:

•A functionality which is crucial for a number

of mobile business processes has to be available 1on client systems.Usually it is not sufficient to keep it running at a central site and to allow mobile access to it.Thus,the availability re-quirement for this functionality may clash with service level agreements which can be obtained for telecommunication infrastructure.

•If huge amounts of data are needed in a mobile business process,then it is usually necessary to provide a only readable copy of these data at the mobile location.This immediately raises the question whether or not these data can be downloaded when needed or if they should be installed at the remote location in advance.•If certain functionality are to be handled con-fidentially,then this may collide with mobil-ity requirements,simply,because components implementing this functionality should not be made available at local devices.

•Updates and patches of mobile components de-mand either precautions from a software up-date process point of view(e.g.to register which versions of which components are in-stalled where)or they require that mobile com-ponents are not made persistent at mobile lo-cations.

Further challenges exist in the context of compati-bility between mobility on the one hand and other non-functional requirements(robustness,scalabil-ity,security,traceability,etc.)on the other hand. Summing this up,requirements with respect to the mobility of software systems determine not only substantial parts of the software processes for the development and maintenance of these systems,but they also have an important impact on the archi-tecture of these systems.

Our approach to an architecture centric software development of mobile software systems is to ex-press all aspects of software systems’mobility in a software architecture.This software architec-ture is described from different views(application view,software view,system view[GT00]).All these view onto the architecture should clearly de-scribe which components and which business pro-cesses are mobile,which kind of mobility applies to them(ranging from mobile web-based access to central systems to mobile code in the sense of

[HBSG99,WBS02])and which side effects a change of mobility properties could have.

Our approach is based on the architecture de-scription language Con Moto.Describing all three views onto a software architecture in terms of Con Moto means to grasp all aspects of mobility.A Con Moto-based description of a software architecture is used for different purposes:

•understanding the architecture of the system,

paying particular emphasis on mobility issues •using the architecture description for configu-

ration management and change management

tasks

•scheduling work including tasks and packages

related to infrastructure support enabling mo-

bility

•analysis of the architecture,in particular for

the identification of performance bottlenecks

and incompatibilities between mobility re-

quirements and non-functional requirements. In the next section we will introduce the key ele-ments of Con Moto along the lines of a concrete example.This examples covers a point-of-sale soft-ware system for insurance agents.We put empha-sis on those elements of Con Moto which focus on mobile components and mobile business processes. We are aware of the fact,that this introduction does not give a complete overview about Con Moto, but it hopefully will provide sufficient insights into the role a Con Moto based architecture description plays in an architecture centric development of mo-bile system.

During the introduction of Con Moto we refer to other ADLs(a good overview can be found in the paper of Medvidovic and Taylor[MT00])wherever appropriate in order to point out what the bene-fits of considering mobility asfirst class property of software systems look like.

3A Case Study:uniPOS

In the insurance sector,insurance andfinancial products are more and more sold actively to the customer instead of being bought by the customer. From the viewpoint of insurance agents,this pro-duces new requirements like rapidity,actuality,and 2demand for diversity(for services as well as for products),which have to be fulfilled by sales sup-porting systems.

Existing sales support systems are mostly real-ized as fat client applications,which are extensive and allow autarkic(offline)execution.However, their architecture implies a number of problems (e.g.replication issues)and fails to meet the needs described above.

Since a mobile system is predestinated to provide the above-mentioned sales support,we decided to create the case study uniPOS.

The uniPOS system supports different target groups.Tied intermediaries,brokers,multiple agents,policy holders as well as staffmembers of the insurance company are provided with an op-timized view on stock data and support for their daily business processes by

•integration of different lines of business in one view,

•support for different sales channels,and •online deal-or case-closing transactions. Additionally,uniPOS provides fast introduction of new products and avoids multiple plausibility checks.

Figure1:Offer Component in uniPOS

Figure1shows a part of the uniPOS system,ex-pressed in Con Moto.There are two devices,

namely the uniPOS Server and the uniPOS Client, both represented by rectangles with triangles in their edges.On these devices,there are several components,namely Offer,OfferLogic and Offer-Storage,which are represented by ovals.The lines between the components and the components and the devices depict connectors,meaning that e.g. Offer can use services from OfferLogic,since the two devices are connected by a attributed connector representing a GSM telecommunications channel.

With this representation(and the formalism be-hind)it is possible to judge about the operational availability of e.g.the Offer component.If we re-quest a high availability,a Con Moto checker can show the clash between the GSM channel(which is not always available)and the requirement of avail-ability.Hence,availability requirements for sys-tems and subsystems can be answered by analyzing a system depicted in Con Moto.

Since we want to use Con Moto during the com-ponent based design of mobile systems,the need for a graphical representation is clear.Although the ADL Wright[AG97]can model distributed sys-tems,it lacks a graphical representation.Other ADLs allow the graphical modeling of distributed systems.Darwin[MDEK95]is based on the pi calculus.From these precise semantics,Darwin’s strength lies in modeling hierarchically constructed systems and distributed systems.Rapide[LV95] provides comprehensive modeling,analysis,simu-lation and code generation capabilities.However, as it is strictly event based,it clashes with the service-oriented nature of component-based sys-tems.C2SADL[MRT99]is an ADL also suitable for dynamic systems.

However,none of these ADLs implies constructs similar to our notion of the device and the at-tributed channel.These ADLs also do neither support such constructs directly nor support non-functional properties,which could be used to ex-tend the existing ADLs in a way to make them suitable for mobile systems.

Weaves[GR91]addresses issues that seem to be related to mobile systems like dynamic rearrange-ment.However,the concept of the stream-like weaves do notfit into the component-based ap-proach that is used for mobile systems today.A recent contribution to thefield of ADL research is xADL2.0[DvT02].It is an XML-based ADL that is highly extensible.Although it does not support 3the mobility issues we want to point out,that ideas

from it can be valuable contributions to Con Moto. We now continue with our example.Before,we detected a system which fails to meet the avail-ability requirements.Con Moto also supports the modeling of a system variant that is superior at this point.

Figure2:Mobile Code

In Figure2two equivalent variants of uniPOS are depicted that make use of mobile code.The compo-nents that shall be mobile have been marked with an‘M’.But to check whether a system configura-tion with this kind of mobile component is possible, we need to introduce the notion of execution envi-ronment,depicted by a rounded rectangle.This indicates that there are different execution envi-ronments like application servers,lightweight ap-plication servers etc.,that may be compatible in a way that they allow mobile components to move around the system in order to cope with availability problems.With the specified mobile components, the availability requirement for the component Of-fer can be satisfied,as the mobile components may travel from the server to the client. Furthermore,with this representation Con Moto allows the discussion of security aspects as it is able to determine which components on which execution areas and hence on which devices a component will be available.

However,just making a component mobile is not sufficient:in our example,the OfferStorage com-

Figure3:Replication

ponent has to work on some data from a database. This directly leads to replication problems in the mobile case.Hence,we have to express in our ar-chitecture,where replication has to be performed. Figure3shows two equivalent architectures this for the database with offer data.The replication ar-eas have been drawn as dotted rounded rectangles. The dotted line between them indicates that they belong together and that there is a replication re-lationship between them.

In Figure2and in Figure3,two equivalent vari-ants of the same system have been displayed.In both cases,the left variant(with the components OfferLogic and OfferStorage on the server)is suf-ficient,as the“mobile”cases(the right sides of the figures)are covered as well due to the specification of the execution environments and the replication areas.

4Conclusion

In this paper we discussed the idea of an archi-tecture description language matching the require-ments of mobile architectures.Due to the increas-ing mobility and distribution of business processes and supporting software systems,the mobility as-pect deserves emphasis and this emphasis should be reflected in the architecture description.Even though our ADL Con Moto is no formally defined 4yet and even though the language itself is still sub-ject to minor updates,it turned already out to be useful in managing the complexity of mobile archi-tectures.We consider this and—even more—the strong demand for architectures whose mobile as-pects are easy to recognize as encouragement on the further way to full syntax and semantics definition for Con Moto.

References

[AG97]R.Allen and D.Garlan.A Formal Ba-sis for Architectural Connection.ACM

Transactions on Software Engineering

and Methodology,6(3):213–249,1997. [DvT02] E.M.Dashofy,A.van der Hoek,and R.N.Taylor.An infrastructure for the

rapid development of xml-based archi-

tecture description languages.In Pro-

ceedings of the24th International Con-

ference on Software Engineering,pages

266–276.ACM Press,2002.

[GR91]Michael M.Gorlick and Rami R.Ra-zouk.Using weaves for software con-

struction and analysis.In Proceed-

ings of the13th international confer-

ence on Software engineering,pages

23–34.IEEE Computer Society Press,

1991.

[GT00]V.Gruhn and A.Thiel.Komponenten-modelle(in German).Addison-Wesley,

2000.

[HBSG99]Ophir Holder,Israel Ben-Shaul,and Hovav Gazit.Dynamic Layout of Dis-

tributed Applications in FarGo.In

Proceedings of the21st international

conference on Software engineering,

pages163–173.IEEE Computer Soci-

ety Press,1999.

[LV95] D.C.Luckham and J.Vera.An Event-Based Architecture Definition Lan-

guage.IEEE Transactions on Software

Engineering,21(9):717–734,September

1995.

[MDEK95]J.Magee,N.Dulay,S.Eisenbach,and

J.Kramer.Specifying Distributed Soft-

ware Architectures.In W.Schafer

and P.Botella,editors,Proc.5th

European Software Engineering Conf.

(ESEC95),volume9,pages137–

153,Sitges,Spain,1995.Springer-

Verlag,Berlin.

[MRT99]Nenad Medvidovic,David S.Rosen-

blum,and Richard N.Taylor.A

language and environment for

architecture-based software devel-

opment and evolution.In Proceedings

of the21st international conference

on Software engineering,pages44–53.

IEEE Computer Society Press,1999. [MT00]N.Medvidovic and R.N.Tay-

lor.A Classification and Compari-

son Framework for Software Architec-

ture Description Languages.IEEE

Transactions on Software Engineering,

26(1):70–93,January2000.

[RPM00]Gruia-Catalin Roman,Gian Pietro

Picco,and Amy L.Murphy.Software

Engineering for Mobility:A Roadmap.

In Proceedings of the conference on The

future of Software engineering,pages

241–258.ACM Press,2000.

[WBS02]Yaron Weinsberg and Israel Ben-

Shaul.A Programming Model and Sys-

tem Support for Disconnected-Aware

Applications on Resource-Constrained

Devices.In Proceedings of the24th in-

ternational conference on Software en-

gineering,pages374–384.ACM Press,

2002.

5下载本文