Title: CS 15-849E: Wireless Networks (Spring 2006) MAC Layer Discussion Leads: Abhijit Deshmukh Sai Vinayak Instructor: Srinivasan Seshan
1CS 15-849E Wireless Networks (Spring 2006) MAC
LayerDiscussion Leads Abhijit Deshmukh
Sai VinayakInstructor Srinivasan Seshan
2Papers
- An Energy-Efficient MAC Protocol for Wireless
Sensor Networks - Wei Ye, John Heidemann, Deborah Estrin
- The Case for Heterogenous Wireless MACs
- Chung-cheng Chen, Haiyun Luo
- Design and Evaluation of a new MAC Protocol for
- Long-Distance 802.11 Mesh Networks
- Wei Ye, John Heidemann, Deborah Estrin
3Outline
- Motivation
- MAC Wireless Sensor Networks
- Heterogenous Wireless MACs
- MAC for Mesh Networks
- Take Aways
- Similarities and Differences
- Q A
4Motivation
- Last Lecture
- MACAW, Carrier Sense, Idle Sense
- Basic Terms, Algorithms
- Major Focus on Fairness
- Very Generic
- Special Requirements for
- Sensor Networks
- Heterogeneous
- Mesh Networks
5MAC for Sensor Networks
- Sensor Networks
- Sensors, Embedded processor, Radio, Battery
- Ad hoc fashion
- Proximity, short-range multi-hop communication
- Committed to One or few applications
- MAC Protocol
- Energy Efficiency
- Scalability
- Accommodate network changes
- Fairness, Latency, Throughput and Bandwidth
6Sensor Networks
- Sources of Energy Waste ?
- Collision
- Overhearing
- Control packet overhead
- Idle Listening
- Tradeoff of fixing these
- Reduction in per-hop fairness and latency. How?
- Message Passing, Fragment long message
- Why not a big concern in Sensor Networks?
- Application-level performance
7- Energy Saving turn off radio . Issues?
- Latency
- In-network processing. Power consumption?
8Related Work
- PAMAS
- Avoid overhearing among neighbors
- Two independent radio channels
- Suffers from idle listening
- TDMA
- Natural Savings
- Scheduling
- Static
- Piconet
- Periodic Sleep
9Sensor-MAC Protocol Design
- Periodic Listen and Sleep
- Message Passing
- Collision and Overhearing Avoidance
10Periodic Listen and Sleep
- Basic Scheme
- Turn off Radio, set timer to wake up, sleep
- Clock Drift
- Sync using relative timestamps
- Long listen period
- Reduce Control Overhead
- Sync with neighbors, exchange schedules
- Advantage over TDMA ?
- Looser Synchronization
- Disadvantage?
- Latency due to switching, RTS/CTS
11Periodic Listen and Sleep
- Choosing and Maintaining Schedules
- Schedule Table
- Synchronizer
- Follower
Rebroadcast
SYNC
Listen
Wait (random)
Wait (random)
12Periodic Listen and Sleep
- Maintaining Synchronization
- SYNC packet
- Listen Interval
- SYNC RTS
13Collision Overhearing Avoidance
- Collision Avoidance
- NAV
- Virtual vs. Physical Carrier Sense
- Overhearing Avoidance
- Listening to all transmissions
- Who all should sleep?
- All neighbors of sender and receiver
x
x
E
C
A
B
D
F
14Message Passing
- Long vs. Short Message Length
- Stream of Fragments, single RTS-CTS
- Problem?
- No Fairness
- 802.11 Methodology?
- Why send ACK after each fragment?
- Prevent hidden terminal problem
15Implementation
- Rene Motes Tiny OS
- Simplified IEEE 802.11
- Message Passing (overhearing avoidance)
- S-MAC (Message Passing Periodic Sleep)
- Topology used
16Results
- Low performance for high loads?
- Synchronization overhead (SYNC packets)
- Latency
17Heterogeneous Wireless MACs
- Basic Service Set (BSS)
- Careful Channel Assignment
- Wireless interference
- Limited orthogonal channels
18Motivation
- Exposed Receiver Hidden Sender
CTS / RTS ?
data
data
x
Blocked
S1 ? R1 ?
ACK
194-way Handshake?
20Incomplete vs. Inconsistent
- Channel status at sender
- Incomplete estimate of receiver
- Inconsistent at multiple competing senders
- Incomplete channel status high packet loss
- Inconsistent channel status unfair channel
sharing
21Intra-BSS Interference Mitigation
- When to use 4-way handshake?
- Client detecting data transmission vs.
- Clients data transmission being detected
- Access point to initiate channel access?
- BSS in center
- Less chance of interference from other BSS
22Inter-BSS Interference Mitigation
- RTR (Request to receive)
- RTR-DATA vs. RTS-CTS-DATA
- ACK in form of next RTR
- Stateless Approach
- Alternating between MAC protocols
- Simple Design and Implementation
- Low Channel Utilization
23Fairness
- Why is flow 2?3 getting unfair treatment?
- Client 3 is exposed receiver
- Receiver 1 is not interfered by 2?3
- How to solve it ?
- Switch to receiver initiated protocol
- Increase power levels of CTS/RTS
24MAC for Long Dist. 802.11 Mesh
- Motivation
- Extend 802.11 for long haul
- Challenges
- Use off-the shelf hardware
- Low cost
25Overview
- Basic Principle
- SynRx SynTx
26Design and Implementation
- Design decisions driven by
- Low cost considerations
- Usage of off-the-shelf 802.11 hardware
- Achieving SynOp
- Get rid of immediate ACKs
- Get rid of carrier sense backoffs
27Design and Implementation (contd.)
- Immediate Acks
- Use IBSS mode of operation
- Convert IP unicast to MAC broadcast
- No ACKs for broadcast packets in IBSS mode
- Broadcast Unicast since link is 1-1
- ACKs can be implemented at the driver level
- Carrier Sensed Backoffs
- Make use of feature provided by Intersil Prism
chipsets
282P Operation on Single Link
- Marker acts as a token
- Loose Synchrony
292P Operation on Single Link (contd.)
- Need to handle 2 scenarios
- Temporary loss of synchrony (loss of marker)
- Link recovery after failure
- 2P handles both using timeouts
- Advantages
- Link-resync process is quick
- CRC errors do not cause timeout (other than
marker) . Why ?
302P Operation on Single Link (contd.)
- Two ends of a link get out of synchrony at the
same time and timeout together . So? - They would not hear each others marker packets
since both SynTx coincides So? - Repeated Timeouts !!! Solution ?
- Staggered timeouts ? Bumping
31Topology Formation
- What are the topologies in which 2P?
- Bipartite ?
- A tree is trivially bipartite
- Bad in terms of fault tolerance
- Add redundancy but turn on only one tree at a
time (Morphing) - 3 Heuristics
- Reduce length of links used
- Avoid short angles between links
- Reduce hop-count
32Evaluation
- Goal is threefold
- Measure impact of step by step link establishment
- Study effect of 2P in a large topology
- Study performance of TCP over 2P
- Link Establishment
- 12.9 ms for first case (delay due to bumping)
- 4.9 afterwards
33Throughput
342P vs TCP
35Similarities and Differences
- Similarities
- MAC protocol implementations
- Extend 802.11 for a specific environment
- Others?
- Differences
- Deployment scenarios
- Energy Saving, Long haul, Heterogeneity
- Writing Style
- Others?
36