ftp.umcs.maine.edu/2015-02-07.ftp.umcs.maine.edu.tar/ftp.umcs.maine.edu/pub/WISR/wisr6/proceedings/ascii/burd.ascii

Text File | 1993-10-19 | 16KB | 343 lines

            Risk  Management  the  Key  to  Successful  Reuse



                             Elizabeth L. Burd, John A. McDermid



                                 Department of Computer Science,

                                           University of York,

                                                Heslington,

                                                    York,

                                                 YO1 5DD,

                                                      UK



                                         Tel:  [phone]

                                        Fax: [phone]

                                   Email:  [e-mail]



                                                  Abstract


    In this position paper we stress the importance of risk management to the process of software
reuse. We consider the problems that are associated with reuse and have accounted for its failure
to become an important and respected part of the software development process. We discuss,in
brief, the work that we have carried out. In particular we concentrate our efforts for this paper
on the procedure of reusing risk assessments to direct a new software development.  We then
look at some of the implications that increasing a developer's knowledge of risks has on the ac-
ceptance of reuse to project managers. Finally we indicate the new directions our work will take.



Keywords: Reuse, Risk, Software life cycle, Process models.


Workshop Goals: Listening , learning, discussing and meeting people


Working Groups:  Reuse process models, Reuse management, organisation and economics,
Reuse maturity models, Education.



                                                   Burd- 1


1      Background



From a study of the researchconducted into the problems associated with software reuse and also
from the industrial experiences of companies trying to employ reuse we,  in harmony with others,
have identified a basis on which the success of reuse depends.  We believe that for reuse to flourish,
it must be viewed as an integral part of the software life cycle.  Changes must be made to the way
in which industry views reuse.  Industry needs to move away from simply the reuse of code and
certainly away from the attitude that it is acceptable not to consider reuse at all. We examined why
these attitudes have existed and our studies have led us to the conclusion that there are a number
of inhibiting factors which limit the appeal of reuse to project managers and software developers.
These inhibiting factors include, at a technical level:



    ffldevelopment - Knowing what kind of software is reusable and as equally difficult is knowing
       how to develop a software component which is potentially reusable.


    fflstorage - Once we have developed an item of reusable software how, and where, should it be
       retained for future retrieval and reuse.


    fflretrieval - If we are to reuse software then we must be able to easily find what we require
       matching what is available with our needs.


    fflverification - How can we be sure that the component which we are proposing to reuse actually
       performs the functions that it claims it does in the environment in which we use it.


    fflevaluation - How can we judge that the functions we require from our reusable unit and the
       functions that it provides are the same.


    fflmodification - If our evaluations have shown that differences exist in the reuse unit's and our
       requirements then how do we perform the necessary modifications and what effect will this
       have on the reusable unit including the results of previous verifications.


    fflintegration - What will the effect be on the reusable unit of attempting to integrate it into
       our development.



We thus came to the realisation that the underlying technical reason why reuse was not as successful
is that developers would have liked as the inhibiting factors which we have described aboveincrease
development risks. Since project managers tend towards the path with the smallest risksthey thus
often, and rightly given the circumstances, tend to develop 'from scratch'.  We therefore consider
that if reuse is to be successful we need to indicate its tangible benefits to project managers by
reducing the risks associated with development and especially when reuseis b eingconsidered.  In
addition it should strive to make reuse a continuous and integral part of the software development
process as this has an effect both on risks and acceptability of risks.


We acknowledge that there are other factors affecting the take up of reuse e.g. individual motiva-
tions but such issues are outside the realm of this paper.



2      Comparison



Barnes and Bollinger [1, 2 ] have indicated similar beliefs about the nature of the problems that are
associated with software developments incorp orating reuse.  They advocate the necessity for a broad


                                                          Burd- 2


spectrum of reuse to be considered and stress the importance of considering human problem solving
as an aspect of software development that is p otentially reusable.  Their work has concentrated on
the need for incentives for performing reuse and have indicated ways in which this can be taken
account  of  in  the  software  development  process.  In  addition  they  have  indicated  that  life  cycle
integration is likely to one of the keyfactors which makes reuse truly effective. This work formed
the foundation from which we embarked in our research One major contribution, and the issue on
which we focus here, is the notion of risk and the reuse of risk assessment - a sort of human problem
solving.


In addition we have also built our work around Boehm's spiral model [3 , 4].  Wewill consider the
additions we have made to the model in more detailb elow, but in brief we have to base our work
on the spiral model because of its risk based approach to software development.  Our extensions
were included to enhance the work of Boehm's to encourage a life cycle approach to reuse and a
broad definition of reuse.



3      Position



We  have  indicated  ab ove  that  development  risks  are  actually  be  increased  with  reuse.  This  is
probably the most compelling reason to develop systems 'from scratch'.  However, the perceptions of
risk may outweigh the actual risks,and social factors will tend to lead project managers towards the
(perhaps illusory) feeling that developing 'from scratch' puts then in control' and hence minimises
risk.  We  consider  these  apparent  increased  risks  an  unnecessary  burden  for  reuse  to  carry  and
believe that given a broader definition of what is reusable from the software development process
then reuse can actually serve to reduce development risks.  Perhaps more importantly by enabling
risk-based decision making about reuse to be explicit, we can help ensure that decisions about reuse
are balanced and well supported.



3.1     Hypothesis



The hypothesis of the PROM project was therefore: reusing all products of software development
will reduce the risks that are associated with reuse and that since develop ers will find reuse more
acceptable, if they are aware of the scale of risks that are involved then reuse will become a necessary
and informative part of software development.



3.2     Research



As  we  indicated  above  that  our  work  has  been  based  on  building  extensions  to  Boehm's  spiral
model.  This allows us to incorporate our aims of providing a reusemetho d which is integral to
software development while supp ortingreuse continually and also encouraging a risk based approach
to development.  These extensions have taken the form of a 'reuse spiral'.  This has a symbiotic
relationship the spiral model; the spiral model encourages a risk based approach to development
and the reuse spiral provides the 'know-how' for reuse.  Thus the information which is collected by
one of the spirals is used by the other and vice versa.


Each of the spirals of the two models are split into four quadrants. The activities that are concerned
with these quadrants are as follows:



                                                          Burd- 3


                           ________________________________________________________________________
                           __Spiral_model_________Reuse_spiral_____________Problems_tackled________

                           _ Q1                 _                         _                        _
                           _ Define system  _     Use information   _      Retrieval            _
                           _  objectives and  _   gathered as         _                            _
                           __constraints__________search_criteria___________________________________

                           _ Q2                 _                         _                        _
                           _ Evaluate         _   Reuse risk          _    Evaluation          _
                           _ alternatives     _   assessment          _    Verification         _
                           _ identify and     _   information to     _                             _
                           _ resolve risks.    _  aid risk              _                          _
                           _______________________management________________________________________

                           _ Q3                 _                         _                        _
                           _ Develop and    _     Reuse products    _      Development       _
                           _ verify next -    _   of previous          _   Modification        _
                           _ level product    _   developments      _      Integration          _
                           _______________________where_possible____________________________________

                           _ Q4                 _                         _                        _
                           _ Plan next        _   Focus record and _       Storage              _
                           _ phases            _  commit decisions  _                              _
                           _______________________to_the_library____________________________________



3.3     Case Study



We consider that one area which is often disregarded in the field of reuse research are those activities
which are concerned with the second quadrant of the reuse spiral; mainly that of the evaluation of
the risks associated with a development. We have indicated how important this is for successful
software development and therefore we will illustrate this process, of reusing risk evaluations, with
the aid of an example derived from a case study of the National Health Service.


In the second quadrant we perform the following activities:



    1. We use information from the similar projects which were located in the first quadrant and
       investigate what options are available for developing a system,  and secondly what risk as-
       sessments activities were associated which each option. From this information weare able to
       indicate some of the risk criteria which are associated with our new development. In terms of
       the health service, typical risk criteria for developing a system which informs doctors which
       drugs  are  functionally  equivalent  may  be:  the  doctors  reaction  to  the  system;  the  cost  of
       development; the accuracy of the information it generates; or the feasibility of building the
       system.   Typical  options  may  be  to:  develop  a  database  application, an  expert  system  or
       perhaps nothing at all i.e.  retain amanual system.

    2. We can now use sensitivity analysis to allow us to evaluate which is the best option to take
       for our proposed development. In order to do this then we evaluate risk criteria individually
       to obtain the utility for each risk criteria. The risk criteria for each option may be the same
       but their utility will be different. (Here we use utility in a technical sense, a concept unifying
       benefit and risk.)  In the caseof our drug substitution program the utility for cost will depend
       on the expected cost of the individual options and the utility for doctors acceptance of the
       system  may  for  instance  be  greater  for  the  database  option  than  the  expert  system  since
       doctors  may  feel  an  expert  system  is  de-skilling  their  job.  Weightings  are  then  placed  on



                                                          Burd- 4


       the individual criteria (again information gained from previous projects) in order that the
       interaction effects between the individual risk criteria can be evaluated.

    3. By folding back the decision trees the utility for each option can then be considered and the
       option with the highest utility, or greatest expected payoff, can b e selected.



3.4     Conclusions



Our studies have lead us to conclude that an all round approach to reuse improves the chances of
a product or process definition of a previous development being reusable. Without taking a risk
based approach to development the typ eof reuse which we have described would in most cases be
infeasible. Thus we believe that reusing risk assessment,weightings and prototypes etc. allow more
informed judgements to be made as to the feasibility of various options of software development
and therefore reuse to be, in general, more successful.



References



[1]  T. B. B.H. Barnes, "Making Reuse Cost-Effective," IEEE Software, pp. 13-24, January 1991.


[2]  S. P. T.B. Bollinger, "Economics of Reuse: issues and alternatives," Information and Technol-
     ogy, December 1990.


[3]  B. Boehm, "A Spiral Model for Software Development and Enhancements," IEEE Computer,
     pp. 61-72, May 1988.


[4]  B. Boehm, "Software Risk Management: Principles and Practices," IEEE Software, pp. 32-41,
     January 1991.



4      Biography



Elizabeth Burd's background is in education; she is a qualified teacher. In 1989 she entered the
research  community  and  after  completing  a  masters  in  Information  processing  she  subsequently
proceeded to research for a Ph.D. in the Information Systems group at the University of York. She
is currently working on the PROM project which isconcerned with a life cycle approach to reuse.
Her interests in education are very much reflected in the options she is considering to aid some of
the problems associated with reuse.


Elizabeth is also a member of the British Computer Society Reuse Special InterestGroup commit-
tee.


John McDermid has been Professor of Software Engineering at the University of York since 1987.
Prior to coming to York he worked for SD-Scicon and the Royal Signals and Radar Establishment
(now the DRA Malvern). At the University he runs the High Integrity Systems Engineering group,
concentrating on safety-critical and dependable systems. He is also the technical director of the
Dependable Computing Systems Centre set up by British Aerospace at the Universities of York
and Newcastle and a director of York Software Engineering Ltd.



                                                          Burd- 5


He is active in the professional community, and has assisted the IEE and BCS in major studies of
safety-critical systems. He has published widely, being the author or editor of six books, and the
author or co-author of about one hundred and twenty papers and articles.



                                                          Burd- 6