doc.: ieee 802.11-05-0367-00-0000s submission may 2005 jan kruys,ciscoslide 1 mesh networking...
DESCRIPTION
doc.: IEEE s Submission May 2005 Jan Kruys,CiscoSlide 3 Purpose Show some of the main factors that determine mesh performanceTRANSCRIPT
May 2005
Jan Kruys,CiscoSlide 1
doc.: IEEE 802.11-05-0367-00-0000s
Submission
Mesh Networking
Performance Considerations
+31 20357 2447
May 2005
Jan Kruys,CiscoSlide 2
doc.: IEEE 802.11-05-0367-00-0000s
Submission
Notice: This document has been prepared to assist IEEE 802.11. It is offered as a basis for discussion and is not binding on the contributing individual(s) or organization(s). The material in this document is subject to change in form and content after further study. The contributor(s) reserve(s) the right to add, amend or withdraw material contained herein.
Release: The contributor grants a free, irrevocable license to the IEEE to incorporate material contained in this contribution, and any modifications thereof, in the creation of an IEEE Standards publication; to copyright in the IEEE’s name any IEEE Standards publication even though it may include portions of this contribution; and at the IEEE’s sole discretion to permit others to reproduce in whole or in part the resulting IEEE Standards publication. The contributor also acknowledges and accepts that this contribution may be made public by IEEE 802.11.
Patent Policy and Procedures: The contributor is familiar with the IEEE 802 Patent Policy and Procedures <http:// ieee802.org/guides/bylaws/sb-bylaws.pdf>, including the statement "IEEE standards may include the known use of patent(s), including patent applications, provided the IEEE receives assurance from the patent holder or applicant with respect to patents essential for compliance with both mandatory and optional portions of the standard." Early disclosure to the Working Group of patent information that might be relevant to the standard is essential to reduce the possibility for delays in the development process and increase the likelihood that the draft publication will be approved for publication. Please notify the Chair <[email protected]> as early as possible, in written or electronic form, if patented technology (or technology under patent application) might be incorporated into a draft standard being developed within the IEEE 802.11 Working Group. If you have questions, contact the IEEE Patent Committee Administrator at <[email protected]>.
AuthorsName Address Compan
yPhone Email
Jan Kruys Haalerbergweg 10, CH1101Netherlands
Cisco Systems Int’l
+ 31 348 453719
Mesh Networking – Performance Considerations
May 2005
Jan Kruys,CiscoSlide 3
doc.: IEEE 802.11-05-0367-00-0000s
Submission
Purpose• Show some of the main factors
that determine mesh performance
May 2005
Jan Kruys,CiscoSlide 4
doc.: IEEE 802.11-05-0367-00-0000s
Submission
A Simple Example
main mesh links
back-up connections
Mesh link interference area for 12 Mb/s (Pl exp = 3.3)
AP coverage
Feed point
May 2005
Jan Kruys,CiscoSlide 5
doc.: IEEE 802.11-05-0367-00-0000s
Submission
Observations• Traffic aggregation near the fibre pop:
– requires high throughput – see next slide• The interference area shown is for 12Mb/s links and
omni antennas– all mesh links share the capacity of ONE RF
channel is this area• because the defer threshold must be set low enough to assure
sufficient SIR• Possible starvation of further away nodes
– their packets have to content for channel access many times• every contention stage decreases throughput exponentially• does not play a role at low load levels (reduced contention)
• This is closely related to the fairness issue brought up by Violeta in the March meeting
May 2005
Jan Kruys,CiscoSlide 6
doc.: IEEE 802.11-05-0367-00-0000s
Submission
Traffic aggregation (topology effect)• The tree type
mesh is a good model for a group of nodes supported by a wired root node – multiple trees
build a 3600 coverage.
• Link level loads increase rapidly towards the wired root node– even a small fan-
out at each hop causes exponential traffic growth at the root link
Required Mesh Link capacity
0.0
50.0
100.0
150.0
200.0
250.0
1 2 3 4 5 6
Hops from the edge
Rela
tive
load
Serial (fan-out=1)
fan-out=2
fan-out=2.2
Required Mesh link Capacity
0.0
10.0
20.0
30.0
40.0
50.0
1 2 3 4
Hops from the edge
Rela
tive
Load
Series1
Series2
Series3
fan-out=1fan-out=1.7fan-out=2.2
May 2005
Jan Kruys,CiscoSlide 7
doc.: IEEE 802.11-05-0367-00-0000s
Submission
Starvation of the edge (single RF channel)
• This graph is for a 5 hop deep mesh with a fan-out of 1.6– for a total of 26 APs– a packet from the edge has to contend 4 x with 25 others to get
through to the root = 2.10^-6• this effect becomes noticeable at medium to high traffic loads
Relative Medium access probability as function of nr of hops
0.0
20.0
40.0
60.0
80.0
100.0
120.0
1 2 3 4 5 6
nr of hops
dB o
f pro
babi
lity
low fan out
high fan out
May 2005
Jan Kruys,CiscoSlide 8
doc.: IEEE 802.11-05-0367-00-0000s
Submission
Down/up ratio limitations (topology effect)
• For the same data rate, delivering packets to the next hop down requires progressively less time – Higher fan-out allows higher down/up ratio– Lower traffic loads also allow higher down/up ratio
1.5
2.5 4
1.00
0.400%
20%40%60%80%
100%
link load
fan-out ratio
Maximum Down traffic % as function of fan-out and link load
1.00
0.80
0.60
0.50
0.40
0.30
0.20
May 2005
Jan Kruys,CiscoSlide 9
doc.: IEEE 802.11-05-0367-00-0000s
Submission
Taking stock• “Deep” and/or “wide” meshes have a problem with
traffic aggregation near the root– the effect is exponential: ~1/fan-out-1)*(fan-
out^hops)• “Deep” meshes have a problem with down/up traffic
ratio• RF channel access contention causes starvation near
the edge – even if the mesh is narrow– the effect is exponential: (1/nodes)^hops for a single
RF channel• At low offered loads, none of this may be very clearly
noticeable
May 2005
Jan Kruys,CiscoSlide 10
doc.: IEEE 802.11-05-0367-00-0000s
Submission
Benefits of Sectorization
Reduces the aggregation effect – in proportion to # of
sectorsReduces the starvation effect
– by reducing the contending population
Reduces the re-use exclusion area (for the same EIRP)– improves spectral
efficiency
Feedpoint
600 sector
May 2005
Jan Kruys,CiscoSlide 11
doc.: IEEE 802.11-05-0367-00-0000s
Submission
Multiple RF channels per node
Requires two backhaul radios per nodeFrequency planning reduces/eliminates re-use and starvation issuesAllows optimization of data rate wrt SIRRetains sharing behaviour of CSMA/CAAggregation and down/up limitation still applyFeed
point
600 sector
May 2005
Jan Kruys,CiscoSlide 12
doc.: IEEE 802.11-05-0367-00-0000s
Submission
A second look at starvation• Starvation is caused by repeated
contention for resources– keeping the contending population low reduces
starvation• one way: spread the contenders over multiple
RF channels: e.g. (1/nodes)^hops becomes (1/fan-out)^hops
• second way: operate the link at the lowest possible SIR = lowest possible data rate– throughput requirements drop with distance from the root– the effect is less in a narrow mesh but it still has benefits
r.e. sharing with adjacent networks– keeping the number of hops down is highly
effective• conversely, sharing the RF channel, omni
fashion, is asking for trouble
May 2005
Jan Kruys,CiscoSlide 13
doc.: IEEE 802.11-05-0367-00-0000s
Submission
What about small meshes?• Small: fan-out < 3, hops <2, <9 devices, one
root• Performance with one RF channel:
– aggregation: < 2^3 or a factor 8 max.• means “no two HDTV streams on the second hop from the
DSL feed”– down/up ratio is limited to 60% or so at full
load• same impact as aggregation
– starvation for a 6 node network: 1/6^2 = 1/36• hurts most on the second hop nodes: no HDTV viewing and
slow uploads.– probably ok for the home and SMB
• Larger meshes have major performance issues– notably when operated at a single RF channel
May 2005
Jan Kruys,CiscoSlide 14
doc.: IEEE 802.11-05-0367-00-0000s
Submission
Summary• Mesh network performance depends strongly on three
factors:– mesh depth and width (nr of hops/nr of links per node)
• drives aggregation and starvation effects– traffic pattern (down/up ratio)
• limits the mesh depth– RF channel
• limits data rates and increases starvation• can be improved with fire brigade or multi-radio designs
– these factors further complicate the QoS problem• The optimum configuration depends on specific
requirements: – underloaded networks can support a “large” mesh depth or
width– high performance networks require a low fan-out and
shallow mesh• Bottom line: there is no “best” solution - TGs standard
should be flexible enough to support a wide variety of mesh solutions