Construction d'applications dans la jungle de l'IOT

DIVERSE : DIVERsity-centric Software Engineering

Software intensive systems

Context: Buidling services on top of the IoT Jungle

• Heterogeneous and Distributed Services • Heterogeneous: The infrastructure on which the service

runs is composed of a set of different nodes and networks. •  The “IOT Jungle" which ranges from microcontroller based

sensors and devices to cloud.

• Distributed: The implementation of the services is composed of a set of independent processes communicating asynchronously. •  Truly distributed services implementation is required in order to

provide useful and reliable services which take advantage of the infrastructure.

Examples n  Health domain and ambient assisted living n  Energy domain and smart grids n  Environmental monitoring and oil and gas n  Safety in hazardous environments n  Intelligent Transport Systems (ITS) n  ...

Heterogeneous infrastructure

New Services on top of the IoT Jungle

Legacy Partner

Proprietary product

External service

System under development

Things

8

Arduino RaspberryPi

IntelEdison STNucleo ESP8266

Cost $3-$30 ~$35 $40 $8 ~$3

Wifi Shield($60)orESP8266

USBDongle Built-in Shield($60)orESP8266

Built-in

Programming C/C++ Python/Java/C++,Javascript,…

C,C++,Javascript,Python,Java,..

C,C++ C++/Lua

CodedistribuAon

USB/SPI/Serial In-Situ In-Situ USB/SPI/Serial Serial/OTA

Storage Built-In SD-Card Built-In Built-In Built-In

I/O ArduinoUno:13GPIO/6ADC

17GPIO 20GPIO(PWM,SPI,one-wire) •6analoginputs. 1UART(Rx/Tx).

GPIO(50) 12-bitADCwith16 Timers(8) I2C(3),SPI(3) USART(3)

10GPIO/1ADC

PowerconsumpAon

12V 50mA Deepsleep?

1A5V peak:984mWusingWifi5GHzband:680mWusingserialport:346mW Idle88mW poweredoff45mW

128µA/MHz.InStopmode, thepower consumpeoncanbeaslowas9µA. 3,3V

3,3V Transmit802.11b,CCK11Mbps,POUT=+18.5dBm197mA Standby0.9mA Deepsleep 10uA

9

Model-driven software engineering to handle

diversity

Our Main Objective

• Enable efficient exploitation of the broad diversity of the future computing continuum for rapid service innovation of advanced services, by providing service developers with new agile tool supported software engineering methods enabling short innovation cycles

10

Heads project outputs Two new modelling languages to tame the

IoT jungle

11

What is ThingML ? n A DSL to model distributed reactive systems

n  IoT systems, embedded systems, sensor networks, ... n Components, State machines and action language

n  « Main stream » MDE n Contributions of ThingML

n  « Complete » action language n  Slots, Mixins and Aspects instead of Inheritance and Composites n  Enforced encapsulation and actors semantics

n  Target Platforms and Applications n  MDE for resource constrained systems (microcontrollers, IoT) n  Development of applications distributed across heterogeneous

hardware n  Other types of reactive systems?

Why ThingML ? n  Typical MDE benefits

n  Reduce development, maintenance and evolution costs n  Perform verifications and analysis on the models n  Model application at a platform independent level

n  No existing approach can deal with microcontrollers n  ThingML can run on hardware less than 1ko of RAM

n  No existing approach is really platform independent n  Since actions are written in the target language

ThingML Goals

• Provide tools and methods • For each actor to concentrate on his task

• For decoupling the tasks of different actors

• Using state of the art software engineering practices • Modularity, reusability, runtime deployment, continuous

integration, validation, etc…

• Cost efficient and practically usable • No large overhead, integrated with legacy systems, etc…

Blink example state machine thingBlinkincludesLedMsgs,TimerMsgs{requiredportHW{sendsled_toggle,timer_startreceivestimer_timeout}statechartBlinkImplinitBlinking{stateBlinking{onentryHW!timer_start(1000)transition->BlinkingeventHW?timer_timeoutactionHW!led_toggle()}}}

Blink

Stopped ?timer_timeout !led_toggle()

on entry !timer_start(1000)

HW

<= timer_timeout() => led_toggle() => timer_start(delay:int)

Blink example and instance groups

Timer:Timer

do: DigitalOutput pin: IOPin

Arduino Board

app: BlinkApplication

led: Led

led group

Led HW

HW timer

io group

configurationBlinkArduino{groupled:LedArduinosetled.io.digital_output.pin=DigitalPin:PIN_13//Thetimerinstancetimer:TimerArduino//Theblinkapplicationinstanceapp:Blinkconnectorapp.HW=>led.led.Ledconnectorapp.HW=>timer.timer}

The ThingML tools n  Based on Eclipse / EMF Metamodel n  Textual Syntax with EMFText

n  For good usability and productivity n  To keep the development cost of the editor(s) reasonable

n  Graphical exports (graphML, graphviz, …) n  Static well formedness and type checker n  Equivalent compilers for a set of platforms

n C/C++ for different microcontrollers, linux, embedded linux n  Java for computers, smartphones, … n  Javascript (NodeJS) n Maybe others if needed

n  Generators for communication channels n  Easy to distribute ThingML IDE

n Standalone and lightweight IDE n Eclipse plugins

ThingML Editor

Kevoree: a DSML to drive software configuration and deployment

• Heterogeneous execution platforms • how to handle deployment?

• how to handle reconfiguration?

• how to verify system level properties?

• Effective support for runtime models •  limited resources

•  time constraints

• Distributed model dissemination

Models@runtime for distributed and heterogeneous systems

Lightweight M@R framwork for building Distributed Adaptive System

M@R Runtimes for Distributed and Heterogeneous adaptive systems

Extensible Virtual System Infrastructure

Resilient Software

infrastructure based on diversity

Kevoree in nutshell 1/3

Kevoree in nutshell 2/3

•  Kevoree Abstraction model for architectural diversity

•  Model@Runtime eases the management of distributed systems

•  Homogeneous adaptation for heterogeneous CPS devices

•  Large scale adaptation dissemination (Gossip, Paxos, etc.)

•  Ready for cloud and elasticity algorithms development

•  Open source

•  lost, www.kevoree.org

Kevoree in nutshell 3/3

Current Works

24

correctness envelop

Reasonable correctness envelop

25

P

P'

• Automatic synthesis of software diversity • sosies are different

programs that still belong to the correctness envelop

Automatic diversity synthesis

• Automatic synthesis of software diversity • sosies are different programs that still belong to the

correctness envelop

26

P

correctness envelop

PP'

correctness envelop

Software intensive systems

27

Software intensive systems

Continuous design for

Heterogeneous, variable, adaptive within reasonable

correctness boundaries