ARUM: an Approach for the Resilience of Ubiquitous Mobile Systems Marc-Olivier Killijian David Powell Nicolas Rivière Matthieu Roy Gaétan Séverac Christophe Zanon Workshop Internet of Things - october 21st 2008 1
Internet of Things! Internet of Things " Ambiant Intelligence, " Ubiquitous Systems! So what? " Connecting zillions of nodes " Always evolving! Disconnections/new services, versioning, etc. " Open to new (untested/malicious/...) things " Mobile things/objects/devices " Tightly integrated into users environment/life " Critical as a whole 2
ARUM Motivation! Need to increase resilience of AmI/IoT/UbiSys! Evaluation of such systems is difficult " Usability, Acceptance, Performance, Resilience " Analytically: need for models, tools, etc. " Experimentally: need for platforms, prototypes, benchmarks, metrics, data, applications, etc.! Why is it so difficult? " What are ubiquitous systems? " Scale (#/size devices/environment) " What to evaluate? (user xp, network, etc.) 3
Evaluation vs. Simulation! Evaluation usually based on simulation " Network simulators (ns/2, glomosim, etc.) " Simple mobility models! Random drunken! Manhattan! Simulation is not satisfactory for critical systems " Cavin et al. 2002 : significant divergence exist between the [different] simulators " Very little work on simulation/fault-injection at node level [Goswami97]! Mobility models meaningless for most applications 4
ARUM : emulation & realistic models/ parameters! Emulation as a complement to simulation " Real prototype of a complex software stack! Runs on real hardware, i.e. uses real device drivers, communication stacks, etc.! Embeds real middleware, real FT mechanisms! Is a full prototype of the application " No details can be simplified " Fault-injection! Use of real mobility/usage data for " Realistic mobility/failure/usage models " Realistic analytical evaluation parameters 5
ARUM : emulation & realistic models/ parameters! Emulation as a complement to simulation " Real prototype of a complex software stack! Runs on real hardware, i.e. uses real device drivers, communication stacks, etc.! Embeds real middleware, real FT mechanisms! Is a full prototype of the application " No details can be simplified " Fault-injection Emulation platform! Use of real mobility/usage data for Mobility models " Realistic mobility/failure/usage models " Realistic analytical evaluation parameters 5
ARUM : Mobility/Fault Models in UbiSys! For resilience evaluation in ubiquitous scenarios! Mobility/Connectivity models! Fault models! Energy consumption models! The ARUM LAAS Project! owner fails OD (owner down) DL (data lost) OU (owner up) DS (data safe) FC (fragments to create) # m(mf).#' m(mf).!' owner meets infrastructure m(mf)+m(sf) < k! R. Diaz s PhD started september 2008! Collect real data on different pops! Humans using GPS Smartphones! Animals using GPS collars! Build realistic models! for simulation, emulation, prototyping, etc contributor meets infrastructure owner meets contributor n " SF (safe fragments) m(sf)! k MF (mobile fragments) contributor fails 6
ARUM Emulation Platform - Scale issue! Ubiquitous systems can be very diverse " VANETs " Urban social networking " Nano-robots! Can we have a unified approach " Scale, scale, scale 7
Experimental platform! Typically composed of fixed and mobile devices " Programmable mobile platform " Light processing unit " Wireless comm. interfaces (adhoc+infra) " Positioning device! Dedicated laboratory " Infrastructure! Computing, communication, positioning " Reproducible experiments " Fault injection 8
Scalable lab!! Different prototypes # different scales " Communicating vehicles (VANETs)! Device: 3m! Environment: 1km road! Communication range: 100s m " Cooperating nano-robots! Device:!1µm! Environment:!20µm vessel! Communication range: 10s µm! Scale increase or decrease " Factor from 10-6 to 10 2 50 10-6 9
Distributed BlackBox Example! Exemplifies the emulation platform " critical application " mobile devices (C2CC) " distributed protocol (P2P)! Collects and saves car/env data " Status of dr.wheel/pedals " Speed, orientation, position " Surrounding vehicles " etc.! Not hardened hardware but distributed algorithm 10 www.hidenets.aau.dk 10
What are the issues to solve! Technical issues " Precise indoor positioning " Programmable mobility " Range-controlled communication 11
t Precise Indoor Positioning! Desired precision " In-vehicle GPS " 5m " Scale reduction factor 50! Indoor precision " 10cm! Several technologies " Scene analysis (motion capture) " Triangularization (RF, ultra-sound, UWB, etc.)! Evart positionning system " sub-mm precision @ 100 Hz 43600 2D(xt) evart-stops precise-scale 43550 43500 43450 43400 43350 43300 43250-500 -400-300 -200-100 0 100 200 300 400 500 x 12
Programmable Mobility! Reproducible Mobile experiments " Small robot platforms! Carry PDA or laptop! LynxMotion 4WD! Different designs " Tape tracks! 20cm/s for a few hours " Autonomous version under development! Increased programmability/precision/speed! Use of LAAS robotic architecture 13
Programmable Mobility! Reproducible Mobile experiments " Small robot platforms! Carry PDA or laptop! LynxMotion 4WD! Different designs " Tape tracks! 20cm/s for a few hours " Autonomous version under development! Increased programmability/precision/speed! Use of LAAS robotic architecture 13
The BIG issue: wireless communication! How to scale down WIFI? " Hundreds meters # few (2-3) meters " Precisely and controlled! Potential solutions " Via emulation " Reducing Tx power of adapters! Access driver! Faraday cages, tents, etc.! Signal attenuators 14
The BIG issue: wireless communication! How to scale down WIFI? " Hundreds meters # few (2-3) meters " Precisely and controlled! Potential solutions " Via emulation " Reducing Tx power of adapters! Access driver! Faraday cages, tents, etc.! Signal attenuators 14
The BIG issue: wireless communication! How to scale down WIFI? " Hundreds meters # few (2-3) meters " Precisely and controlled! Potential solutions " Via emulation " Reducing Tx power of adapters! Access driver! Faraday cages, tents, etc.! Signal attenuators 14
Scaled WiFi : Results 100% exp2.32db! Final experiments! Static and mobile objects! C2C: 1.5m -> 32db % communication 75% 50% 25% 0% 4m 3m 2m 1m 0 1m 2m 3m 4m 120 exp2.32db/exp2.32db.3.b 100 80 60 40 20 0 4m 3m 2m 1m 0 1m 2m 3m 4m 15 15
LAAS Hidenets Platform Cooperative Data Backup Middleware Proximity Map Location Svce Networking Svce Trust&Coop. Ocle OS Linux kernel 2.6.17 Hardware Location HW Reduced WiFi SmartCard Robot Platform 16 www.hidenets.aau.dk 16
Scenario A B C Signal 2 Signals 1 Vehicles A B signal 1 = accelerate/decelerate signal 2 = change mode: a-next turn left b-next turn right C go a little faster, when obstacle = go straight forward and crash 17 17
Future Work! Experimental evaluation " Complete the evaluation prototype " Extend the mobility aspect " Develop fault-injection techniques " Develop benchmarks! Analytical evaluation and mobility models " Develop a data collection platform " Extract and anonymize data " Synthesize models from data " Evaluate models 18
Conclusion! Ambiant Intelligence/Ubiquitous/Internet of Things " Tremendous need for user trust/confidence " Resilience (Dependability + Evolution)! Evaluation of Resilient UbiMob Systems " Large open field! Evolvability! Privacy " ARUM Contribution! Analytical evaluation! Experimental evaluation 19