Performance issues

1
Performance issues improvement on 802.11 MAC

• review of 802.11 MAC
• performance issues
• improvements
• idle sense
• an overlay approach
• more

2
An Overlay MAC Layer for 802.11 Networks

• Ananth Rao Ion Stoica
• UC Berkeley
• Mobisys 2005

3
Motivation

• 802.11 hardware provides initial ease of
  deployability for many applications
• mesh networks
• long haul links
• large Infrastructure Networks
• these apps stretching 802.11 beyond its design
  goals (Wireless LANs)

Internet Gateway
4
Problem 1 different data rates
Data Rate Throughput
Case I A 11 Mbps 3.09 Mbps
Case I B 11 Mbps 3.36 Mbps
Case II A 11 Mbps 0.76 Mbps
Case II B 1 Mbps 0.76 Mbps
R
B
A
5
Problem 2 unpredictability
2
1
3
4
5
6
Problem 3 forwarding on behalf of others
Forwarders get same share as others
Ethernet
1/2
1/6
1/6
1/6
This problem cannot be solved by local scheduling
or queue management algorithms like WFQ
7
Approaches

• workarounds in routing/transport layer
• easy to deploy
• cannot address some issues
• change/replace MAC
• new protocols, new standard
• more powerful, hard to deploy
• overlay MAC layer (OML)
• directly on top of 802.11 MAC
• no need to change hardware
• directly use interfaces exposed by 802.11 cards

8
Advantage of an overlay approach

• easy to deploy
• easy to modify
• implemented in software
• easy to modify for diverse requirements
• tighter integration between MAC and upper layers
• performance benefits
• utilize information from higher layers

9
Bigger picture overlay network
10
Bigger picture overlay network
Focus at the application level
11
Even bigger picture virtualization

• Virtualization of resources powerful abstraction
  in systems engineering
• computing examples virtual memory, virtual
  devices
• virtual machines e.g., java
• IBM VM os from 1960s/70s
• Networking examples
• connecting local heterogeneous networks
• IP over ATM
• overlay networks
• VPN

12
Overlay MAC Layer (OML) design goals

• efficient
• fair or differentiated allocation
• flexible and low cost
• avoid modifying MAC

13
Overlay MAC Layer (OML) what can it control?

• no control in upper layer
• cannot decide when getting a packet
• no control in MAC
• cannot decide when packet is actually sent
• can control only when to send packet to network
  card
• packet scheduling policy FIFO, strict priority
  scheduling, weighted fair queuing

14
Detour strict priority scheduling

• transmit highest priority queued packet
• multiple classes, with different priorities
• class may depend on marking or other header info,
  e.g. IP source/dest, port numbers, etc..
• real world example reservations versus walk-ins

arrivals
time
packet service
time
departures
15
Detour weighted fair queuing

• each class gets weighted amount of service in
  each cycle
• equal weight Round Robin scheduling

16
OWL main idea use TDMA-like schedule

• divide time into slots
• weighted slot allocation (WSA) allocate slots to
  nodes according to weighted fair queuing policy
• assigns a weight to each node
• allocate slots proportion to nodes weights -gt
  weighted allocation
• a slot is only assigned to one node in an
  interference region -gt reduce packet loss

17
Questions

• clock synchronization?
• slot length?
• interference region?
• weighted slot allocation
• how to choose weight?
• decide a winner w/o communication?

18
Clock synchronization slot size

• loose time synchronization
• leader-based
• estimate one way delay
• slot size transmit 10 packets of maximum size
• larger than clock synchronization error
• larger than packet transmission time
• as small as possible

19
Interference region

• ideally node i applies WSA to all nodes that
  interfere with i
• how to determine who interfere with me?
• assume a node can interfere with all nodes within
  k-hop distance
• only an approximation, not accurate
• how to determine interference relationship is an
  active research

20
Weighted Slot Allocation decide winner w/o
communication

• each node uses pseudo-random function to generate
  a random number
• Hi H(ni, t) 1/wi
• t time slot, wi weight of node ni
• can generate random number for all nodes in the
  collision domain (e.g., 2-hop neighborhood)
• the highest number wins

21
Evaluation methodology

• Simulation in Qualnet
• Implementation in Atheros Madwifi driver Click
  router

22
Summary of results

• Overhead OML thruput comparable to native 802.11
• reduced contention and retransmissions
• Fairness Fairness index for OML network much
  higher
• a nodes weight flows passing thru it
• Limitations Impact of mobility Interference
  from native 802.11 clients

23
Simulation results

• Similar throughput to 802.11
• Control overhead is small

24
Simulation results (cont.)

• Improved fairness over standard 802.11

25
Summary

• overlay approach
• coarse-grained scheduling on top of 802.11
• less contention
• provide flexibility of assigning priorities to
  senders
• enables experiment with new scheduling and
  bandwidth management algorithms