networking infrastructure and data management for cyber ... · networking infrastructure and data...
TRANSCRIPT
![Page 1: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/1.jpg)
Networking Infrastructure and Data Management for Cyber‐Physical
S tSystems
Song HanSong Han
Computer Science Department, University of Texas at [email protected]
www.cs.utexas.edu/~shan
![Page 2: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/2.jpg)
What is Cyber Physical System (CPS)?What is Cyber‐Physical System (CPS)?
Cyber‐physical system is a system featuring a tight combination of, and coordination between, the system’s computational and h i l l tphysical elements.
2
![Page 3: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/3.jpg)
CPS Application Cyberphysical AvatarCPS Application – Cyberphysical Avatar
Skill AcquisitionDynamic Model Skill Acquisition through Machine
Learning
Dynamic Model and Control
Structure Design
R l iReal‐time Avatar‐Human Interaction
Prototype Testbed
Cyberphysical Avatar: A semi‐autonomous robotic system (joint project with
3
Cyberphysical Avatar: A semi autonomous robotic system (joint project with UT Human Centered Robotics Lab)
![Page 4: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/4.jpg)
CPS Application – Network‐based Mobile Gait Rehabilitation System
• Integrating heterogeneous sensors into real‐time wirelesssensors into real time wireless platform
• Low level motion control of• Low‐level motion control of rehabilitation device over wireless network
• Development of high‐level decision making algorithm
Network based Mobile Gait Rehabilitation System (joint project
g g
4
Network‐based Mobile Gait Rehabilitation System (joint project with Mechanical Systems Control Laboratory, UC Berkeley)
![Page 5: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/5.jpg)
Research Overview
Theoretical Framework for Real‐time Data Management Techniques
Real‐time Wireless Communication Platform
5
![Page 6: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/6.jpg)
Guiding Applicationsg pp
Cyberphysical AvatarNetwork‐basedRehabilitation System
Remote and Real‐time Welding System
6
![Page 7: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/7.jpg)
OutlineOutline• Research Overview
• Reliable and Real‐time Wireless Platform for CPS– Wireless real‐time communication protocol– Network management techniques
S d i d i l i– System design and implementation
• Real‐time Data Management in CPSReal time Data Management in CPS– Model and assumptions– Algorithms and analysisg y
• Summary and Future Work
7
![Page 8: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/8.jpg)
Wireless Reliable and Real‐time Communication Platform
![Page 9: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/9.jpg)
Design Space and Required FeaturesDesign Space and Required Features
LLow‐power– 802.15.4‐based radio
Real timeReal‐time– TDMA Data Link Layer (DLL) – Centralized management
Reliable– Mesh networking– Data link layer ACK– Channel hopping mechanism
SecureSecure– Data integrity on DLL– Data confidentiality on
9
ynetwork layer (NL)
![Page 10: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/10.jpg)
Overview of Our Real‐time Protocol StackOverview of Our Real‐time Protocol Stack
• TDMA‐based Data Link Layer– Guarantee timely deliveryGuarantee timely delivery
• Channel Hopping and Blacklisting– Spread communication in all active physical channels– Reduce interference to provide reliable communication
• Confidential and Secure Communication– Use both public and private keys to secure communication inUse both public and private keys to secure communication in
both join process and normal operations
10
![Page 11: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/11.jpg)
TDMA based Data Link LayerTDMA‐based Data Link Layer
10ms
11
![Page 12: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/12.jpg)
TDMA based Data Link LayerTDMA‐based Data Link Layer
• Link: activity in a time slot Idle link
– Neighbor– Send/Receive
Send link
– Communication channel Receive link
12
![Page 13: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/13.jpg)
TDMA based Data Link LayerTDMA‐based Data Link Layer
• Superframe: a group of links– Repeat itself infinitely– A device can support several superframes
13
![Page 14: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/14.jpg)
TDMA based Data Link LayerTDMA‐based Data Link Layer
Li k S h d lLink Scheduler
14
Priority queues for data link layer packets
![Page 15: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/15.jpg)
How to Achieve Reliable and Real‐time Services in CPS
• Network Manager• Authenticating devicesut e t cat g de ces• Forming the network• Constructing routing graphs
S h d li DL t i i• Scheduling DL transmissions
• Gateway• Collecting/caching sensor data• Process queries from other systems
• Security Manager• Manage key information
15
![Page 16: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/16.jpg)
How to Achieve Reliable and Real‐time Services in CPS
• Communication task definitionNeed to solve two related sub problems:– Need to solve two related sub‐problems: 1. communication graph design 2 link scheduling
• Technical Objectives
2. link scheduling
Technical Objectives– Achieve reliable routing in wireless mesh networks– Achieve real‐time communication by deterministic link andAchieve real time communication by deterministic link and
channel assignment– Evaluate their performance in real industrial environments
16
![Page 17: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/17.jpg)
Communication Graph Design to Achieve Reliable Graph RoutingGraph Routing
To avoid forwarding loop:1) Only one cycle of length 2 in Gv2) Each DEV on the cycle has direct
edges to v
17
![Page 18: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/18.jpg)
Constructing Reliable GraphsConstructing Reliable Graphs
• Reliable Broadcast Graph and Uplink Graph– Grow the graph by greedily selecting the reliable node with g p y g y g
minimum latency to the Gateway
• Standard Reliable Downlink Graph– Construct a completely new graph from GW to DEV v– Configuration in intermediate nodes cannot be reused– High configuration cost and poor scalability
18
![Page 19: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/19.jpg)
Sequential Reliable Downlink Routing q g(SRDR)
• Key PrinciplesEach node only keep a small local graph– Each node only keep a small local graph
– Local graphs are reusable building blocks for constructing reliable downlink graph for multiple destinationsreliable downlink graph for multiple destinations
Low configuration costLow configuration cost
High ScalabilityHigh Scalability
High Reliability
19
High Reliability
![Page 20: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/20.jpg)
An Example of SRDRpG
A1 A2
G
A1 A2
G
A1 A2A1 A2
21 3
A1 A2
21 3
A1 A2
21 3
4 5
(b) Downlink graph: g2
4 5
(c) Downlink graph: g3
4 5
(a) Original network topology (b) Downlink graph: g2Sequential route for Dev 2: g2
(c) Downlink graph: g3Sequential route for Dev 3: g3
G G G
(b)
Avoid node failure at DEV2
A1 A2
21 3
A1 A2
21 3
A1 A2
21 3
Local graph
21 3
4 5
21 3
4 5
21 3
4 5
20
(d) Downlink graph: g1Sequential route for Dev 1: g2, g1
(e) Downlink graph: g4Sequential route for Dev 4: g2, g1, g4
(f) Downlink graph: g5Sequential route for Dev 5: g2, g5
![Page 21: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/21.jpg)
SRDR ExtensionsSRDR Extensions
A 2A2 4
A1 1
2 422A2
g2 g4
11A1
A1
gg1
TTLControl ASN Snippet
Graph ID
DestAddr
Source Addr
Proxy Route Payloadg2 g1 g4
21
Extended Routing Information
![Page 22: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/22.jpg)
SRDR Optimizationp
A 2A2 4
A1 1
2 422A2
g2 g4
11A1
A1
gg1
TTLControl ASN Snippet
Graph ID
DestAddr
Source Addr
Proxy Route Payloadg2 g1 g4
22
Extended Routing Information
![Page 23: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/23.jpg)
Communication Link SchedulingCommunication Link Scheduling
• The general scheduling problem is known to be NP‐hardness [Saifullah et al 2010]hardness [Saifullah et al. 2010]
• Key Principles:y p
– Spread out the channel usage in the networkApply Fastest Sample Rate First policy (FSRF)– Apply Fastest Sample Rate First policy (FSRF)
– Allocate the links iteratively from Src to Dest– Split traffic (bandwidth) among all successors– Split traffic (bandwidth) among all successors
23
![Page 24: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/24.jpg)
Example Schedule Construction Using the Key PrinciplesPrinciples
Ch Offset16
Channel offset will be converted into practical channel number in the runtime16
2
.
. Global Channel‐Time Slot Matrix
practical channel number in the runtime
Slot0
1
Slot100 200 300 400
Dev 1
Device Schedule
GDev 1
AP A
A BDev 2
AP B241 2 3
1 sec 2 sec 1 sec
AP B
Dev 3
![Page 25: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/25.jpg)
Performance Evaluation
25
![Page 26: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/26.jpg)
System Design, Implementation and Deployment
26
![Page 27: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/27.jpg)
System Design, Implementation and Deployment
Hardware Platforms
Freescale 1322x SRB Evaluation Board
Custom Designed Mother Board with Sensor Support
Custom Designed Board with EnergyMicro EFM32 MCU
27
![Page 28: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/28.jpg)
System Design, Implementation and Deployment (Cont.)
Compliance Testing Suite
Testing Engine 16‐Channel Sniffer Virtual Network Approach
28
![Page 29: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/29.jpg)
System Design, Implementation and Deployment (Cont.)
Network Manager and Simulator
Simulating a real‐time wireless gnetwork with 100 devices:
‐ reliable broadcast graph d h d l‐ device communication schedule
29
![Page 30: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/30.jpg)
System Design, Implementation and Deployment (Cont.)
Network Manager and Simulator
Simulating a real‐time wireless Simulating a real‐time wirelessgnetwork with 100 devices:
‐ reliable broadcast graph d h d l
gnetwork with 100 devices:
‐ reliable uplink graphd b d d h l‐ device communication schedule‐ device bandwidth utilization
30
![Page 31: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/31.jpg)
System Design, Implementation and Deployment (Cont.)
Application Layer CoAP APP Layer
Socket API
Transport Layer6LoWPAN
UDP ICMP
Network Topology CoAP‐HTTP Server
802 15 4 PHY
Data Link Layer
Enhanced NWK Layer
802.15.4 PHY
Intra‐system Service Web Service
![Page 32: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/32.jpg)
System Design, Implementation and Deployment (Cont.)
UT Austin ACES 5th floor UT Pickle Research UWO10 Device Testbed UT Austin ACES 5 floor UT Pickle Research Center
UWO Power House
10 Device Testbed
32
![Page 33: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/33.jpg)
Higher Sampling Rate Required in Network‐based h b lRehabilitation System
• Challenges• Challenges– Mechanic modules need high frequency and low jitter control
A platform for a wide range of wireless control applications: a good– A platform for a wide range of wireless control applications: a good balance among sampling rate, energy consumption and real‐time performance
33
![Page 34: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/34.jpg)
High‐speed Real‐time Wireless ControlHigh speed Real time Wireless Control
• Real‐time Wi‐Fi to support high speed controlpp g p– Replacing 802.15.4 PHY with 802.11 PHY– Network‐wide synchronization and power saving
Application Layer CoAP APP Layer
Transport LayerUDP ICMP
Socket API
RT‐WiFi ApplicationsApplication Domain
Enhanced NWK Layer
p y6LoWPAN
Currently supported Applications
802.15.4 PHYData Link Layer
Enhanced NWK Layer
802.11 PHY
Currently supported Applications
34
![Page 35: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/35.jpg)
Real‐time Data Management in CPSg
![Page 36: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/36.jpg)
Maintaining Data Quality in CPS is KeyMaintaining Data Quality in CPS is Key
– CPS are in essential sensing and control systemsCPS are in essential sensing and control systems – Data quality is the key to the success of sensing and control
applications– Sensor data have time semantics, and their quality degrade with time
36
![Page 37: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/37.jpg)
Maintaining Data Quality in CPS is KeyMaintaining Data Quality in CPS is Key
• Need to enable tradeoff between data quality and sampling rate– High sampling rate ‐> high network traffic & CPU workload– More power consumption & shorter network lifetime– Reduce sampling rate but maintain data and control quality
• Will exploit concept of validity interval to make the tradeoff
37
![Page 38: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/38.jpg)
Task Model
Real‐time DatabaseSensor Data
Update Task <Ciu,Vimin,Vi
max>Periodic Control Task
<Cic, Pic, Dic, Ωi>
A t k i b t ti f
i , i , i , i
Control Data
• A task is an abstraction of resource consumer; a task can be a computing task (consuming CPU cycles) or a communication task (consumingcommunication task (consuming network bandwidth)
• Validity intervals quantify theValidity intervals quantify the quality of sensor data
• Control data quality is a function of
38
sensor data quality
![Page 39: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/39.jpg)
Task Model
S d t t k t Tu { u} nSymbol Meaning
Task Model
• Sensor update task set Tu = {τiu}i=1n– τiu is a 4‐tuple: τiu = (Ciu, Vi
min, Vimax, Qi
u(t)).– Qu(t) is application‐dependent
τiu(c) Update/Control Task i
Ciu(c) WCET for τiu(c)
Q u(c)(t) Quality function for τ u(c)– Qi (t) is application‐dependent.
• Control task set Tc = {τic}i=1mQi
u(c)(t) Quality function for τiu(c)
Vimin (max) Min(max) validity interval
Dic (Pic) Deadline (Period) of τic
– τic is a 5‐tuple: τic = (Cic,Dic, Pic, Ωi, Qi
c(t)) – Ωi is the update tasks that τic will access and Qi
c(t) is application‐dependent
i i i
G l M i t i th t l d t lit bGoal: Maintain the control data quality above threshold while Minimizing update workload
39
![Page 40: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/40.jpg)
Task Model
S d t t k t Tu { u} nSymbol Meaning
Task Model
• Sensor update task set Tu = {τiu}i=1n– τiu is a 4‐tuple: τiu = (Ciu, Vi
min, Vimax, Qi
u(t)).– Qu(t) is application‐dependent
τiu(c) Update/Control Task i
Ciu(c) WCET for τiu(c)
Q u(c)(t) Quality function for τ u(c)– Qi (t) is application‐dependent.
• Control task set Tc = {τic}i=1mQi
u(c)(t) Quality function for τiu(c)
Vimin (max) Min(max) validity interval
Dic (Pic) Deadline (Period) of τic
– τic is a 5‐tuple: τic = (Cic,Dic, Pic, Ωi, Qi
c(t)) – Ωi is the update tasks that τic will access and Qi
c(t) is application‐dependent
i i i
• Simplifying Assumptions
Validity Constraint: An update job must finish before its previous job’s validity interval expiresSimplifying Assumptions
– No control task in the system for now– Vi
min = 0 and Vi = Vimax
Goal: Guaranteeing validity constraint while minimizing the
40
update workload.
![Page 41: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/41.jpg)
From Validity Interval Model to Periodic Task ModelFrom Validity Interval Model to Periodic Task Model
Validity Interval Task Model How to pick the time point toValidity Interval Task Model
V
How to pick the time point to perform sensing?
C C C
V
C C C
Periodic Task Model
0 t1 t2
P P P
C C C
D D D
41
0 t1’ t2’
![Page 42: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/42.jpg)
Maintaining Update Data Freshness‐ Baseline Scheduling Techniques
• HH (Half‐Half) Algorithm– Period (Pi) and relative deadline (Di) of an update task i are each set to
be one‐half of the data validity length (Vi).
Validity Length Vi
Pi = DiPi = Didi,1di,0 = ri,1
Pi Diri,0
i i
42
Ti,0 Ti,1
![Page 43: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/43.jpg)
Maintaining Update Data Freshness‐ Baseline Scheduling Techniques
• ML (More‐Less) Algorithm– Relative deadline (Di) of an update task i is set to be its worst‐case ( i) p
response time (WCRT). Period Pi = Vi ‐ Di
Validity Length ViWCRT
DiPiD di,1ri,1Di
ri,0i
di,0Di
43
Ti,0 Ti,1Release time is deferred to ri,1
![Page 44: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/44.jpg)
Deferrable Scheduling with Fixed Priority (DS‐FP)‐ From Periodic to Sporadic Task Model
PrinciplesPrinciples• Adopts the sporadic task model.
• Defers the sampling time of the update job as late as possible to increases the distance of two consecutive jobs.
![Page 45: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/45.jpg)
Deferrable Scheduling with Fixed Priority (DS‐FP)‐ From Periodic to Sporadic Task Model
ML S h d l (P i di )ML Schedule (Periodic)P1 = 10 D1 = 2P2 = 14 D2 = 4
T1: { C1= 2, V1= 12 } T2: { C2= 2, V2= 18 }
V 18P2 = 14 P2 = 14D2 = 4
V2 = 18
D2 = 4 D2 = 4
Release time of T2,1 is deferred from 14 to 16
0 5 10 15 20 25 30
DS‐FP Schedule (Sporadic)(Sporadic)
0 5 10 15 20 25 30V2 = 18Separation time is increased
![Page 46: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/46.jpg)
Deferrable Scheduling with Fixed Priority (DS‐FP)‐ From Periodic to Sporadic Task Model
ML S h d l (P i di )ML Schedule (Periodic)P1 = 10 D1 = 2P2 = 14 D2 = 4
T1: { C1= 2, V1= 12 } T2: { C2= 2, V2= 18 }
V 18P2 = 14 P2 = 14D2 = 4
V2 = 18
D2 = 4 D2 = 4
Deadline of T2,2 is deferred to 34
0 5 10 15 20 25 30
DS‐FP Schedule (Sporadic)(Sporadic)
0 5 10 15 20 25 30V2 = 18Separation time is increased
![Page 47: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/47.jpg)
Deferrable Scheduling with Fixed Priority (DS‐FP)‐ From Periodic to Sporadic Task Model
ML S h d l (P i di )ML Schedule (Periodic)P1 = 10 D1 = 2P2 = 14 D2 = 4
T1: { C1= 2, V1= 12 } T2: { C2= 2, V2= 18 }
V 18P2 = 14 P2 = 14D2 = 4
V2 = 18
D2 = 4 D2 = 4
0 5 10 15 20 25 30Release time of T2,2 is deferred from 28 to 32DS‐FP Schedule
(Sporadic)(Sporadic)
0 5 10 15 20 25 30V2 = 18Separation time is increased
![Page 48: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/48.jpg)
Deferrable Scheduling with Fixed priority (DS FP)Deferrable Scheduling with Fixed priority (DS‐FP)
DS‐FP significantly reduces the CPU workload incurred by update trans.
Lower priority tasks have larger relative avg. sampling periods
48
![Page 49: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/49.jpg)
Deferrable Scheduling with Fixed priority (DS FP)
C i f DS FP d ML
Deferrable Scheduling with Fixed priority (DS‐FP)
• Comparison of DS‐FP and ML– THEOREM. Given a synchronous update transaction set T with known Ci and Vi
,if for all i, fi 0ml ≤ Vi / 2, then T is schedulable with DS‐FP., , i,0 i ,
• Necessary and Sufficient Schedulability TestTHEOREM Gi d k T if i b h d l d b DS FP i h– THEOREM. Given an update task set T, if it can be scheduled by DS‐FP in the bounded time interval [0, Vm – Cm + ∏i=1
m (Vi – Ci + 1) ‐ 1], then the schedule has a repeating pattern that must occur at least once in the bounded time interval [V C V C +∏ m (V C + 1) 1]interval [Vm – Cm, Vm – Cm + ∏i=1
m (Vi – Ci + 1) ‐ 1].
• Overhead Reduction Algorithms– DS with Hyperperiod by Schedule Construction (DESH‐SC)– DS with Hyperperiod by Schedule Adjustment (DESH‐SA)
49
![Page 50: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/50.jpg)
CPS Real‐time Data Management Research RoadmapCPS Real time Data Management Research Roadmap
Half‐Half
Maintaining data freshness for fixed Periodic
k d l
Half‐Half
More‐Less Sporadic task model
No Jitter
freshness for fixed update task set task model JBML
SJBML
task model
Jitter
Algorithm and lanalysis
Overhead Maintaining data
freshness in flexible Co‐scheduling of update tasks and Deferrable
h lreduction
Schedulability
cyber‐physical systemspcontrol tasks Scheduling
50
test
![Page 51: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/51.jpg)
Research Summaryy
• Reliable and real‐time wireless platform for CPSReliable and real time wireless platform for CPS– Wireless real‐time communication protocol– Network management techniquesNetwork management techniques– System design and implementation
• Theoretical framework for real‐time data management in CPSmanagement in CPS– Models and assumptions– Algorithms and schedulability analysisAlgorithms and schedulability analysis
51
![Page 52: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/52.jpg)
Ongoing and Future WorkOngoing and Future Work
• I believe that the next Internet resolution will be about the delivery of• I believe that the next Internet resolution will be about the delivery of physical services in addition to information services over long distances.
• The economic and social impact will be enormous.
52
![Page 53: Networking Infrastructure and Data Management for Cyber ... · Networking Infrastructure and Data Management for Cyber‐Physical StSystems Song Han Computer Science Department, University](https://reader030.vdocuments.mx/reader030/viewer/2022041222/5e0c6207ad206f7ab4419b2a/html5/thumbnails/53.jpg)
Thanks and Questions?