- A
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 1
- Seminar 2:
Multi-Channel, Multi-Radio Wireless Mesh Networks
Andreas J. Kassler kassler@ieee.org
Seminar 2: Multi-Channel, Multi-Radio Wireless Mesh Networks r e - - PowerPoint PPT Presentation
Seminar 2: Multi-Channel, Multi-Radio Wireless
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 1
Andreas J. Kassler kassler@ieee.org
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 2
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 3
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 4
– needs “wired” connectivity to access points. – Deployment slow and expensive
Node
Reach
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 5
Node Reachable! Node Reachable!
Every node is now Access Point AND Router Every node is now Access Point AND Router
– mesh routing backbone created by grid of wireless APs – Clients can associate with any access point. – Complete transparency: nodes forward voice, video and data traffic to and from nearby nodes wirelessly and ultimately to the internet
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 6
– Small number of wireless hops to gateway
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 7
– No Wires! – Properties:
management
– Empowers the individual and community
A WMN is dynamically self-organized and self-configured, with the nodes in the network automatically establishing and maintaining mesh connectivity among themselves A WMN is dynamically self-organized and self-configured, with the nodes in the network automatically establishing and maintaining mesh connectivity among themselves
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 8
– City of Chaska (8000 homes, 23.000 residents) 28% uptake after 2 years – Nomadic broadband service for $17.99 per month – Based on Tropos mesh products
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 9
which do not have wired networking infrastructure.
backbone network (Mesh)
economic development
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 10
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 11
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 12
no parallel transmission no parallel reception
Goal for MAC layer design:
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 13
– Stations have to equally compete for access to the medium – Acknowledgment scheme is used for error indication
Time Time Time Node 1 (transmit) Node 2 (receive) Node 3 (transmit) Busy Busy DIFS DIFS CW = 4 CW = 9 Data Frame Busy SIFS ACK DIFS DIFS CW = 11 CW=9-4=5 NAV
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 14
hidden nodes
– A mesh node is hidden for an ongoing transmission if it is not able to sense the
– A node not in the sensing range of the transmitter but within the interference range
– Throughput degradation – Unfairness
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 15
IEEE 802.11 RTS/CTS
– RTS: Request to send – CTS: Clear to send
Nodes receiving RTS or CTS might not get involved in new transmissions RTS/CTS
– solves hidden node problem – induces increased overhead and delay – also virtual carrier sensing
hidden nodes RTS CTS
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 16
Time Time Time sender receiver
Busy Busy DIFS DIFS CW = 4 CW = 9 RTS SIFS CTS NAV(RTS) SIFS DATA SIFS ACK DIFS DIFS DIFS Time
Busy DIFS CW = 13 NAV(CTS) DIFS sender receiver
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 17
receiving range interfering range sender receiver hidden node CTS
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 18
blocked nodes RTS CTS Parallel transmissions might occur (except for ACK)
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 19
– Generally hardware is not flexible enough – All you hear is your own signal
– wireless can’t do collision detect like Ethernet
n r
r
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 20
– 2400 – 2483.5 MHz (2.4 GHz), – 5150 – 5250 MHz (lower U NII), – 5250 – 5350 MHz (mid U NII), and – 5725 – 5825 MHz (upper U NII)
equipment, and incidental radiators
– 802.11{b,g} in the 2.4 GHz band; 802.11a in the U NII band – 802.15 WPAN (Bluetooth) in the 2.4 GHz band – 802.16 WirelessHUMAN in the mid and upper U NII bands
– Field disturbance sensors, cordless telephones, low power devices, and microwave ovens – Non 802.11 Part 15 devices: cordless telephones, A/V repeaters, security cameras, baby monitors, & digital data links
– Amateur radio in the 2.4 GHz and upper U NII bands; fixed microwave in the 2.4 GHz band; and satellite in the lower U NII band
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 21
Phone on
TCP download from a 802.11 AP Performance worsens when there are large number of short range radios in the vicinity
Panasonic 2.4GHz Spread Spectrum Phone 5 m and 1 wall from receiver
802.11 in presence of BT
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 22
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 23
– MRN need to relay traffic AND serve attached clients – In single radio WMNs, clients and MRNs operate on same channel
Internet
MRN1 MRN2 MRN3 MRN4
Single radio (e.g. 802.11b) for backhaul and client access
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 24
2 3 4 5 7 8 9 1 11 10
RTS RTS CTS CTS
6
2 packets in flight! Only 4 out of 11 nodes are active…. 2 packets in flight! Only 4 out of 11 nodes are active…. Backoff window doubles!
RTS RTS RTS
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 25
Key issues:
Cannot Tx and Rx in parallel (single radio) More problems due to collisions (hidden nodes) and interference Need to serialize reception and transmission Reduces capacity
Key issues:
Cannot Tx and Rx in parallel (single radio) More problems due to collisions (hidden nodes) and interference Need to serialize reception and transmission Reduces capacity Per MN Capacity=1/N , (N=hops) Per MN Capacity=1/N , (N=hops)
&
'% '( ') '
Single Radio Throughput (Best Case) 2 4 6 8 10 12 14 16 18 20 22 24 1 3 5 7 9 Hops Available Bandwidth (Mbps) 802.11b 802.11a
# *""#
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 26
– and , on " Radios
– BUT: Wireless Backhaul still shared All 802.11a MRNs operate on same channel
, Nico Bayer, Marcel Cavalcanti de Castro, Peter Dely, Andreas Kassler, Yevgeni Koucheryavy, Piotr Mitoraj and Dirk Staehle, in: Proceedings of the IEEE ICCSC 2008, Shanghai, China, May 26 28 2008.
VoIP background VoIP background
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 27
Key Idea:
Multi-radio, multi- channel Backhaul required for Carrier- Grade Send and receive in parallel on different channels Channel qualities and traffic demand vary
priori How to find the “best” channel for given link? How to coordinate which channel to use between what nodes at a given time?
*""
&
'% '& '- '(
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 28
Internet
MRN1 MRN2 MRN3 MRN4
Internet
MRN1 MRN2 MRN3 MRN4
Internet
MRN1 MRN2 MRN3 MRN4
(
(
Per MRN Capacity (different WMN types)
1 2 3 4 5 6 1 2 3 4 5 6 7 8 9 10 Number of MRNs (first one wired to internet) Per MRN Access Capacity (Mbps)
Single Radio Per-MRN Capacity (Mbps) Dual Radio Per-MRN Capacity (Mbps) Multi Radio Per-MRN Capacity - String (Mbps) Multi Radio Per-MRN Capacity - Loop (Mbps)
Internet
MRN1 MRN2 MRN3 MRN4
(
(
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 29
Today’s US Spectrum Map – 300 MHz to 30 GHz
/ # # " $ / # # " $
#
*"
0,
1 1
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 30
sufficient channels, interference can be completely eliminated.
$ "
becomes crucial and influences topology
#
"
!
& *"
!
% %
&
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 31
Channel Separation 5 2 1 Non-overlapping channels, A = 1, B = 6 Partially Overlapped Channels, A = 1, B = 3 Partially Overlapped Channels, A = 1, B = 2 Same channel, A = 1, B = 1 LEGEND 3 4 5 6 10 20 30 40 50 60 Distance (meters) UDP Throughput (Mbps)
3 435 + 6789
' ' ' '-
throughput, can observe minimal distance
Banerjee-SIGMETRICS-2006
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 32
– Manageability:
– Contention mitigation:
– Better performance via use of more spectrum
', ' "
1
c 1 m 1 m m m+1
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 33
– Use more channels shall be beneficial for dense networks – BUT: Cannot add infinite number of radios – What improvements are there when adding more channels? – How many radios are then needed?
– Available spectrum bandwidth – Connectivity constraints [Gupta Kumar]
– Interference constraints [Gupta Kumar]
;"$ #,3#0<= ;"$ #,3#0<=
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 34
– Interface constraint [Kyasanur Vaidya]
– Destination bottleneck constraint [Kyasanur Vaidya]
1
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 35
– captures link interference between pair of links, which
– Helps in
– Can use to model fractional interference and variable traffic
– packet transmission from nodes that are not “visible” – physical location of nodes within the network – whether or not multiple transmissions increase or decrease interference 1 2 6 4 5 3
1 - 4 1 - 2 2 - 3 4 - 5 2 - 5 3 - 6 5 - 6
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 36
1 2 6 4 5 3
Ch=1 Ch=6 Ch=11
1 2 6 4 5 3
1 - 4 1 - 2 2 - 3 4 - 5 2 - 5 3 - 6 5 - 6
[Subramanian08]
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 37
– Today: 4 ms with optimisations – Possible: 80 microsec
– RTS: Potential channels to be used – CTS: Receiver tells sender which channel to use
– How does the sender know which channel the receiver is listening on?
– Receive on all channels simultaneously costly – Use a dedicated control channel – Use a synchronized hopping protocol – Provide multiple rendezvous opportunities
#
*"
0,
?
1 1
?
#
*"
0,
?
1 1
?
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 38
Negotiating Channel with RTS / CTS Before starting, Node C sends RTS on Control Channel C1 to D including list of potential channels to be used. Node D replies with channel selected. Nodes tune to the selected channel
B C D E G H I A K J
RTS (C1,C3,C7) RTS (C3,C5,C7,C11) CTS (C11) CTS (C3)
F C2 C2 C3 C11 C1 10 nodes are active, 5 packets in flight, 150% improvement! 10 nodes are active, 5 packets in flight, 150% improvement!
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 39
Let C1 be the control channel C can hear traffic on C11 only, doesn’t hear the CTS from B consequently doesn’t know anything about traffic on C6 (D is too far to hear anything from B)
C1 C1 C11
RTS RTS Data on C6 Data on C6 Data on C6
Collision
CTS (C6) CTS (C6) Time
C6 C1 C6 C6 C6 C1
A B C D
Possible solution: Use multiple radios
So-MobiHoc-2004
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 40
B does not hear the RTS from A on C1. As node B has its interface on C11 it doesn’t hear the RTS from A. Also, A cannot sense the carrier as it is on different channel. A falsely concludes that B is not reachable.
A B C
C1 C11 C11
R T S Time R T S
Deafness Problem Deafness Problem
A B C D
Channel Deadlock Channel Deadlock
&
All nodes send RTS in circular fashion to
resolved but system capacity is degraded.
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 41
– number of channels typically greater than the number of radios
– but power is not the primary concern in most mesh networks
– but radios are fairly inexpensive – However, cannot add radios at will – How many cards give a good speedup at a reasonable cost?
– but switching speeds are being reduced
can use partially overlapping channels
mechanisms for concurrent transmissions
– Nodes can send and recieve in parallel using different Radios – Several links can operate in parallel at different nodes
#
*"
0,
?
1 1
?
#
*"
0,
?
1 1
? 1 2 6 4 5 3
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 42
– Example: MeshDynamics MD4000
1. Use standard 802.11 based hardware (BUT: need multiple interfaces). 2. Use 802.11, but customized hardware. 3. Develop minor extensions to 802.11 (AKA layer 2.5) 4. Design new MAC protocol.
– Interface can only use a given channel at a time – For two nodes to communicate they need to share acommon channel – Using multiple Radios, deafness, multi channel hidden terminal and channel deadlock problems can be mitigated – Channel re assignments might be required to improve capacity, minimize interference from external networks, etc – Network Partition Problems might arise
Network poorly connected
A B C D
1,3 2,4 1,2 3,4
A B C D
1,3 2,4 1,2 3,4 1,2 Some channels not used
A B C D 1,2
1,2 1,2 1,2
A B C D 1,2
1,2 1,2
A B C D 1,2
1,2 1,2
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 43
7
A B C D
5 10 15 20 25 30 35 64,64 60,64 56,64 52,64 Channels Throughput (Mb/sec) Netgear: A to B Hop Netgear: C to D Hop
0.00 5.00 10.00 15.00 20.00 25.00 30.00 35.00 40.00 Hop1 = A, Hop2 = G Hop1 = G, Hop2 = A
TCP Throughput (Mb/Sec) Hop 2 Hop 1
Same channel or channel separation of 4 causes 46% 49% reduction in overall throughput 802.11a link causes a 22% reduction in overall throughput, and a 63% reduction in throughput on the 802.11g link.
Interference is significant, RF hardware shielding work is beneficial Interference is significant, RF hardware shielding work is beneficial
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 44
– Connectivity, Spectrum Utilization, Load Awareness, External Interference?
– Routing determines traffic load on the virtual links – Need to consider channel, range, data rate diversity.
– Network Partition Problem – Channel Dependency Ripple Effect Channel Re assignment potentially needs Coordination – Topology Change Routing should be aware of Re assignment – Non Convergent behaviour during Channel Re assignment 0$ : $ 0$ : $
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 45
patterns and/or to reduce internal or external interference.
– Channel might change on large timescale according to traffic demand
patterns and/or reduce interference.
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 46
– A given interface fixed to a given channel
– E.g. C1 assigned to Radio 1, C2 to Radio 2, etc.
– All nodes use common set of channels used by Mesh Connectivity Layer [Draves04] or MUP
11 1 1 11 11 1 1 11 11 11 1
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 47
– Implemented at Link Layer – Exposes a single virtual MAC address – All nodes use common set of channels – Neighbor discovery and classification is done by ARP, channel selection (CS), and the MUP table – Broadcast over all interfaces – Determines channel quality to each neighbor using periodic probing over all channels – Switch to a new (better channel) based on SRTT measurements after time interval
Adya 2004
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 48
– Optimizes locally use of spectrum – Does work with standard hardware – No need for global topology info. – Can use standard protocols – Reduce delay by 40 50%
– Estimating channel quality can be hard – Need to couple with routing and routing metrics to utilize end to end
1 MUP Enabled
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 49
– [Marina 05]Treat Channel Assignment as Topology Control Problem, – use conflict graph to model interference – Assign Channels to minimize maximum conflict weight – [DAS05] Use ILP to maximize # concurrent transmissions given connectivity constraints – [Tang05] Statically bind interface channels by minimizing interference among links
3%
60 52 2 1 64 60 Not possible 52 56 52 60
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 50
– Re Assign channels at slow time scales – External Interference Aware, Centralized [Ramachandran06] – Load Awareness – Centralized [Raniwala04] – Distributed [Raniwala05]
3%
60 52 2 1 64 60 Not possible 52 56 52 60
2 radios / node, 4 channels
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 51
– External interference can severely degrade performance
data and control traffic
interfered to most interfered
channel assignment server (CAS)
– Internal interference between mesh links should be avoided Assign orthogonal channels using Conflict Graph Concept
estimates for all channels supported by the router’s radios – Channel sensing and assignment can break network connectivity
– Uses Multi Radio Conflict Graph (MCG) to model interference between mesh links
[Ramachandran06]
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 52
– CAS ranks channels according to their interference levels – With the links adjacent to the gateway as the starting points, the algorithm traverses vertices in MCG in BFS order – Each vertex is assigned the currently highest ranked channel that is not assigned to its adjacent vertices in MCG. – Considering the MCG constraint is to reduce the interference among the mesh nodes – Considering the channel rankings of the external interference level is to reduce the interference between the mesh and the external wireless networks – If such a channel is not available, then randomly assign a channel to this vertex.
– Interference can change rapidly, Depends on existence of traffic patterns to determine interference can lead to incorrect channel assignment – Assumes most traffic towards gateway
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 53
10 50 40 10 50 40
,=
,=
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 54
– Link load determined by routing algorithms Joint channel assignment routing? – Having complete information about network topology and traffic matrix (how?), the traffic aware channel assignment problem is NP hard
– Form multiple spanning trees, each one rooted at a Gateway – Each node joins one (or more) gateways – Routing Metrics: Gateway link capacity, Path capacity – Re assign channels to balance traffic load across channels – Coordinate with direct neighbors
Raniwala05
# $@ *# !"@ $≥ " 2 =
" 5
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 55
– Each interface is used to communicate with multiple neighbors. – Each node balances its children associations across DOWN NICs.
– Each node periodically exchanges its channel utilization with neighbors – constructs a neighborhood channel usage map – Periodically re assigns DOWN NICs’ channels to balance traffic load across – Adapt channel assignment to traffic patterns
– A weighted combination of aggregated traffic load and number of nodes using the channel
[Raniwala05]
Start with single channel routing for initial load estimation Iterate over channel assignment and routing until convergence Factor 6 to 7 throughput improvement Start with single channel routing for initial load estimation Iterate over channel assignment and routing until convergence Factor 6 to 7 throughput improvement
Ch=36 Ch=36
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 56
– Interface can switch channel when needed [So MobiHoc 2004 , Bahl04]
( 11 1 1 11 11 1 1 11 11 11 1Backup
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 57
– One common control channel (e.g. On radio 1), many data channels (switchable, e.g. on radio 2) – Control channel used to negotiate, which data channel to use – Advantages:
– Disadvantages
Control Channel: 1 Control Channel: 1 Data Channel: 2 4 Data Channel: 2 4
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 58
Each node has at least 1 fixed, 1 switch able interface Connectivity is maintained, all channels used Every node picks a channel as it’s fixed channel Different nodes use different fixed channels Once a “connection” is made, there may not be a reason to switch channels again for that particular flow Per Channel Packet Queue
A
Fixed (ch 1) Switchable
A
Fixed (ch 1) Switchable
B
Fixed (ch 2) Switchable
B
Fixed (ch 2) Switchable
C
Fixed (ch 3) Switchable
C
Fixed (ch 3) Switchable
1 2 3 2
B D C
A
B D C
A
Packet to D Packet to C
Packet to C arrives buffer packet Interface switches to channel 3
[Kyasanur06]
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 59
Multi channel broadcast support, Scheduling for channel switching Hybrid Multichannel Control Protocol (HMCP)
On startup each node picks a random fixed channel Periodically send a “hello” pkt. containing fixed channel & 1 hop neighbors info.
Maintain a NeighborTable containing fixed channels being used by neighbors Select the channel with fewest nodes as a candidate Not traffic aware! Change fixed channel to candidate channel probabilistically to avoid oscillations
A
Fixed (ch 1) Switchable
A
Fixed (ch 1) Switchable
B
Fixed (ch 2) Switchable
B
Fixed (ch 2) Switchable
C
Fixed (ch 3) Switchable
C
Fixed (ch 3) Switchable
1 2 3 2
B D C
A
B D C
A
[Kyasanur06]
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 60
– Select routes that have channel diversity WCETT
– Switching interfaces results in packets being queued and delayed – If a node is on more routes, might require more switching – Try to minimize the amount of switching while maximizing channel diversity
4 +
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 61
Intuition: balance switching overhead with channel diversity For each channel (i): measure InterfaceUsage(i), average over 1s interval Measure probability that switchable interface is on different channel i != j when packet arrives on channel j Switching Cost Path Metric
≠
=
j i s
i sage InterfaceU j p ) ( ) ( elay switchingD j p j SC
s
* ) ( ) ( =
( )
+ + − =
≤ ≤ =
c j j n i i i
X c SC ETT MCR
1 1
max ) ( ) 1 ( β β
ci = channel used on i th hop
2 1 2 1 3
4 +
[Kyasanur06]
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 62
2 4 6 8 10 12 14 16 1 2 3 4 5 6 7 8 9 10 Normalized throughput Topology number (2,2) (2,5) (5,5) (2,12) (12,12)
(m,c)
50 nodes, 50 flows, 500m x 500m area
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 63
– Based on Net X, Linux 2.6, Cambria Platform (Gateworks) – Three 802.11a radios per mesh node (m = 2), Legacy clients with 1 radio 802.11b/g – Nagios Network Management Platform
http://www.cs.kau.se/cs/prtp/pmwiki/pm wiki.php?n=Resources.MeshTestbed http://www.cs.kau.se/cs/prtp/pmwiki/pm wiki.php?n=Resources.MeshTestbed
Multi Channel Routing, Hybrid Channel Assignment
Interface and Channel Abstraction Layer, Aggregation IP Stack QoS Interface Device Driver User Applications ARP QoS Interface Device Driver
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 64
1-
Route A B C in use D needs route to F Route D E F better
– Routing larger time scale
loss rate, etc...
– Brown route ”better” than green
– Can use separate broadcast channel needs additional radio
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 65
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 66
interference?
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 67
single radio, dual radio, multi radio mesh deployments (slide 63), under the assumption that available capacity is 5 Mbps (for 802.11b, single radio) and 23 Mbps for 802.11a backhaul for dual radio and multi radio. Assume interference range = 2 * transmission range. Also assume a string topology where the MRN providing internet access is located at one end of the string and all clients transmit data towards nodes in the internet. In Dual and Multi Radio deployments, assume that 802.11b radio is used for client access and 802.11a for backhaul.
Wireless Networks”, Pradeep Kyasanur and Nitin H. Vaidya, in SIGMOBILE Mobile Computing and Communications Review, Volume 10, Number 1, pages 31 43, January 2006, discuss the following two questions:
– Authors compare their approach against the case where One Channel is used. Do you think this comparison is fair? Provide arguments for each potential answer. If you think it is not fair, against what solution should it be compared to make a more fair comparison? – For the chain topology Evaluation results (Figure 5), why the throughput does not increase linearly with the number of channels used? Discuss several reasons.
Jorge Crichigno, Min You Wub, Wei Shu, In: Ad Hoc Networks 6 (2008) 1051–1077, discuss the following questions:
– Explain the Channel Dependency Problem and the Ripple Effect. – For the Receiver Fixed Hybrid Channel Assignment strategy (4.2.4), is the Ripple Effect an Issue? What about the Non convergent behavior problem?
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 68
considered harmful. , Proceedings of the joint international conference on Measurement and modeling of computer systems, Saint Malo, France 2006
Wireless Mesh Networks, IEEE Transactions on Mobile Computing (TMC), Vol 7. Number 11. November 2008.
2004.
International Conference on Broadband Networks (Broadnets 2005), Boston, Massachusetts – USA, October 2005.
radios, in: 2nd IEEE Communications Society Conference on Sensor and Ad Hoc Communications and Networks (SECON), Santa Clara, California – USA, September 2005.
International Symposium on Mobile Ad Hoc Networking and Computing (Mobihoc 2005), Urbana Champaigne, Illinois – USA, 2005.
networks, in: 25th Conference on Computer Communications (Infocom 2006), Barcelona – Spain, April 2006.
networks, Mobile Computing and Communications Review 8 (2) (2004) 50–65.
Conference on Computer Communications (Infocom 2005), Miami, Florida – USA, March 2005.
networks, in: 10th ACM International Conference on Mobile Computing and Networking (MobiCom 2004), Philadelphia, Pennsylvania – USA, 2004
10th International Conference on Computer Communications and Networks (ICCCN 2001), Scottsdale, Arizona – USA, 2001.
Mobile Computing and Communications Review 10 (1) (2006) 31–43.
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 69
n d r e a s J . K a s s l e r Topics in Computer Networks 2009 • Seminar 3 70
Andreas J. Kassler kassler@ieee.org