Scaling Mesh for Real

Ed Knightly

ECE Department

Rice University

http://www.ece.rice.edu/~knightly

Ed Knightly

Scalable Mesh

High bandwidth – 400 Mb/sec to residences and small businesses

High availability– Nomadicity– Large-scale deployment– High reliability and resilience

Economic viability– $$/square mile

Ed Knightly

Research Challenges

1. Physical layer– 400 Mb/s

2. Media access– Target multi-hop and exploit PHY capabilities

3. Fairness and traffic control– Prevent starvation, remove spatial bias

4. Prototypes, Testbeds, and Measurement Studies– Platforms for experimentation and proof-of-concept

5. Architecture– Node placement, security, economics, etc.

Ed Knightly

Rice Transit Access Point (TAP) Platform

400 Mb/sec via 4x4 MIMO custom design– Single 20 MHz WiFi channel at 2.4 GHz and 20 bits/sec/Hz efficiency– Feedback-based algorithms for beam-forming MIMO

Custom MAC design and FPGA implementation

Ed Knightly

Rice Transit Access Point (TAP) Platform

400 Mb/sec via 4x4 MIMO custom design– Single 20 MHz WiFi channel at 2.4 GHz and 20 bits/sec/Hz efficiency– Feedback-based algorithms for beam-forming MIMO

Custom MAC design and FPGA implementation

Ed Knightly

Technology For All Deployment

Technology For All – Houston, Texas (non-profit) Empower low income communities through technology

– Neighborhood: income 1/3rd national average, 37% of children below poverty

Applications– Education and work-at-home

Ed Knightly

Technology For All Mesh Deployment

Multi-hop IEEE 802.11 wireless network covering 40,000 residents– Single wireline Internet backhaul– Long-haul directional links– OTS programmable platform– $25k/square mile

Ed Knightly

TFA Research Issues

Architecture– Node/wire placement

Sustainable non-profit business model

Protocol deployment– traffic management

Security

Measurement studies

Ed Knightly

Two Tier Architecture

Access: connects homes to mesh nodes Backhaul: connects mesh nodes to wires

Ed Knightly

Parking Lot Scenario

One branch of the access tree is shown

Parking lot is dominant traffic matrix

Ed Knightly

Parking Lot Measurements (FTP/TCP upload)

Single flow scenario widely studied Concurrent flows

– Without RTS/CTS, hidden terminals starvation– With RTS/CTS, multi-hop flows achieve 20% of 1-hop flows

Ed Knightly

Parking Lot Measurements (FTP/TCP bi-directional)

Near starvation with 3 or more hops– TCP unable to throttle short flows to leave capacity for long flows– MAC hidden terminals and Information Asymmetry [GSK05]

Ongoing work: – congestion control over an imperfect MAC– MAC redesign

Ed Knightly

Hidden Terminals in Access Networks

Ethernet

Ethernet

Ethernet

Ethernet

Internet

TAP1 TAP2 TAP3 TAP4

collision no collision

Ed Knightly

Information Asymmetry

Ethernet

Ethernet

Ethernet

Ethernet

Internet

TAP1 TAP2 TAP3 TAP4

RTS

TAP2 sets its NAVNo CTS

RTS

• Asymmetric view of channel state

• Node with more information knows when to contend; other attempts randomly

Ed Knightly

Result on Information Asymmetry [GSK05]

Analytical model to predict throughput

If randomly place nodes:– IA scenario is the most probable resulting in

severe throughput imbalance

– Previous studies in mobile settings missed by focusing on average throughput

Information Asymmetry is a fundamental property of wireless: state cannot be perfectly shared

Ed Knightly

Conclusions

Communications advances enabling 400 Mb/s links

At 3-4 hops, TCP/WiFi utilizes 1% of this

We can do better!

Challenges– MAC – multi-hop protocols– Fairness – distributed fairness algorithms– Prototypes – testbeds and proof-of-concept– Architecture – placement, economics, security, …