2012. Eko Didik Widianto. Teknik Sistem Komputer - Universitas Diponegoro

Jaringan dalam Sistem Embedded Terdistribusi @2012,Eko Didik Widianto Sistem Embedded Terdistribusi

Jaringan dalam Sistem Embedded Terdistribusi Kuliah#5 TSK-612 Sistem Embedded Terdistribusi - TA 2011/2012

Eko Didik Widianto Teknik Sistem Komputer - Universitas Diponegoro

Event- & Time-Triggered System Evolusi Komputing Terdistribusi Why Distributed Might Be Better Lisensi

Review Kuliah Sebelumnya

Jaringan dalam Sistem Embedded Terdistribusi @2012,Eko Didik Widianto

I Pemodelan sistem embedded terdistribusi menggunakan UML I I

Keterkaitan antara UML dengan metodologi desain yang diambil Tipe diagram UML: I

I

Struktur: component diagram, deployment diagram, class diagram Perilaku: use-case diagram, activity diagram, state diagram

I Prinsip kerja sistem elevator: prinsip dasar, profil dan unjuk kerja,

arsitektur kontrol, sistem keselamatan, antarmuka pengguna dan pertimbangan desain I

Implementasi UML untuk desain sistem elevator

I Mendesain sistem embedded terdistribusi secara komprehensif, mulai

dari kebutuhan sampai pengujian sistem I

Kasus: mesin penjual soda otomatis (soda vending machine, automated retail kiosk)

Sistem Embedded Terdistribusi Event- & Time-Triggered System Evolusi Komputing Terdistribusi Why Distributed Might Be Better Lisensi

Tentang Kuliah #5


I Pendekatan komunikasi/jaringan dalam sistem embedded terdistribusi I I

Event-triggered: transfer message asinkron Time-triggered: transfer message sinkron

I Teknik sistem terdistribusi: I I I I

Sumber daya listrik terdistribusi (power switch) Sinyal kontrol termultipleks Komputasi terdistribusi “Smart nodes” – kasus ekstrim dari terdistribusi

I Pengembangan sistem dari elektromekanikan menjadi distributed I Link I

Website: http://didik.blog.undip.ac.id/2012/03/06/

kuliah-tsk-612-sistem-embedded-terdistribusi-2011/ I

Email: [email protected]

I Acknowledgement: I

Beberapa gambar yang ada di slide ini diambil dari http://www.ece.cmu.edu/~ece649/[ECE649]


Bahasan Sistem Embedded Terdistribusi Event- & Time-Triggered System Pengiriman Message di Sistem Terdistribusi Pendekatan Event-Triggered Pendekatan Time-Triggered Evolusi Komputing Terdistribusi Menambah Mikrokomputer Remote Power Switched System Use Multiple Computers Fully Distributed Control Network Why Distributed Might Be Better Kelebihan Sistem Tersentral Modularitas Dalam Sistem Terdistribusi Diagnosabilitas dalam Sistem Terdistribusi Kelebihan Sistem Terdistribusi Lainnya Lisensi

Jaringan dalam Sistem Embedded Terdistribusi @2012,Eko Didik Widianto Sistem Embedded Terdistribusi Event- & Time-Triggered System Evolusi Komputing Terdistribusi Why Distributed Might Be Better Lisensi

Multi-system Autonomous Terhubung Network


Sistem Caterpillar 797


I Caterpillar Inc. produsesn peralatan/mesin berat (http://www.caterpillar.com/)


Sistem Caterpillar Braking/Cooling Control


X-By-Wire dalam Otomotif


I

X-By-Wire (Drive-by_Wire) merupakan teknologi kunci dalam kontrol komputer dalam otomotif menggantikan kontrol mekanik I

Mempunyai requirement yang ketat (Safety-Related Fault-Tolerant)



Jaringan dalam Sistem Embedded Terdistribusi @2012,Eko Didik Widianto Sistem Embedded Terdistribusi Event- & Time-Triggered System Pengiriman Message di Sistem Terdistribusi Pendekatan Event-Triggered Pendekatan Time-Triggered

Evolusi Komputing Terdistribusi Why Distributed Might Be Better Lisensi

Contoh Sistem: Tombol Pemanggil Elevator


I

Up Button I

I

Up Button Light I I

I

Light on Light off

Down Button I

I

Button press

Button press

Down Button Light I I

Light on Light off

Event- & Time-Triggered System Pengiriman Message di Sistem Terdistribusi Pendekatan Event-Triggered Pendekatan Time-Triggered


Elevator Call Button: Event-Based Messages I

Button Press I I

I

Turn on button light Send a “button pressed” message every time button pressed Keep button light turned on for 500 msec after button pressed I

I

I

Turn button light off when told to by hallway computer

Hallway computer (hall controller, one for all floors) I

I

I

I

Then turn button light off unless an acknowledgement message is received

When button is pressed, send acknowledgement to keep light on When button is pressed, notify dispatching computer to send elevator car When car doors open, send “turn button light off” message

For dependable systems, need handshake messages at every step I I

Might be at application layer Might be at network protocol layer



Elevator Call Button: Time-Triggered Messages I

Send “button” and “button light” messages every 150 msec I

I

Button Press I

I

I

I

Each message has current state of “on” or “off” Set “button state” variable to “pressed” (gets copied to hallway computer) Turn on button light and force on for 500 msec; then follow message value Whenever you can, check the button light state and set light accordingly (perhaps every 150 msec)

Hallway computer (one for all floors) I

While button is pressed, notify dispatching computer to send elevator car

I

When car doors open, set button light state to “off” & button state to “off”

I

I

Set button light state to “on”

Message acknowledgements often not required I I

Missed button light state messages are self-correcting User must correct a completely missed “button press” message by pressing again



Event-based vs Time-Triggered Message


I

Event-based messages (“events”) I

Send messages in response to an event I I

I

Similar to event-triggered systems Think “asynchronous” state machine transitions

Can have high message frequency with an “event shower”

Sistem Embedded Terdistribusi Event- & Time-Triggered System Pengiriman Message di Sistem Terdistribusi Pendekatan Event-Triggered Pendekatan Time-Triggered

I

Time-Triggered messages (“state variables”) I

Message sending is invisible to application programmer I I

I I

I

Send messages periodically with the latest information Think “synchronous” state machine transitions

Reduces problems with missed messages Matches communication & computing load to system control loop speeds rather than external environment Have to infer that an event happened based on observing a value change I

Might be easy to miss a sequence of quick events





Sistem Event-Triggered I

Sensors feed computers


Evolusi Komputing Terdistribusi

I

I

I

Computers are demand driven I I

I

They process values from queues when available Network messages are queued for transmission

Network is demand driven I

I

But their values arrive asynchronously – whenever values are available So, sensor values mustbe queued

Transmission queues release messages according to some network protocol

Actuators are demand driven I

Actuator outputs occur in response to event messages

Why Distributed Might Be Better Lisensi

Problem - Coordinating Event Automatic Transmission Car


I

http://en.wikipedia.org/wiki/Automatic_transmission

I

Car example: transmission unlocks to shift out of park if: I I I

(Engine is running) AND (Parking brake is OFF) AND (Brake is ON)



Salah Satu Urutan Event



I

But, this is too simplistic! I

There are many arrival orders

Problem di State Machine Asinkron


I

Asynchronous state machines have trouble with complex condition I

I

Need to use states to keep track of which events have been seen if order is flexible (This is also a problem with UML sequence diagrams)






Pendekatan Time-Triggered

Jaringan dalam Sistem Embedded Terdistribusi @2012,Eko Didik Widianto Sistem Embedded Terdistribusi Event- & Time-Triggered System Pengiriman Message di Sistem Terdistribusi

I Computers poll sensors

Sensors queried periodically I Sampled whether or not there has been a change I Computers are time triggered I Sensors are sampled when it is time perform a computation I State variables are placed in buffers for later transmission on network I Network is periodic I Copies of buffers are sent to all nodes periodically I Receiving nodes store most recent values, even if previous value unused I The “blackboard architecture? I The receive buffers are that node’s locally stored copy of the current blackboard contents I Actuators are periodically driven I Actuators outputs asserted periodically regardless of change/no change I

Pendekatan Event-Triggered Pendekatan Time-Triggered


Time-triggering Menyederhanakan Koordinasi Event


I

Example: transmission unlocks to shift out of park if: I I I

(Engine is running) AND (Parking brake is OFF) AND (Brake is ON)



Updating Isi Buffer I

Design based on periodically updated state variables I

I

I

Every compute/communication block has a state variable buffer Buffer contents are updated whenever source sends an update Buffer values are read whenever consuming functional block needs them



Perbedaan Desain: Event vs Time-Triggered


I Event triggered: I

State machine only changes states when event occurs I

I

Each arc can have ONLY ONE incoming event/message I I I

I

Actions within state are executed exactly once, when state is entered

Instantaneous state changes Events arrive via network message or are serialized in some other way Requirements statements can depend on exactly ONE message

In many cases these are simpler to understand & design I I

But, they are easily confused by dropped, duplicated, or out-of-order messages Coordinating multiple messages requires messy temporary variables

I Time triggered: I

State machine changes periodically based on variables and most recent inputs I I I

I

State machine has to infer events by noting state changes Can change state based on multiple messages and/or local variable values Actions within each state are executed continually

Tend to be more robust to dropped or out-of-order messages



Step By Step Evolution

Jaringan dalam Sistem Embedded Terdistribusi @2012,Eko Didik Widianto Sistem Embedded Terdistribusi Event- & Time-Triggered System

I

how systems evolve from electromechanical to distributed ? 1. Add a microcomputer 2. Remote Power Switched System 3. Use Multiple Computers I

Mostly-Distributed Control Network

Evolusi Komputing Terdistribusi Menambah Mikrokomputer Remote Power Switched System Use Multiple Computers Fully Distributed Control Network


4. Fully-Distributed Control Network

Bahasan Sistem Embedded Terdistribusi Event- & Time-Triggered System Pengiriman Message di Sistem Terdistribusi Pendekatan Event-Triggered Pendekatan Time-Triggered

Jaringan dalam Sistem Embedded Terdistribusi @2012,Eko Didik Widianto Sistem Embedded Terdistribusi Event- & Time-Triggered System Evolusi Komputing Terdistribusi Menambah Mikrokomputer

Evolusi Komputing Terdistribusi Menambah Mikrokomputer Remote Power Switched System Use Multiple Computers Fully Distributed Control Network Why Distributed Might Be Better Kelebihan Sistem Tersentral Modularitas Dalam Sistem Terdistribusi Diagnosabilitas dalam Sistem Terdistribusi Kelebihan Sistem Terdistribusi Lainnya Lisensi

Remote Power Switched System Use Multiple Computers Fully Distributed Control Network


No Computer at All


I

You don’t need a computer at all to operate light bulbs with switches!

I

Non-computer control is possible I

RLC circuits, analog transistors, etc.

Sistem Embedded Terdistribusi Event- & Time-Triggered System Evolusi Komputing Terdistribusi Menambah Mikrokomputer Remote Power Switched System Use Multiple Computers Fully Distributed Control Network


Why Add A Computer? I

I

I

Permits optimization I Using a many-to-many relationship for actuators to control sensors requires relays or other signal isolation (isolation relays would also be “computation” ) I Can change control loop behavior depending on system operating modes Enables adding sophisticated features in software I Safety interlocks I Require brake pedal to be depressed before shifting into gear I Active safety operations I Car locks doors once shifted into gear I Timers, counters, conditional logic I Variable interval wipers that adapt to rain intensity Enables context-sensitive operation I Switches can be momentary closure “soft switches” I Permits controller to set value even when switch isn’t pressed I Example: rear view window heaters off after timeout or if outside temp is hot

Jaringan dalam Sistem Embedded Terdistribusi @2012,Eko Didik Widianto Sistem Embedded Terdistribusi Event- & Time-Triggered System Evolusi Komputing Terdistribusi Menambah Mikrokomputer Remote Power Switched System Use Multiple Computers Fully Distributed Control Network


Sistem Tersentral


I

Central computer I I I

Reads input sensors Provides motive power to actuators Common ground (e.g., metal vehicle frame) used as electrical return path



Wiring



Wiring



Apa yang Bisa Ditingkatkan Berikutnya?

Jaringan dalam Sistem Embedded Terdistribusi @2012,Eko Didik Widianto Sistem Embedded Terdistribusi Event- & Time-Triggered System

I

Computer has to switch power to actuators I

I

I

Power switching components can be much larger than microcontroller Power-carrying wires can form a thick, heavy wire bundle – every conductor has to carry activation power, not just a signal

So, let’s move power switching out to periphery of system

Evolusi Komputing Terdistribusi Menambah Mikrokomputer Remote Power Switched System Use Multiple Computers Fully Distributed Control Network







Remote Power Switched System I

Actuator outputs from computer are control signals I

I

I

Power is fed to remote switching modules that obey control signals from CPU No computational power in power switches – can be just power transistors Blue lines are low-current logic signals, not high-current switched power



Mengapa Menggunakan Remote Power Switching? I

Thinner, lighter centralized cable bundles I

I

I

Pure centralized approach requires all power to touch CPU, causing congestion Control signal wires can be thinner, lighter, and more flexible than power-controlled actuator wires Remote switched power can be run from power supply directly to switching nodes I I

I

Cable can daisy-chain from region to region Yes, the power cable needs to be thicker since it handles all power loads, but not as thick as a bundle of switched power lines

This sets the stage for networking in the next evolutionary step I

I

Remote power switching isn’t a convincing win on wiring, but is a good start What if we replace all the analog control wires with a network?








Use Multiple Computers I

It doesn’t take much more to put a CPU at each remote power switch I I

I

We already have housing, circuit board, power distribution Compared to that, a $1 microcontroller can be an acceptable additional cost

But, analog signals from central CPU to remote CPUs is inefficient I

So, let’s change multiple analog signals to serial digital communications



Menambah CPU di Blok Sensor - Aktuator


I

Many sensors are near actuators I I

I

Sensor/actuator pairs are used to close control loops Many embedded systems aren’t that big, so it’s hard to be too far from a CPU

Add CPUs so that all sensors & actuators connect to a remote CPU I I

I

This changes to a fully distributed approach “Central” CPU might not really exist, just “big” and “small” distributed CPUs While we’re at it, switch to a network instead of point-to-point wiring

Event- & Time-Triggered System Evolusi Komputing Terdistribusi Menambah Mikrokomputer Remote Power Switched System Use Multiple Computers Fully Distributed Control Network


Jaringan Kontrol Terdistribusi


I

Both power and control signals flow across a shared wire I

Remote power switch nodes have a small computer for networking



Contoh: Central Control + 4 Corner Controllers








Fully Distributed Control Network


I

Main difference is no central CPU I

I

No “Brain” node calling the shots – all the computation is distributed If all the remote CPUs can handle functionality, do you need a central CPU? I

(answer depends mostly on the software architecture)



Tradeoff Kontrol Embedded Terdistribusi I

Multiplexing control wires saves weight, wire cost, cable thickness I I I

I

One digital wire replaces multiple analog wires Network must be fast enough to keep control loops closed Can use one wire per distribution node if network bandwidth is a concern

Network interface computer added to remote switching nodes I

Interfacing to even a simple network requires computer-like capability I

I

I

In simplest case, computer just “muxes” wires

Local computer’s job is to translate control signals and switch power locally

More complicated computers permit functions to migrate I

I

Once we have a remote computers, why not do computation there beyond just network interface? Carried to its logical conclusion, don’t even need the central computer anymore I

But, doing this requires a significant rethinking of software architecture



Fine Grain Distributed System


I

Each sensor and actuator also has CPU, power switching, network connection

Event- & Time-Triggered System Evolusi Komputing Terdistribusi Menambah Mikrokomputer Remote Power Switched System Use Multiple Computers Fully Distributed Control Network


Why Distributed Might Be Better

Jaringan dalam Sistem Embedded Terdistribusi @2012,Eko Didik Widianto Sistem Embedded Terdistribusi Event- & Time-Triggered System Evolusi Komputing Terdistribusi

I

Lots of little things can be better than one big thing I

Extensibility / Flexibility / Task Partitioning

Why Distributed Might Be Better Kelebihan Sistem Tersentral Modularitas Dalam Sistem Terdistribusi Diagnosabilitas dalam Sistem Terdistribusi Kelebihan Sistem Terdistribusi Lainnya

Lisensi

Evolusi Elektronika Otomotif Source: [Kassakian96]) I

Currently 12V DC power; 800 Watts I

Old style: centralized power fed to actuators via electromechanical switches I

I

New style: distributed computers each with several sensors/actuators I

I I

I

Wiring harnesses with point-to-point connections

Local wiring from sensors/actuators to each ECU (Electronic Control Unit) Global wiring/bus among ECUs Power switched within ECU for light loads, with relays for heavy loads

Future: perhaps 42V DC power; 2.5 KW average / 12 KW peak I I

Higher voltage is more efficient for motors and distribution Use local switching at sensors/actuators to reduce wiring I

I

Also put a CPU at every sensor/actuator

Maybe an AC distribution bus so transformers can be used

Jaringan dalam Sistem Embedded Terdistribusi @2012,Eko Didik Widianto Sistem Embedded Terdistribusi Event- & Time-Triggered System Evolusi Komputing Terdistribusi Why Distributed Might Be Better Kelebihan Sistem Tersentral Modularitas Dalam Sistem Terdistribusi Diagnosabilitas dalam Sistem Terdistribusi Kelebihan Sistem Terdistribusi Lainnya

Lisensi



Lisensi

Kelebihan Sistem Tersentral A Centralized System Justification


I

Simple programming model I

I

Powerful CPU(s) I

I

I

Can use CPU for any needed function (can use desktop PC sometimes) Can adapt CPU loading to operating mode

Better operating environment for digital electronics I

I

Ability to think about distributed architectures is an uncommon skill

Put machine in sheltered area away from combustion, environment

Arguably simpler software configuration I

All changes are made in one place in the system

Event- & Time-Triggered System Evolusi Komputing Terdistribusi Why Distributed Might Be Better Kelebihan Sistem Tersentral Modularitas Dalam Sistem Terdistribusi Diagnosabilitas dalam Sistem Terdistribusi Kelebihan Sistem Terdistribusi Lainnya

Lisensi

Kapan Sistem Tersentral Perlu “Dipertimbangkan”


I

Total system cost/weight I I

I

Housing + cooling costs may outweigh wiring savings Distributed system has components in harsher environment than central systems

System expandability I I

Central system has limit on I/O connectors Distributed system has limit on bus fanout (typically 32 nodes) I

I

I

But, arguably easier to install repeaters/bridges

Distributed system has limited communication bandwidth (compared to backplane)

Inventory costs I

Distributed systems have cheaper components, potentially

Event- & Time-Triggered System Evolusi Komputing Terdistribusi Why Distributed Might Be Better Kelebihan Sistem Tersentral Modularitas Dalam Sistem Terdistribusi Diagnosabilitas dalam Sistem Terdistribusi Kelebihan Sistem Terdistribusi Lainnya

Lisensi



Lisensi

Modularitas Dalam Sistem Terdistribusi I

Modular system development, support, and evolution I I

I

I

A different team designing each node Well-defined, tightly enforced interface (system message formats) Can upgrade individual models and limit effect of changes on rest of system

Limits competition for resources among different features I I

Can add compute+I/O power incrementally But, wastes resources on a node that might be inactive most of the time I

I

Difficult to share compute resources

Reduces interactions I

I I

Easier to make worst-case guarantees on a per-module basis Can re-certify only modules that have changed Can have “critical” and “non-critical” modules, reducing certification effort


Lisensi

Contoh: Subsistem Mobil


I

Different subsystems often provided by different vendors I

I

Engine, transmission, anti-lock brakes, power windows, etc.

Different CPU for each subsystem means: I

I

I

Each vendor has a CPU all to itself – minimal software integration issues Can change any component without worrying about details of internal software affecting other subsystems Change to convenience subsystem (windows) can be easily shown to have no effect on safety critical subsystem (brakes)

Sistem Embedded Terdistribusi Event- & Time-Triggered System Evolusi Komputing Terdistribusi Why Distributed Might Be Better Kelebihan Sistem Tersentral Modularitas Dalam Sistem Terdistribusi Diagnosabilitas dalam Sistem Terdistribusi Kelebihan Sistem Terdistribusi Lainnya

Lisensi



Lisensi

Diagnosabilitas dalam Sistem Terdistribusi


I

Very often this is the decisive advantage I

I

Decreasing maintenance and repair costs can make a big difference!

Remote diagnosability I

Can isolate problems based on available processors I

I

Especially useful on systems where cables and connectors have high failure rates

General ideas: I I

I

I

Each controller can diagnose its own sensors/actuators Each controller can diagnose its local connection to the network Each controller can determine which other controls seem to be alive Sampling a few controllers or network segments identifies problem areas


Lisensi

Contoh Diagnosabilitas


Lisensi



Lisensi

Kelebihan Sistem Terdistribusi Lainnya


I

Flexibility I

I

Robust data transmission I

I

Single bus means you can’t hook the wrong wires up – there is only one “wire”!

Enables fault tolerance I

I

Digital network lets you use error coding, controlling noise on signals

Simpler to build and maintain I

I

Can modify or upgrade systems by changing components

A single CPU is a single point of failure – multiple CPUs support fault tolerance

Improves safety certifiability I

Separate CPU for critical functions means non-critical CPU can’t cause safety faults


Lisensi

Flexibility di Sistem Terdistribusi


I

Can add new components more easily I

I

I

Multiple vendors can add components to a well defined HW+SW standard interface New components can have different physical size/shape as long as they can interface to the network

Scalable systems can be created on a pay-as-you-scale basis I

More copies of components added as system grows I

I

(But, there are limits before repeaters are needed for network)

But, individual node packaging might be too much overhead if most systems have only 2 or 3 copies of a component I

A single module with a couple long signal wires might be cheaper than a couple modules with a network wire


Lisensi

Contoh: Tombol Panggilan Elevator


I Can build using two standard units: I I

Two-button unit for middle floors One-button unit for end floors (spin arrow around for top floor)

I Adding more floors involves adding more button boxes to network I

Cost of long wiring runs makes using a network worthwhile

I But, for 2 or 3 floor lifts (which are the most prevalent): I I

Middle floor buttons aren’t the majority of those used Might be cheaper just to use a single controller box with discrete wires

Lisensi

Transmisi Data yang Handal


I

Analog data suffers from noise

Event- & Time-Triggered System

I

Digital data can be noise resistant

Evolusi Komputing Terdistribusi

I I

I

Digital sensor/actuator data provides better tradeoffs I I

I

Error detect & retransmit Error correcting codes

Lower bit rate gives better SNR Arbitrarily high precision at cost of bit rate

Once you have digital data transmission, you also get a distributed processor

Why Distributed Might Be Better Kelebihan Sistem Tersentral Modularitas Dalam Sistem Terdistribusi Diagnosabilitas dalam Sistem Terdistribusi Kelebihan Sistem Terdistribusi Lainnya

Lisensi


Lisensi

Jaringan dalam Sistem Embedded Terdistribusi

Lisensi

@2012,Eko Didik Widianto

Creative Common Attribution-ShareAlike 3.0 Unported (CC BY-SA 3.0) I Anda bebas: I I

untuk Membagikan — untuk menyalin, mendistribusikan, dan menyebarkan karya, dan untuk Remix — untuk mengadaptasikan karya

I Di bawah persyaratan berikut: I

Atribusi — Anda harus memberikan atribusi karya sesuai dengan cara-cara yang diminta oleh pembuat karya tersebut atau pihak yang mengeluarkan lisensi. I

Sumber asli tidak boleh dihilangkan, yaitu

http://didik.blog.undip.ac.id/2012/03/06/ kuliah-tsk-612-sistem-embedded-terdistribusi-2011/ I

Pembagian Serupa — Jika Anda mengubah, menambah, atau membuat karya lain menggunakan karya ini, Anda hanya boleh menyebarkan karya tersebut hanya dengan lisensi yang sama, serupa, atau kompatibel.

I Lihat: Creative Commons Attribution-ShareAlike 3.0 Unported License


2012. Eko Didik Widianto. Teknik Sistem Komputer - Universitas Diponegoro

Recommend Documents