A ServiceDifferentiated RealTime Communication Scheme for WSNs

1
A Service-Differentiated Real-Time Communication
Scheme for WSNs

• Y. Xue, B. Ramamurthy M.C. Vuran
• University of Nebraska-Lincoln
• Presented _at_ SenseApp 2008

2
Outline

• Motivation and Design Considerations
• Design of A Service-Differentiated Real-Time
  Communication Scheme (RCS)
• RCS Performance Evaluation
• Conclusions and Future Work

3
Motivation

• Event-Based Real-time Applications
• Intrusion detection, border control, industrial
  process control
• Traffic type Upstream converge-cast
• Restricted end-to-end deadline requirements ?
  Diverse end-to-end/per-hop latency

4
Design Goals

• Service differentiated soft real-time
• E2E latency subject to the required deadlines
• Accurate priority classification
• Admission control and early-drop policy
• Ready to use on current hardware platform
• Constrained memory and energy
• No accurate synchronization and localization
• Adaptive to network dynamics
• Node failure
• Wireless channel fading
• Local traffic jamming

5
Related Work

• Prioritized Medium Access Control (MAC)
• TDMA? 1
• Require accurate synchronization
• High control overhead for light traffic
• Or CSMA/CA?
• Related work 802.11EDCA 2
• Challenge tradeoff between average throughput
  available service level

1 B. D. Bui, R. Pellizzoni, and etc., Soft
real-time chains for multi-hop wireless adhoc
networks, in Proc. of RTAS 2007, April 2007. 2
IEEE802.11WG, Draft supplement to IEEE
standard802.11-1999 Medium access control (MAC)
enhancements for quality of service (QoS), 2003.
6
Related Work

• Optimized routing for minimizing end-to-end
  latency
• Table-based Routing? SPEED 3, MMSpeed 4
• Network Dynamics -gt High control overhead
• Incompatible to duty-cycle design
• Geographic Forwarding? RAP5
• no channel quality aware
• Or Dynamic Forwarding? GeRaf 6, XLM 7,
• no real-time requirements have been considered in
  forwarding metrics

3 T. He, J. Stankovic, and etc., A
spatiotemporal communication protocol for
wireless sensor networks, IEEE Tran on Parallel
and Distributed Systems, May 2005. 4 E.
Felemban, C. Lee, and E. Ekici, MMSPEED
Multipath multi-speed protocol for qos guarantee
of reliability and timeliness in wireless sensor
networks, IEEE Transactions on Mobile Computing,
June 2006. 5 C. Lu,, T. Abdelzaher, and etc,
RAP A real-time communication architecture for
large-scale wireless sensor networks, in Proc.
of RTAS 2002, September 2002. 6 I. F. Akyildiz,
M. C. Vuran, and O. B. Akan, A cross layer
protocol for wireless sensor networks, in Proc.
of CISS 06, Princeton, NJ, March 2006. 7 M.
Zorzi and R. Rao, Geographic random forwarding
(geraf) for ad hoc and sensor networks multihop
performance, IEEE Transactions on Mobile
Computing, no. 4, December 2003.
7
RCS Overview

• Hop-based geographic grouping
• Light-weight localization for minimizing e2e
  latency
• Per-hop deadline based prioritized queuing
• Traffic classification for service
  differentiation
• Polling contention period based real-time MAC
• Decreasing average Inter Frame Space (IFS) for
  higher throughput
• Receiver contention based dynamic forwarding
• Lower control overhead on network dynamics
• Adaptive to duty cycle design

8
Hop-based Geographic Grouping

• Using limited broadcast
• The sink initializes a broadcast with groupID
  0.
• The sensor nodes assign their groupID received
  groupID 1.
• Any node rebroadcasts the grouping message with
  its own groupID once.
• The sensor nodes with lower group ID have smaller
  back-off window.
• Operate in pre-deployment stage
• Light-weight approach compared with precise
  localization

Sink
GroupID 1
GroupID 2
GroupID 3
GroupID 4
GroupID 5
9
Per-Hop Deadline Based Priority Assignment
Sink
tA 8, GroupID 3 Deadlineper-hop (15-8)/3, P
2
tA 5, GroupID 3 Deadlineper-hop (15-5)/3, P
3

• Example
• Deadlinee2e 15ms
• Deadlineminper-hop 1ms
• Priority level P 0 ? packet drop

tA 2, GroupID 4 Deadlineper-hop (15-2)/4 ,
P 3
tA 0, GroupID 5 Deadlineper-hop 15/5, P 3
10
Per-Hop Deadline Based Priority Queue
11
Polling Contention Period Based Real-Time MAC
802.11EDCA

• Less Average Inter-Frame Space (IFS)
• Smaller Back-off Window (BW)

12
Polling Contention Period Based Real-Time MAC

• Using Polling Contention Period instead of
  extended IFS in IEEE802.11 EDCA for prioritized
  MAC
• Priority Level PRTS, (PRTS 1, 2, , N)

WIN!
POLLRTS
13
Receiver Contention Based Dynamic Forwarding

• Dynamic Forwarding embedded in RTC/CTS exchange
• Upon receiving the RTS packet, only the node with
  highest priority can compete for transmitting the
  CTS packet
• Forwarding Metrics to determine the CTS Priority
  PCTS
• Local contention level
• Channel quality
• Queuing delay
• Forwarding distance

Average transmission time
Moving average of queue length
GroupID difference from the sender
14
Prioritized Real-time Packet Transmission
Operation
Arbitrary IFS
Slot
Slot
Short IFS
POLLRTS
RTS
WINCTS!
Data Pkt
WINRTS!
POLLCTS
15
Performance Evaluation

• RCS V.S. RAP, MMSpeed
• RAP 802.11EDCA Geographic Forwarding
• MMSpeed 802.11EDCA Table-based routing -
  Multicast
• Simulation settings in GlomoSim
• Each event lasts 300 s
• 10 simulations with different random seeds for
  each scenario
• Use the average value collected from all 10
  simulations for performance evaluation

16
Scenario I RCS V.S. RAP
Sink
Source 2 E2E DL 80ms

• Pure geographic forwarding cannot adapt to
  channel fading ? RCS provides much better
  end-to-end delay compared to RAP
• The channel quality along the route, instead of
  the priority level, dominates the end-to-end
  delay ? RAP fails to provide service
  differentiation

Source 1 E2E DL 40ms
17
Scenario II RCS V.S. RAP
Sink

• RCS
• 1
• P1RTS 1, P2RTS 3, P3RTS 5
• RAP
• P1RTS 1, P2RTS 3, P3RTS 4

Source 2 E2E DL 60
Source 1 E2E DL 30
Source 3 E2E DL 90
18
Scenario II Average E2E Delay

• GroupID provide accurate enough location
  information for dynamic forwarding with lower
  control overhead
• RCS provides better end-to-end delays for low
  priority traffic because of lower IFS with higher
  throughput.
• RCS provides better service differentiation
  capability with larger number of supported
  priority levels

19
Scenario II On-time Delivery Rate

• RCS provides better on-time delivery rate for all
  priority levels
• For high priority traffics by using better
  dropping policy, RCS achieves a little better
  performance even with longer IFS
• For low priority levels, with much smaller IFS
  and BW, RCS achieves better average throughput
  and results in 5-10 better on-time delivery
  rate.

20
Conclusions Future Work

• RCS Feature
• A novel design to provide service differentiated
  soft real-time guarantees for end-to-end
  communication in WSNs
• Requires minimum hardware support and adapts well
  to network dynamics
• Achieve lower end-to-end latency, better on-time
  delivery rate, finer service-differentiated
  granularity in unsynchronized WSNs, compared with
  RAP and MMSpeed.
• Future Work
• Extend to duty cycle design
• Implementation in WSN testbed