ftp.umcs.maine.edu/2015-02-07.ftp.umcs.maine.edu.tar/ftp.umcs.maine.edu/pub/WISR/wisr4/proceedings/detex/martin.detex

Text File | 1992-04-05 | 14KB | 382 lines

 [12pt] article








  Software Reuse Across Continents

  Roland Martin
       Hewlett-Packard GmbH
       Waldbronn Analytical Division
       Hewlett-Packard Str.
       D7517 Waldbronn, Germany
       email: [e-mail]

       Gary Jackoway
       Hewlett-Packard Co.
       Avondale Division
       P.O. Box 900
       Avondale PA. 19311
       email: [e-mail]

       Charu Ranganathan
       Hewlett-Packard Co.
       Scientific Instrument Division
       1601 California Avenue
       Palo Alto, CA. 94304
       email: [e-mail]





  The Analytical Products Group of Hewlett-Packard Co. is practicing
software reuse at its three sites, located in the US and Europe.
This paper describes the organization, goals, processes and results of a
reuse program. Data from two example projects illustrate the approach.
Lessons learned from these and other examples, as well as key factors
for economic success, are presented.

  0.3in

   Keywords:  software reuse, reuse without modification,
component library, economics




  Introduction

Hewlett-Packard (HP) Analytical Product Group (APG) supplies
instrumentation, software
and systems  for a wide range of analytical techniques including Gas
Chromatography, Liquid
Chromatography, Mass Spectrometry, Spectrophotometry and Automated Sample
Preparation.

Single-user workstations
and multi-user data systems with software for instrument control, data analysis,
automation
and information management are essential elements of APG's business.

APG has divisions at Palo Alto, CA., Avondale, PA. and Waldbronn, Germany. Each
site
has worldwide business responsibilty for a subset of above techniques, and does
research and
development (R+D) for instrumentation and software for their business segments.

Sharing of software workproducts across APG organizations began as early as 1982
 with
voluntary design for reuse and cooperative efforts among design teams.
APG's motivation for practicing reuse has two major elements:

1. Enhance value to customers through consistency and compatibility across
products

2. Improve productivity  by reducing redundant development efforts

In late 1988, the first author was assigned to further stimulate and lead these
reuse activities
with small project teams at all three sites funded by and reporting to the
APG R+D manager.

Activities include improving the infrastructure for multi-site reuse,
identification of workproducts for reuse,
re-engineering existing workproducts for reuse,
new designs for reuse,
consultation to projects designing with reuse, and maintenance of reusable
and reused workproducts.

Activities are organized in a multi-site program, i.e. a set of related
and coordinated, but
independently managed projects.

   Reuse Program Outline

The mission of this multi-site reuse program is to create a set of
software components for
reuse without modification in APG products. A subset of these
components is designed for multiple platforms (i.e. hardware and operating
system), others are
specifically designed for one platform. APG software products are developed for
HP-UX/X-Window, MS-DOS/MS-Windows, and firmware platforms.

Reusable components are expected to meet the following criteria:

* Functionality is generic within our domain: meet the need of at least
three products.

* Platform dependencies are isolated.

* Internal elements, in particular data structures, are hidden through
encapsulation.


Workproducts and services other than just code are essential to facilitate
reuse across platforms, products and organizations. The following items are
provided together with reusable components:

* Documentation: Specifications, Application Programmer's Interface,
Installation Guide.

* Test suites at component level for quality control independent of an
application.

* An example application.

* Consultation to consumers.

* Commitment for maintainance over several years of component lifetime.


The content of a reusable component is very similar to that of a software
product. The same is true for its development process, i.e. the component
life cycle.
We are using a scaled down software product life cycle to create and maintain
software components. A panel of R+D section managers from all three sites
chaired
by the  program manager reviews the projects and approves phase transitions.
This panel holds telephone or video conferences every two week for this
purpose.
A council consisting of all APG R+D Managers meets
three times per year and provides strategic
guidance and funding for the reuse program.

   Results

The program started in 1989 and produced an asset of about 50 000 lines of
code plus the additional workproducts for reuse without modification in multiple
software products. This asset covers more than 10 percent of the lines of
code released
with APG software products. With the currently active producer and
consumer projects we
expect to increase the asset to near 150 000 lines covering more than
30 percent of the code shipped with APG software products.
On average each component is reused in 4 products.

   Examples

   Signal Processing Algorithms
A set of signal processing algorithms to calculate quantitative information from
chromatograms was created in the early 1980's in Pascal. It was carefully
designed
for portability. Several software products made use of this library.
The creator was reassigned to other projects after completion.
The library was subsequently ported to new operating systems by product
development
teams and adapted to other analytical techniques. This resulted in six different
versions with subtle differences. Once a defect was found in one of the
versions,
it was a challenge to verify and fix all others.

Another effect was observed: the code size grew with each port or adaption.
The biggest growth in code size came during a port from Pascal to C using
a translator program. The original 3 kNCSS grew to 9 kNCSS, essentially without
any additional functionality.

The first project within the components program was set up to re-engineer this
library for strict black box reuse in all software products.

The result: size is down to 5 kNCSS, speed improved by a factor of two.

The project took about one year for two engineers (one of them the original
creator).
This effort included a detailed assessment of the requirements of all
potential consumer projects and the creation of an extensive test suite
with test cases
from all analytical techniques which need this type of algorithm.

After release of this component, four products on two platforms replaced their
individual versions with this one common stream of code. More products are under
development using the same source without modification.

Meanwhile, the component development team added new capabilies to the
library and
is responsible for all maintenance activities for this component.


   Command Processor

Like the first example, the command processor (CP) concept was created in
early 1980's. It started as a mechanism to coordinate a collection of
small programs and evolved into a language interpreter.
During this evolution process the CP concept was adopted by the majority
of APG R+D teams and became a key element for the
architecture of APG's software products and systems. The development teams
adapted code from other projects to fit the requirements of their product.

Different from the algorithm example, the perceived needs for CP
capabilities
are significantly different from project to project. Over time, projects
teamed up and shared the same code for a few product releases but afterwards
felt the pressure to implement those extensions that were most important for
their product's customers. The extent of adaption ranged from small enhancements
to almost complete re-writes. In the late 1980's we found ourselves
with around a dozen sets of CP code.
All versions do support the same core capabilities but with small,
sometimes annoying differences.

It was impractical to create one single set of code for Unix and DOS.
For each environment one version was identified for re-engineering.
The core capabilities were identified and described in a rigid syntax guide.
Product-specific elements were eliminated from the core. Mechanisms
were designed and implemented that allow addition of product-specific
capabilities
without modifying the core source code.

Both versions of the CP library are currently in the test phase and each have
delivered pre-release versions to two consumer projects. More project teams
plan to use the same code without modification.
An essential
shared deliverable is one common validation suite to check compliance with
the syntax guide mentioned above.

Size of both libraries is about 20 kNCSS.
We spent about two person years for each of the two libraries.
This included the creation of the syntax guide and a
validation suite shared between the two.

The CP teams will continue to enhance and maintain the two components
for a growing number of consumer products.

   Other Reuse Activities

Ten other software component libraries of similar nature and size covering
other funtions needed by three or more APG projects are in progress and
could have
served as examples as well. Some of them are new designs, others do
re-engineering of existing functions. Many of them are sponsored and funded
by APG divisions and complement those funded by Group R+D.


   Some Lessons We Learned

From the above examples our organization learned a few lessons, including:


1. White-box reuse (i.e. reuse with modification) has marginal economic
benefits.

2. Black-box reuse (i.e. reuse without modification) help us to achieve
our goals, i.e. (a) to achieve consistency across products and (b) to
reduce redundant development and maintenance efforts.
However, it requires sustained attention
over the full life cycle of components and products using these components.

3. Unsustained black-box reuse degenerates within 2 years.

4. Voluntary and cooperative reuse efforts do not last in a tough business
environment. Management support and some organizational structure is
necessary to sustain reuse.

   Key Success Factors

   Products
APG products cover a diverse range of analytical techniques, but share many
functions, not only in software, but also in hardware.
Identifiction of these commonalities is rather obvious.
End users strongly demand consistency across products for
these common functions.

   Culture
During years of steady growth, APG built a strong base of experienced people.
Best practices are exchanged among teams worldwide through formal and
informal channels.
This exchange of best practices results for example in:

* One product development life cycle is practiced by all teams

* The same operating systems are used by all new products

* All new code is developed in C.

* Reasonably consistent software development tools exist at all sites.


   Management Support
Upper management supports and expects reuse and allows for
the necessary investment within reasonable constraints.
Significant effort has been spent, and will be necessary in the future to
provide the environment for reuse across projects, organizations and
continents, including:

* Communication tools (telephone and video conferencing, fax)

* Common development tools

* Global controlled access to reused workproducts

* Defect tracking for components

* Keeping track of who is using what (configuration management)

* Common business goals for the participants of a reuse program

* Consensus on technical concepts (architecture, user interface style)

* Compatible ways of doing things (processes)

* Validation of compatibility


   People
Definitely the most important factor for the success of a reuse program
are good working relationships among qualified,
dedicated and experienced people in all roles and levels.

No technology can compensate for the lack of this key factor.

Unfortunately
we cannot give a simple recipe for achieving this.
Regular face-to-face meetings
between the key people are definitely important to build and maintain such
relationships. We have experienced some difficulties in our program whenever
more than two months passed without personal contacts between two
project teams.



   Conclusions

Software reuse is more than just creating and/or obtaining workproducts for/from
a public domain. In order to get economic benefits, it requires upfront investment
and sustained effort. More than any particular technology, the consensus
between producers and consumers of reusable workproducts is essential for success.

Hewlett-Packard Analytical Products Group (APG) has made a two year
investment in software reuse and is beginning to see economic benefit from this
program. We are prepared to sustain that effort.

   Biography
Roland Martin, born 1946 is managing the APG reuse program. He holds a Ing.(grad)
degree in electrical engineering from Fachhochschule fuer Technik, Esslingen, Germany.
He joined HP's Boeblingen division in 1969 and relocated to the newly
created Waldbronn Analytical Division in 1975. His past assignments included
production engineer, starting up a PC board test facility, design of hardware,
firmware and software for liquid chromatography, project- and section manager.
He is single, lives 5 minutes from work in a small village
at the edge of the black forest. Between his
frequent trips to the US, he tries to keep his skills as a glider pilot
up to date.

Gary Jackoway, is an R+D software engineer.  He holds a Master's degree in
Computer Science from Duke University, and a Bachelor's degree in Mathematical
Sciences from Stanford University.  Before joining Avondale Division, Gary
worked on automatic routing and placing algorithms for printed circuit
board design.  Gary is an avid volleyball player, and a contract bridge
Life Master.

Charu Ranganathan has been a software engineer at Scientific Instuments Division
of Hewlett-Packard (Palo Alto) for the past 2 1/2 years.
She is currently working on the Command Processor and will continue to be
part of the reuse program at HP. She holds a BS in Computer Science from
Bangalore University, Bangalore, India and an MS in Computer Science from
University of South Western Louisiana, Lafayette, LA. She is an Indian
classical musician (vocalist) and is interested in painting/sketching.