For the Real-‐Time Internet of Things

Transcription

For the Real-‐Time Internet of Things
RIOT: For the Real-­‐Time Internet of Things
Team 7 : Michael Warrick, Ethan McGee
Internet of Things
My teddy bear is connected to the internet.
Internet of Things
 The Internet of Things (IoT) is the interconnecEon of uniquely idenEfiable embedded compuEng devices within the exisEng Internet infrastructure
 It comprises a variety of fields, domain, applicaEons and protocols
 Expected to usher in a huge degree of automaEon across nearly every discipline imaginable
Examples
 Environmental Monitoring
  Electric Clams in Ocean to monitor Clam PopulaEon
  Intelligent River Project
 Household
  Washer / Dryer Monitoring
  Thermostats / Fire Detectors
 Health
  Pace Maker Monitoring
  Health Blueprints
RIOT: Real-­‐Time Internet of Things
One OS to Rule Them All, to Find Them, One OS to Bring Them, and in the Cloud Bind Them
Why RIOT?
 Scenario: Airport
 IoT Requirments
  Hardware
 
 
Memory and Processing Power
Pla3orm Support
  Autonomy
 
 
 
Power Efficiency
Reliability
Networking Stack
  Programmability
 
Language Support
A Comparison
Features
MarkeEng Time!!!!
Hardware Support
Somewhat Limited Range
Airfy Beacon
Arduino Due
Arduino Mega 2560
Atmel samr21-­‐Xplained Pro
f4vi
HiKoB FOX
mbed NXP LPC1768
MSB-­‐IoT
Nordic nrf51822 (DevKit)
OpenMote
Spark-­‐Core
STM32F4DISCOVERY
STM32F3DISCOVERY
STM32F0DISCOVERY
ScaferWeb MSB-­‐A2
ScaferWeb MSB-­‐430H
TelosB
Texas Instruments cc2538 Developer Kit
Texas Instruments EZ430-­‐Chronos
UDOO Board (Cortex-­‐M3 part)
WSN 430 (v1.3b and v1.4)
yunjia-­‐nrf51822
ZolerEa Z1
Sensors
-­‐ Radio Transceivers
-­‐ Environmental Sensors: Humidity and Temperature Sensor, Pressure Sensor, Alcohol Gas Sensor
-­‐ Bafery Gas Gauge
-­‐ AcceleraEon Sensors: Tri-­‐axial Sensor, Compass and Accelerometer
-­‐ Gyroscopes: Three-­‐Axis Digital Gyroscope
-­‐ Ultra Sonic Range Finders
-­‐ Light: RGB LED, Light Sensor
-­‐ Servo Motor
Reliable
By Inheritance & Design
 OS derived from an FeuerWare (used by most embedded devices in rescue scenarios)
 Kernel is kept extremely small
  MulEthreading
  Mutexes
  Inter-­‐Process CommunicaEon
 Device drivers cannot run in kernel   Must run in thread
  Cannot crash kernel MulE-­‐Threading
Full Support
 Does not rely on explicit yields (unlike TinyOS)
 Threads can be switched out of context (unlike ConEki)
Networking Support
Network Stacks
Other Features
 Wiselib Support
  Algorithm library designed for embedded devices
 
Rou?ng, clustering, security, ?me sync, localiza?on
  MulEplajorm
 
 
 
RIOT
Con?ki
Shawn, etc
 JSON Support
 Security Support– (ie SHA 256, 3DES, RC5)
C++ 11 Support
Awesome Stuff!!
 Futures / Promises
  Extremely popular in the JavaScript world
  Helps to make handling return values from threads easier
  h#p://en.cppreference.com/w/cpp/thread/future  Tuples
  Easier to return mulEple values from a single funcEon
  std::tuple<bool, int> someFuncEon(int x, int y);
 
yes, you could just use a struct
NaEve Mode
NaEve Mode
 Runs on Mac or Linux
 Provides simple tesEng and experimentaEon
 InstallaEon (Ubuntu Linux)
  gcc-­‐mulElib
  bridge-­‐uEls
  (tuntap needed for mac, preinstalled on ubuntu)
 Makefile templates provided
  Specify target hardware/modules
Bloom Filters
Bloom Filter
 An extremely fast data structure that allows you to test for inclusion within a set
 ProbabilisEc data structure
 Extremely fast comes at a price
  Bloom filters can tell you if something is not in the set
  Bloom filters can tell you if something might be in the set
 h#p://billmill.org/bloomfilter-­‐tutorial/ Shell
Shell
 Provides basic shell access
  Not a full shell – specific commands available
 Runs on serial port/uart
 Commands:
  addr
  reboot
  route
ImplementaEon
And you thought the stuff on Bloom Filters was bad...
Modularity
h#ps://github.com/RIOT-­‐OS/RIOT FireKernel
Real Eme support
Scheduler
Scheduler
 Uses a pre-­‐empEve Eming scheduler
  A Eme-­‐slice scheduler is not needed because there is no need to maintain the appearance of a responsive GUI
  Reduces interrupt load on system
 Main idea
  Priority based scheduling
  Every task gets a unique priority
  Tasks run unEl compleEon unless someone with higher priority comes along
RIOT's Scheduler
 Tasks with same priority are possible
  Must cooperaEvely share Eme
 32 possible prioriEes
 PrioriEes on same level are kept in linked list
 Switches done in almost constant Eme
Blocking
ObservaEons
 If a high priority task does not finish in Eme, it would not have finished in Eme even on a single-­‐threaded system
  All processing Eme is devoted to it
  Allowed to run to compleEon
 If a lower priority task does not finish in Eme, probably doesn't have the right priority
  Increased priority means more processing Eme
  Won't be interrupted
Energy Efficiency
Idle Task
 It is a special task created during the OS boot  Does not end but has lowest priority
 Used to help achieve ulEmate power efficiency
 Only task is to determine which low-­‐power mode the OS can enter
 Can only be exited by hardware interrupt
Tickless Timer
 Most kernels: a repeated Emer interval set (e.g. 1 microsecond)
  Generates and interrupt
  System wakes to update data structure, compare
 Tickless:
  No Emer is kept by the kernel
  Timer Architecture provided to applicaEons
 
Mapped to hardware
  MulEple registers used
Interrupts
Overview
 Most systems have interrupts bypass OS because the OS can't handle interrupts
 RIOT is different
  Interrupt made via OS API call
  Allows for befer context switching
  Easier return to original context
 Can be wrifen using ISR notaEon used with other systems, or using RIOT's API
  ISR in RIOT is a wrapper around it's API
Diagram
Latency
 Needs to be as small as possible
 In RIOT, obviously higher because OS is involved
 However sEll extremely low
 50 clock cycles or 700 nanoseconds
Mutexes
Mutexes
 Implemented as a Stack (-­‐ish)
 When a mutex is encountered
  Push self onto stack unless top of stack has higher priority
  Then use random access to find the first entry with a lower priority than you
  Enter blocked mode
  Issue context switch
 When criEcal region is done
  Pick first thread off top of stack
Inter-­‐Process CommunicaEon
IPC
 Synchronous message passing
 Send thread blocks unEl a thread receives
  Similar to mutex implementaEon
Examples
Yes, we are actually going to present something pracEcal for change .
QuesEons / Comments?
Whew... bet you're glad that's over!