Eugenio Magistretti † Krishna Kant Chintalapudi * Božidar Radunović # Ramachandran Ramjee *
WiFi-Nano: Reclaiming WiFi Efficiency through 800 ns Slots
† Rice University * Microsoft Research India # Microsoft Research Cambridge
WiFi-Nano: Reclaiming WiFi Efficiency through 800 ns Slots Eugenio - - PowerPoint PPT Presentation
WiFi-Nano: Reclaiming WiFi Efficiency through 800 ns Slots Eugenio Magistretti Krishna Kant Chintalapudi * Bo idar Radunovi # Ramachandran Ramjee * Rice University * Microsoft Research India # Microsoft Research Cambridge Problem
Eugenio Magistretti † Krishna Kant Chintalapudi * Božidar Radunović # Ramachandran Ramjee *
† Rice University * Microsoft Research India # Microsoft Research Cambridge
Problem Overview
[Mbps]
~55 Mbps
§ 802.11 data-rates have increased from 1 Mbps
to 1 Gbps
§ Throughput performance has not seen a
commensurate increase
1997 802.11 1999 802.11a/b 2003 802.11g 2009 802.11n 2012 802.11ac/ad
Contribution
WiFi-Nano increases 802.11 throughput up to 100%
Why Throughput << Data-rate?
DIFS
Medium Access
101.5µs
Preamble
20µs 224µs
Data
SIFS
ACK
7 6 5 4 3 2 1 0
405.5 µs
44µs
54 Mbps
7 6 5 4 3 2 1 0
225.5 µs
300 Mbps
Medium Access
101.5µs
Preamble
32µs 40µs
Data
SIFS
ACK
36µs 7 6 5 4 3 2 1 0
221.5 µs
600 Mbps
Medium Access
101.5µs
Preamble
40µs 20µs
Data
SIFS
ACK
44µs DIFS DIFS
Why Throughput << Data-rate?
DIFS
Medium Access 101.5µs Preamble 20µs 224µs Data
SIFS
ACK
7 6 5 4 3 2 1 0
405.5 µs 60µs
54 Mbps
7 6 5 4 3 2 1 0
225.5 µs
300 Mbps
Medium Access 101.5µs Preamble 32µs 40µs Data
SIFS
ACK
52µs
7 6 5 4 3 2 1 0
221.5 µs
600 Mbps
Medium Access 101.5µs Preamble 40µs 20µs Data
SIFS
ACK
60µs
Overhead ~45%
Overhead ~82%
802.11 overhead does not scale with data-rate
Preambles 40% ACK 2%
Motivation
Single Link Case
Overhead Components
SIFS 16% Slot Duration 42% Communications Overhead Cannot be removed!
7 6 5 4 3 2 1 0
Medium Access
101.5µs
Preamble
40µs 20µs
Data
SIFS DIFS
Preamble+ACK
44µs
Motivation
Preambles 34% ACK 2%
Overhead Components
SIFS 14% Slot Duration 23%
Multiple Links Case (30)
Collisions 27% Collisions 27%
Objective § Reduce slot duration
q and reduce the occurrence of collisions q … while preserving fairness
§ Remove SIFS
Collisions 27% Collisions 27% Slot Duration 23% SIFS 14% Preambles 34% ACK 2%
Challenge: Slot Duration
§ 802.11a/n slot duration of 9 µs is close to the
minimum feasible
PropagaDon
4 ¡µs
A B 1 802.11 Slot
Clear ¡Channel Assessment ¡(CCA)
PropagaDon Clear ¡Channel Assessment ¡(CCA)
Rx/Tx ¡Hardware Turnaround
4 ¡µs
4 ¡µs
A 3 B 1
Key assumption:
Preamble detection and transmission are serial
Preamble detection and transmission occur in parallel
Speculative Preambles
Clear Channel Assessment may take multiple slots
The slot time can be reduced to a round-trip propagation delay
Slot Time (800 ns)
Preamble
Preamble
1 2 1 0
Preamble
3 2 1 0 Node ¡C Node ¡B Node ¡A
Speculative Preambles
1.
As soon as the backoff expires, a node transmits its preamble
2.
CCA: A node transmitting a preamble continues to attempt to detect incoming preambles
3.
A node aborts its transmission if it detects a preceding preamble
Slot Time (800 ns)
Medium Access time decreases from 101.5 µs to 7.6 µs
WiFi-Nano Design
§ Slot Time Duration § Collisions § Fairness § SIFS § Speculative Preambles
Objectives Techniques
§ Speculative ACKs § Counter Roll-back § Minimum Slot Size § Probabilistic Collision
Resolution
Implementation
§ Detect preambles and their starting time under
interference
§ Simultaneously transmit and detect preambles
Lattice Correlator
[PN1, PN2,PN3]
PN1 PN2 PN3
Δ
[PN1] [PN2] [PN3]
[PN1,PN2]
[PN2,PN3]
Δ Δ Δ Δ
QH Noise Canceller
Analog Radio Front-end
Rx Tx
Analog Self-Interference Canceller
§ Interference may require longer preambles
Results
§ Experiments
q Reliability of Preamble
Detection
q Efficiency Gain and Analysis q Fairness
Lyrtech Qualnet
Preamble Detection
Can nodes Can nodes reliably detect preambles reliably detect preambles despite self-interference? despite self-interference? Slightly longer preambles permit to maintain reliability
Efficiency
Efficiency f(data rate, #nodes) WiFi-Nano increases the throughput up to 100%
Efficiency
Efficiency f(data rate, slot time)
.11a/n .11b WiFi-Nano
WiFi-Nano increases the throughput up to 100%
How to Achieve More?
q Small packets (TPC ACKs) q Short flows (HTTP) q Delay sensitive applications (TPC, VoIP)
Frame Aggregation
§ Works only for single-link bulk downloads § Practically difficult to achieve
Frame Aggregation
At 18 kB, WiFi-Nano gains 25% over 802.11 at 600 Mbps
§ Practically difficult to achieve
§
Related work reports 18 kB as average aggregation
Summary
§ WiFi-Nano permits to
q Reduce the slot time to 800 ns q Reduce the occurrence of collisions to nearly 0 q Remove SIFS
802.11 overhead can be > 90%
WiFi-Nano increases 802.11 throughput up to 100%
Eugenio Magistretti † Krishna Kant Chintalapudi * Božidar Radunović # Ramachandran Ramjee *
† Rice University * Microsoft Research India # Microsoft Research Cambridge
WiFi-Nano: Reclaiming WiFi Efficiency through 800 ns Slots