Industrial experiences on Domain-Specific Modeling

Industrial Experiences on Domain-Specific Modeling

Contents

Domain-Specific Modeling Languages: Introduction

Review of industry cases

– Touch screen

– Home automation control

– Sports computer

– Military radio

Summary and discussion

1

4,5 56 6

40

0

5

10

15

20

25

30

35

40

Number of new product

features implemented in a given time (productivity proportional to Assembler)

A rise in productivity is overdue

"The entire history of software engineering is that of the rise in levels of abstraction"

New general-purpose programming languages have not increased productivity

Abstraction of development can be raised above current level...

... without losing control or accepting substandard results

*Software Productivity Research & Capers Jones, 2002

What data is available behind the statements* like:

"5-fold productivity increase when compared to standard development methods"

"The quality of the generated code is clearly better, simply because the modeling language rules out errors"

"The DSM solution makes development significantly faster and easier than the old manual coding practices"

* source: www.metacase.com/cases/

Four cases in more detail

Touch screen device (Panasonic)

Home automation (Ouman)

Sports computer (Polar Electro)

Military radio (Elektrobit)

Case 1: Panasonic’s touch screen devices (Safa, 2007)

Home automation solution installed by construction firms

Features for controlled

– Lights

– Heating

– Air-conditioning

– Electricity

– Alarms

• Burglar

• Gas

• Smoke

– Reporting (energy saving)

Evaluation method

Compare DSM to current manual coding practice

Two evaluation methods

1. Implemented an existing product with DSM

2. Implemented the existing features to a different target platform

• The same models, a new generator

Tool-chains covering: – Generate

– Compile

– Upload

– Boot

– Run

Generators for: – PC simulation

– Touch screen

– Microcontroller

Measuring development time

Build the same application with both approaches

– Measure the time used

Results:

– DSM is 425% faster

Implemented the same product to a new platform

– 3 days for generator development

– 0 days for modeling

Return on investment: Panasonic

DSM solution developed in 15 days

Product development with:

– the current approach: 17 days

– DSM: 4 days

0 5 10 15 20 25 30 35

DSM

Coding

Days

Creating DSM solution

Product 1

Product 2

Product 3

Product 4

Product 5

Case 2: Home automation (Puolitaival, 2011)

Home automation, remote control via mobile phone Ouman manufactures home automation systems

Products are focused mainly on temperature control with many different heating systems

SMS interface for remote control

A language for specifying remote control applications

One model for Ouman EH-60 product remote controller

0 5 10 15 20

DSM

"past"

Days

Creating DSM solution

Product 1

Product 2

Product 3

Product 4

Product 5

Return on investment

DSM solution developed: 2 weeks

Product development: 1-2 days – Comparison to earlier development effort not possible since

outsourced • Cost was 6 figure number

DSM allows a non-programmer to develop applications

Case 3: Polar’s Sport computers

Heart rate measuring, analysis and visualization Calorie calculation, like current, cumulative, expenditure rate, active time Speed: current, average, maximum Distance, based on interval, trip, recovery Altimeter, vertical speed, altitude alarms, slope counter, graphical trend Cycling information like pedaling rate and cycling power Barometer, pressure drop alarm, graphical trend Exercise diaries Sensor connectivity (heart rate, speed, cadence, power, GPS) Compass, Temperature, Odometer, Logbooks, etc.

About the product development (Kärnä et al. 2009)

Polar focused on UI application development

– Single largest piece of software

• Takes 40-50% of the development time

– Typically always vary among products

Software development is constrained by limited resources:

– Memory, processor speed and battery life

Polar created the needed languages and generators internally

Sample of UI application design

Evaluation methods

Compare the use of DSM and the current practice

Two research methods

– Laboratory study

• 6 current developers, 6 implementations

• Implement a small, typical feature

– Pilot project

• Implement large portion of a whole product

• 1 person

Results of the studies

Laboratory study

– Measuring time: at least 750% faster

– Asking opinions: results (scale 1-5, 5 best):

Pilot

– Measuring time: >900% faster

Return on investment: Polar

DSM solution developed in 60 hours

Product development with:

– the current approach: 23 days

– DSM: 2,3 days

Case 4: Military radio (Puolitaival et al., 2011)

EB Tough VoIP Features

Tough VoIP is a wired phone that is using UDP/IP network for connection

Manufacturer: Elektrobit

Main features:

– Easy configuration

– Point-to-Point call

– All call

– War-proof device

Testing problem

ETC...

EB Test Tool Platform + OpenTTCN tester

Two language solution

Model Model

MBT

TTCN-3 TTCN-3

Modeling one test case

Modeling a test logic

Model-Based Testing

generates multiple test

cases

Generating one test case

Executing the test case

Executing test cases

Model example 1: test case

Model example 2: test logic

0 5 10 15 20

DSM

Coding

Days

Creating DSM solution

Test suite 1

Test suite 2

Test suite 3

Test suite 4

Test suite 5

Return on investment: Elektrobit

About 10 times faster with modeling

Set-up time estimation: – 2 weeks for the first version, +1 week to make it better

Other benefits: – Test coverage dramatically increase

– Easy test configuration

Economics of DSM

Repetition:

– # of product variants

– # of similar features

– # of developers

– ”outsourcing” to domain experts

Investment:

– Effort needed to implement DSLs

0

10

20

30

40

50

60

70

80

90

100

0 1 2 3 4 5 6

Repetition

CostCurrent DSM

DSM Solution Development Time

Man days

Concepts Symbols

Generators Rules 1 2 3 4

Steps for implementing DSM

Summary

Domain-Specific Modeling languages provide:

– Better productivity

– Quality improvements

– Easier use and introduction of new developers

MetaEdit+ makes moving to DSLs feasible

– Expert can focus on language design, not on creating tooling

– Models update and work when languages changes

Creation of languages does not take much time!

More details on the cases

Kelly, S., Tolvanen, J.-P., Domain-Specific Modeling: Enabling Full Code Generation, Wiley, 2008. http://dsmbook.com

Kärnä, J., et al. Evaluating the Use of Domain-Specific Modeling in Practice, Proceedings of the 9th Workshop on Domain-Specific Modeling, HSE Print, B-108, 2009.

Puolitaival, O.-P., Home Automation DSL case, Code Generation Conference, 2011

Puolitaival, O.-P., Kanstren, T., Rytky, V.-P., Saarela, A., Utilizing Domain-Specific Modeling for Software Testing, Proceedings of VALID, October 2011

Safa, L., The Making Of User-Interface Designer, A Proprietary DSM Tool, Proceedings of the 7th Workshop on Domain-Specific Modeling, Technical Reports, TR-38, University of Jyväskylä, 2007

For more cases, customer stories, testimonials visit www.metacase.com

Thank You!