Flush: A Reliable Bulk Transport Protocol for Multihop Wireless Networks

1
Flush A Reliable Bulk Transport Protocol for
Multihop Wireless Networks

• Sukun Kim, Rodrigo Fonseca, Prabal Dutta,
  Arsalan Tavakoli, David Culler, Philip Levis,
  Scott Shenker, and Ion Stoica

University of California at Berkeley
International Computer Science Institute
Samsung Electronics
Stanford University




SenSys 2007
2
Motivating Example
Sausalito (north)
SF (south)
500 ft
1125 ft
4200 ft
246 ft
56 nodes
8 nodes
Structural Health Monitoring of the Golden Gate
Bridge

• All data from all nodes are needed
• As quickly as possible
• Collecting data from one node at a time is
  completely acceptable
• Over 46 hop network !

3
Introduction

• Target applications
• Structural health monitoring, volcanic activity
  monitoring, bulk data collection
• One flow at a time
• Remove inter-path interference
• Easier to optimize for intra-path interference
• Built on top of MAC layer
• No merging with MAC layer for easy porting

4
Table of Contents

• Introduction
• Algorithm
• Implementation
• Evaluation
• Discussion
• Related Work
• Conclusion

5
Flush Algorithm Overview

• Receiver-initiated
• Selective-NACK
• Hop-by-hop Rate Control
• Empirically Measure Interference Range

6
Rate Control
8
6
5
4
8
6
4
3
7
5
Interferer
6
5
4
1 / Rate Packet Interval d8 d7 d6 d5
dX Packet transmission time at node X
7
Interference Range gt Reception Range
However,
Jammer
Vulnerable to Jammer
No problem to Jammer
Signal Strength
Noise Floor
Noise Floor SNR Threshold
Noise Floor 2 X SNR Threshold
SNR Threshold minimum SNR to decode a packet
Jammer a node which can conflict with the
transmission, but cannot be heard
8
Identifying the Interference Set
CDF of the difference between the received signal
strength from a predecessor and the local noise
floor A large fraction of interferers are
detectable and avoidable
Fraction of Links
9
Implementation Control Information

• Control information is snooped
• d packet transmission time, 1 byte
• f sum of ds of interfering nodes, 1 byte
• D Packet Interval 1 / Rate, 1 byte
• d, f, D are put into packet header, and exchanged
  through snooping

10
Implementation Rate-limited Queue

• 16-deep Rate-limited Queue
• Enforces departure delay D(i)
• When a node becomes congested (depth 5), it
  doubles the delay advertised to its descendants
• But continues to drain its own queue with the
  smaller delay until it is no longer congested

11
Table of Contents

• Introduction
• Algorithm
• Implementation
• Evaluation
• Discussion
• Related Work
• Conclusion

100 MicaZ nodes Purple nodes Diameter of 67
hops Mirage Testbed in Intel Research Berkeley
Sink
12
Packet Throughput of Different Fixed Rates
Effective Throughput (pkt/s)
Packet throughput of fixed rate streams over
different hop counts No fixed rate is always
optimal
13
Packet Throughput of Flush
Overall Throughput (pkt/s)
Effective packet throughput of Flush compared to
the best fixed rate at each hop Flush tracks the
best fixed packet rate
14
Bandwidth of Flush
Overall Bandwidth (B/s)
Effective bandwidth of Flush compared to the best
fixed rate at each hop Flushs protocol overhead
reduces the effective data rate
15
Fraction of Data Transferred in Different Phases

• Fraction of data transferred from the 6th hop
  during the transfer phase and acknowledgment
  phase
• Greedy best-effort routing is unreliable, and
  exhibits a loss rate of 43.5. A higher than
  sustainable rate leads to a high loss rate

16
Amount of Time Spent in Different Phases

• Fraction of time spent in different stages
• A retransmission during the acknowledgment phase
  is expensive, and leads to a poor throughput

17
Packet Throughput at Transfer Phase
Transfer Phase Throughput (pkt/s)
Effective goodput during the transfer
phase Flush provides comparable goodput at a
lower loss rate which reduces the time spent in
the expensive acknowledgment phase, which
increases the effective bandwidth
18
Packet Rate over Time for a Source
Flush-e2e has no in-network rate control

• Source is 7 hops away, Data is smoothed by
  averaging 16 values
• Flush approximates the best fixed rate with the
  least variance

19
Maximum Queue Occupancy across All Nodes for Each
Packet

• Flush exhibits more stable queue occupancies
  than Flush-e2e

20
Sending Rate at Lossy Link
6
3
4
5
2
1
0
Packets were dropped from hop 3 to hop 2 with 50
probability between 7 and 17 seconds
Both Flush and Flush-e2e adapt while the fixed
rate overflows its queue
21
Queue Length at Lossy Link
Flush and Flush-e2e adapt while the fixed rate
overflows its queue
22
Route Change Experiment
4
5

• We started a transfer over a 5 hop path
• Approximately 21 seconds into the experiment
  forced the node 4 hops from the sink to switch
  its next hop
• Node 4s next hop is changed, changing all nodes
  in the subpath to the root
• No packets were lost, and Flush adapted quickly
  to the change

3a
3b
2a
1a
2b
1b
0
23
Scalability Test
Effective bandwidth from the real-world outdoor
scalability test where 79 nodes formed 48 hop
network with 35B payload size Flush closely
tracks or exceeds the best possible fixed rate
across at all hop distances that we tested
Overall Throughput (B/s)
24
Table of Contents

• Introduction
• Algorithm
• Implementation
• Evaluation
• Discussion
• Related Work
• Conclusion

25
Discussion

• High-power node
• reduces hop count and interference
• Not an option on many structural health
  monitoring due to power and space problems
• Interactions with Routing
• Link estimator of a routing layer breaks down
  under heavy traffic

26
Related Work

• Li et al capacity of a chain of nodes limited
  by interference using 802.11
• ATP, W-TCP rate-based transmission in the
  Internet
• Wisden concurrent transmission without a
  mechanism for a congestion control
• Fetch single flow, selective-NACK, no mention
  about rate control

27
Conclusion

• Rate-based flow control
• Directly measure intra-path interference at each
  hop
• Control rate based on interference information
• Light-weight solution reduces complexity
• Overhearing is used to measure interference and
  to exchange information
• Two rules to determine a rate
• At each node, Flush attempts to send as fast as
  possible without causing interference at the next
  hop along the flow
• A nodes sending rate cannot exceed the sending
  rate of its successor
• In combination, Flush provides as good bandwidth
  as fixed rate, and also gives a better
  adaptability

28
Questions
29
(No Transcript)
30
Reliability
Source
Sink
2, 4, 5
2
4
5
4, 9
4, 9
4
9
31
Rate Control Conceptual Model
Assuming disk model N Number of nodes, I
Interference range
Rate
32
Rate Control
1. At each node, Flush attempts to send as fast
as possible without causing interference at the
next hop along the flow 2. A nodes sending rate
cannot exceed the sending rate of its successor
8
6
5
4
d8 d8 H7 d8 d7 f7 d8 d7 d6 d5
8
7
6
5
4
3
33
Implementation

• RSSI is measured by snooping
• Information is also snooped
• d, f, D are put into packet header, and exchanged
  through snooping
• d, f, D take 1 byte each, 3 bytes total
• Cutoff
• A node i considers a successor node (i- j) an
  interferer of node i1 if, for any j gt 1,
  rssi(i1) - rssi(i-j) lt 10 dBm
• The threshold of 10 dBm was chosen after
  empirically evaluating a range of values

34
Implementation

• 16-deep Rate-limited Queue
• Enforces departure delay D(i)
• When a node becomes congested (depth 5), it
  doubles the delay advertised to its descendants
• But continues to drain its own queue with the
  smaller delay until it is no longer congested
• Protocol Overhead
• Out of 22B provided by Routing layer, 2B sequence
  number 3B control field 17B payload

35
Test Methodology

• Mirage testbed in Intel Research Berkeley,
  consists of 100 MicaZ
• -11 dBm
• Diameter of 67 hops
• Average of 4 runs

36
Bottom line performance

• High-power node
• reduces hop count and interference
• Not an option on the Golden Gate Bridge due to
  power and space problems
• Interactions with Routing
• Link estimator of a routing layer breaks down
  under heavy traffic
• Bottom line performance???

37
Average Number of Transmissions per node for
sending 1,000 packets
38
Bandwidth at Transfer Phase
Transfer Phase Throughput (B/s)
Effective goodput during the transfer
phase Effective goodput is computed as the
number of unique packets received over the
duration of the transfer phase
39
Details of Queue Length for Flush-e2e
40
Memory and Code Footprint
41
(No Transcript)
42
(No Transcript)
43
6
3
4
5
2
1
0
44
4
5
3a
3b
2a
1a
2b
1b
0
45
Motivating Example

• Every data from every node is needed
• Partial data has no or little value
• Should work over 46 hops