Interview talk at various universities and labs

1
Pushing the Limits of Wireless Networks
Dina Katabi
2
In the near future,

• Communication computation will go mobile
• Convergence of voice, image, video, and data
• Run on any wireless infrastructure (Cellular,
  WiFi, WiMax, )

3
But, wireless still struggles with low throughput
This future relies on wireless networks

• Need a solution!

Talk in the WiFi context, but everything applies
to any technology with omni-directional antenna
4
Current Approach
5
Current Approach
6
Current Approach
Router
Bob

• Requires 4 transmissions
• Can we do it in 3 transmissions?

7
Our Approach
Router
Bob

• 3 transmissions instead of 4
• Increase Throughput
• Reduce Power

8
Two Departures

• Accept wireless as a broadcast medium
• Dispose of the point-to-point abstraction
• Network Coding
• Routers mix bits in packets then forward them
• Our work bridges theory with practical and
  addresses multiple unicast flows

9
COPE Coding Opportunistically
10
COPE (1)

• Opportunistic Listening
• Exploit wireless broadcast
• Every node snoops on all packets
• A node stores all heard packets for a limited
  time

11
COPE (1)

• Opportunistic Listening
• Exploit wireless broadcast
• Every node snoops on all packets
• A node stores all heard packets for a limited
  time

A
C
B
12
COPE (1)

• Opportunistic Listening
• Exploit wireless broadcast
• Every node snoops on all packets
• A node stores all heard packets for a limited
  time

• Node sends Reception Reports to tell its
  neighbors what packets it heard
• Reports are annotations on packets
• If no packets to send, periodically send reports

13
COPE (2)

• Opportunistic Coding
• To send packet p to neighbor A, XOR p with
  packets already known to A
• Thus, A can decode
• But how can multiple neighbors benefit from a
  single transmission?

14
Efficient Coding
A
D
C
B

• Arrows show next-hop

15
Efficient Coding
Bad Coding C will get RED pkt but A cant get
BLUE pkt
A
D
Cant decode!
C
B
Decoded!
16
Efficient Coding
Better Coding Both A C get a packet
A
D
C
Decoded!
B
Decoded!
17
Efficient Coding
Best Coding A, B, and C, each gets a packet

XOR
XOR
A
Decoded!
D
C
B
Decoded!
Decoded!
To XOR n packets, each next-hop should have the
n-1 packets encoded with the packet it wants
18
But how does a node know what packets a neighbor
has?

• Reception Reports
• But reception reports may get lost or arrive too
  late
• Make informed guesses based on the delivery rate
  between the two nodes
• Yes, error might occur and we recover by encoding
  and retransmitting

19
COPE s Characteristics

• COPE inserts a coding layer between the IP and
  MAC layers
• Does not delay packets
• Works with 802.11, 802.16, TDMA, etc.

20
Performance
21

• We implemented COPE in Linux

22
Arvind-and-Bob Experiment
Router
Bob
3 transmissions instead of 4 ? 25 throughput
increase
23
Results of Arvind-and-Bob Experiment
Ratio of Throughput with COPE to Current Approach
COPE almost doubles the throughput
24
Why More than 25?
Router
Bob
COPE alleviates the mismatch between MACs
capacity allocation and the congestion at a node

• 802.11 is fair ? 1/3 capacity for each node
• Without COPE, router needs to send twice as
  much as Arvind or Bob ? Router drops packets
• With COPE, all nodes need equal rate

25
Large-Scale Experiments

• Wireless Testbed
• 34 nodes
• 3 floors
• Experiments
• Pick sender and receiver randomly
• UDP flows
• Transfer size based on actual measurements

26
Testbed (one out of 3 floors)
27
COPE vs. Current Approach
Network Throughput (Mb/s)
Current
Total Demands (Mb/s)
28
COPE vs. Current Approach
Network Throughput (Mb/s)
COPE
Current
COPE provides a large throughput increase
Total Demands (Mb/s)
29
Conclusion

• COPE a new approach to wireless
• Broadcast instead of point-to-point links
• Network coding
• Large throughput increase that can reach several
  folds
• Simple and practical
• Our ongoing research examines theoretical bounds
  and more topologies and transport protocols