Optimization Problems in Wireless Coding Networks

1
Optimization Problems in Wireless Coding Networks

• Alex Sprintson
• Computer Engineering Group
• Department of Electrical and Computer Engineering

2
Agenda

• An overview of the COPE
• A new forwarding architecture for wireless
  networks
• by Katti et. al. Sigcomm 2007
• Optimization problems in wireless coding networks
• Joint work with Salim El Rouayheb and Asad
  Chaudhry

3
COPE

• New Architecture For Wireless Networks
• Substantially improves throughput
• Inserts a coding layers between MAC and IP
• Identifies coding opportunities
• Benefits from them by mixing packets

4
Background

• Bob and Alice

Relay
Requires 4 transmissions
5
Background

• Bob and Alice

6
Broadcast medium

• Each packet is broadcasted, for free, in a small
  neighborhood around its path

7
Coding opportunity

• Each node stores the overheard packets for a
  short time
• Each node tells its neighbors which packets it
  has heard by annotating the packets it sends

8
Opportunistic coding

• When a node transmits a packet it uses its
  knowledge of what its neighbors have heard

9
COPE Overview

• Opportunistic Listening
• COPE sets all nodes in promiscuous mode
• Nodes snoop on all communications over the
  wireless medium
• Store the overheard packets for a limited period
  of time T (default T0.5 s).

10
COPE Overview

• Each node broadcasts reception reports to tell
  its neighbors what packets are stored
• Reception reports are sent by annotating the data
  packets the node transmits
• A node that has no data packets to transmit
  periodically sends the reception reports in
  special control packets

11
Opportunistic coding

• The key question is what packets to code together
  to maximize throughput
• The node should maximize the number of native
  packets delivered in a single transmission, while
  ensuring that each intended nexthop has enough
  information to decode its native packet
• Use a simple heuristic
• We analyze this problem in the second part of the
  talk

12
Learning Neighbor State

• How does a node know what packets its neighbors
  have?
• By using reception reports
• Problems
• At times of sever congestion, reception reports
  may get lost in collisions
• At time of light traffic they can arrive too late
• A node cannot relay on reception reports and may
  need whether a neighbor has a particular packet
• May need to guess whether a neighbor has a
  particular packet

13
Intelligent guessing

• Leveraging the routing protocols
• Routing protocols compute the delivery
  probability between every pair of nodes
• This information needed for identifying good
  paths
• Occasionally, a node can make an incorrect guess

14
Coding gain

• Defined as a ratio between the number of packets
  required by the current non-coding approach to
  the minimum number of transmissions used by COPE
• The gain is equal to 1.33 in the Bob and Alice
  experiment

15
CodingMAC gain

• Experiments show that the throughput improvement
  sometimes greatly exceeded the coding gain
• The interactions between coding and MAC produces
  a beneficial side effect
• For example, in the Bob and Alice example, the
  MAC divides the bandwidth equally between the
  three users
• However, the router needs to transmit twice as
  many packets
• Halve the packets transmitted by the edge nodes
  are dropped at the router queue

16
CodingMAC gain

• COPE allows the router to XOR packets and drop
  them twice as fast
• Coding and MAC gain is higher for many topologies

17
Practical considerations

• Never delay a packet
• Transmit when a wireless channel is available
• If no coding opportunities exists, do not wait
  for the arrival of a new packet.
• Let the node to opportunistically overload each
  transmission with additional information

18
Practical considerations (cont.)

• Dealing with packets of different lengths
• Give preference to XORing packets of the same
  length
• Coding small packets with large packets reduces
  bandwidth savings
• Empirical studies show that the packet
  distribution in the Internet is bimodal with
  picks at 40 and 1500 bytes
• Dealing with packets of two different sizes

19
Practical considerations (cont.)

• Internet packet size distribution

20
Practical considerations (cont.)

• Packet reordering
• We would like to limit reordering packets from
  the same flow
• Because TCP considers it as a congestion signal
• Consider packets according to their order in the
  queue
• Reordering can arise because of the need to
  retransmit packets which were lost in quessing
• COPE has a module that puts TCP packets in order
  before delivering them to the transport layer

21
Pseudo-broadcast

• 802.11 has two modes broadcast and unicast
• Broadcast mode cannot be used because of poor
  reliability and lack of backoff
• Unicast mode retransmission of the packet until
  a synchronous ack is received
• Solution unicast packets which are intended for
  broadcast
• COPE require each nexthop node to ack the packet

22
Packet format

• Packet ID- a 32-bit hash of the packet's source
  IP address and IP sequence number

23
Performance

• Large Scale Experiment
• 20 nodes
• 2 floors
• Pick sender and receiver randomly
• Packet size based on actual measurement
• Flow arrival are Possion

24
Large Scale Exp.
Offered load in Mb/s
Scare Coding Opp. At Low Demands Demand Up /
Congestion Up / Gain Up
25
Findings

• Network coding does have practical benefits, and
  can substantially improve wireless throughput.
• When the wireless medium is congested and the
  traffic consists of many random UDP flows, COPE
  increases the throughput by 3-4x.
• Hidden terminals create a high collision rate
  that cannot be masked even with the maximum
  number of 802.11 retransmissions. In these
  environments, TCP does not send enough to utilize
  the medium, and thus does not create coding
  opportunities.