doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 1 Proactive Mesh Network Framework Date: Authors: Notice: This document has been prepared to assist IEEE 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 Patent Policy and Procedures: The contributor is familiar with the IEEE 802 Patent Policy and Procedures, 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 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 Working Group. If you have questions, contact the IEEE Patent Committee Administrator at. NameCompanyAddressPhone Bing ZhangNational Institute of Information and Communications Technology 3-5 Hikaridai, Seika-cho, Soraku-gun, Kyoto, Japan Oyunchimeg ShagdarATR Adaptive Communication Research Laboratories Hikaridai, Seika-cho, Soraku-gun, Kyoto, Japan Suhua TangATR Adaptive Communication Research Laboratories Hikaridai, Seika-cho, Soraku-gun, Kyoto, Japan Youiti KadoOki Electric Industry Co., Ltd Honmachi, Chuo-ku, Osaka, Japan Masanori NozakiOki Electric Industry Co., Ltd Honmachi, Chuo-ku, Osaka, Japan

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 2 Outline Layer-2 proactive routing protocol –Topology discovery –WDS unicast / broadcast / multicast –Routing enhancement with different metrics Strong link selection with RSSI Pseudo flow-based load balancing QoS support –Delay based priority control Multiple-radio support –Autonomous channel assignment Security –Secure link establishment over mesh networks Conclusion

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 3 Proactive Routing Protocol for Mesh Networks

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 4 Two kinds of routing protocols: Reactive routing protocol Low overhead Proactive routing protocol Transmission starts quickly Adaptation to the topology variation Features of Mesh Networks Mesh APs with low mobility Moderate scale (max: 32 MPs) MPs can have multi-interface Proactive routing is beneficial Why proactive routing protocol?

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 5 Architecture - Network Model - Router Server Mesh Portal Mesh AP Legacy STA ・・・ a(5.2GHz) ・・・ b/g(2.4GHz) ・・・ a(5.2GHz) Connecting Mesh Points based on proactive protocol Efficiently accommodating Legacy STA associated with Mesh AP MP: Mesh Point Legacy STA MP

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 6 Local Topology Discovery - Hello Message Exchange - Periodically exchanging Hello message –Advertising 1-hop neighbor information Upon receiving the Hello message from neighbors –Checking the link status with 1-hop neighbors –Acquiring the 2-hop neighbor information MP-A adds MP list from MP-B that is not included in its 1-hop MP list, into 2-hop MP list. MP-A DA MAC:FF SA MAC:A FCDu DA MAC:FF SA MAC:B FCDu Hello frame from MP-A Hello frame from MP-B MP-B FCS 1-hop Neighbor MAC Address list FCS 1-hop Neighbor MAC Address list MP-B adds MP list from MP-A that is not included in its 1-hop MP list, into 2-hop MP list. MC

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 7 Local Topology Discovery - MPR Selection - MP determines the OLSR-based MPR (Multipoint Relay) set from its 1-hop neighbors. MP advertises its MPR information in the periodic HELLO messages. MP-A 1-hop neighbor MPR of MP-A 2-hop neighbor

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 8 Topology Discovery - ASAT & TC Messages - Mesh AP generates an ASAT (Associated Station Address Table) message which records a MAC address table of legacy STA associated with it. MP generates a TC (Topology Control) message which includes the links to all MPs of its MPR selector set. ASAT Leagacy STAs ASAT Leagacy STAs Mesh AP-AMesh AP-C MP-B TC

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 9 Topology Discovery - ASAT & TC Message Forwarding - MPR forwards the ASAT & TC messages from its MPR selectors. Flooding ASAT & TC messages in the network to construct the routes to all of MPs and legacy STAs FCS MP-C FCS MP-A broadcasts the TC frames (1) MP-B retransmitted the TC frame by replacing SA with MAC:B (2) advertised link set advertised link set TC frame from MP-A TC frame from MP-B MP-B (MPR) FCS ASAT SN, MAC:A MP-C DA MAC:FF SA MAC:B FCDu DA MAC:FF SA MAC:B FCDuFCS ASAT SN, MAC:A ASAT frame from Mesh AP-A ASAT frame from Mesh MP-B Mesh AP-A broadcasts the ASAT frames (1) MP-B retransmitted the ASAT frame by replacing SA with MAC:B MP-A DA MAC:FF SA MAC:A FCDu Mesh AP-A DA MAC:FF SA MAC:A FCDu (2) MP-B (MPR) MC

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 10 Example of Unicast Routing 1.STA1 sends a data frame of infra-mode to Mesh-AP1. 2.Mesh-AP1 –learn STA2’s association to Mesh-AP3 from its ASAT. –transform STA1’s data frame into a WDS frame. –send the WDS frame to Mesh-AP2. (RA : Mesh-AP2, TA : Mesh-AP1, SA : STA1, DA : STA2) 3.Mesh-AP2 –learn STA2’s association to Mesh-AP3 from its ASAT. –forward the WDS frame to Mesh-AP3.( RA : Mesh-AP3, TA : Mesh-AP2, SA : STA1, DA : STA2) 4.Mesh-AP3 –de-transform the WDS frame into the data frame of infra-mode. –forward the data frame to STA2. STA1 Mesh-AP1Mesh-AP2Mesh-AP3 STA2 (1) (2)(3) (4)

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 11 Example of Broadcast Routing STA1 Mesh-AP1Mesh-AP2Mesh-AP3 STA4 (1) (2)(3) (4) STA2STA3 (2)(4) 1.STA1 sends a data frame to Mesh-AP1. ( DA : broadcast address) 2.Mesh-AP1 – transform STA1’s data frame into a WDS frame. – broadcast the WDS frame to Mesh-APs. (RA : broadcast address, TA : Mesh-AP1, SA : STA1, DA : broadcast address) – broadcast the data frame to the associated STAs. 3.MPR Mesh-APs broadcast the WDS frame. (RA : broadcast address, TA : own address, SA : STA1, DA : broadcast address) 4.Mesh-APs –de-transform the WDS frame into the original data frame. –broadcast to the associated STAs.

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 12 (3)(3) (4)(4) (1) Multicast Routing - Admission control - –Multicast group leader generates a member list for each multicast address. –Multicast member lists are broadcasted. –Participating in multicast group MP sends a “Join Request” frame to the group leader. (1) Group leader notifies its decision by a “Join Reply” frame. (2) –Leaving multicast group MP sends a “Depart Request” frame to the group leader. (3) Group leader notifies its decision by a “Depart Reply”. (4) A L B DA MAC:MA SA MAC:A FCDuFCS TA MAC:A DATAMC MA: Multicast address (1) (2)(2)

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 13 Routing Enhancement with Different Metrics

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 14 Distribution of Link Metrics Types of Routing Metric: –Traffic load over a link –RSSI –Transmission delay by the link Message SizeTime To LiveMetric type=HOPCOUNT Advertised Neighbor Main Address1 Advertised Neighbor Main Address2 Message SizeTime To Live Metric type Advertised Neighbor Main Address1Metric1 Advertised Neighbor Main Address2Metric2 Original TC message TC message with various specified metrics

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 15 Selection of Strong Links Metric corresponding to RSSI. MPs know the RSSI (Received Signal Strength Indicator) from the Hello frame. In addition to the hop count, RSSI is adopted to calculate the metric. Selection of the strong link Example:

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 16 Load Balancing - Problem Specification - Why load balancing using multiple paths is required –MPs usually use only a path to each destination. –Traffics to the same destination run down into a single path. –It is desirable to balance the load using multiple paths p2 p1 89 heavy Path Index DARA p173 p274 Routing Table with Multiple Paths

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 17 Load Balancing - Related Schemes - Existing Scheduling Schemes –MPs randomly select an RA (= next-hop MP) for each packet destined to the same DA Path Index DARA p173 p274 Routing Table with Multiple Paths Merit –Easy to implement. Demerit –It causes out-of-order frame delivery.

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 18 Load Balancing - Pseudo Flow Based Scheme - Pseudo flow : –Type 1 : Identified by DA and SA –Type 2 : Identified by DA and TA (= previous-hop MP) MPs generate a pseudo flow table. MPs determine next-hop MP for each pseudo flow based on the current load condition f2 f1 f2 89 Flow Index DASARA f1783 f2794 Pseudo Flow Table [type 1] Flow Index DATARA f1713 f2724 Pseudo Flow Table [type 2] Merits –Utilizing WDS header Info. –A pseudo flow’s path is stable. –In-order frame delivery. Demerit –Cross grained.

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 19 QoS Support for Multimedia Applications

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 20 Many applications such as Voice are delay sensitive. Motivation Flow1 Flow2 Flow3 Flow Flow1 has larger number of hops than Flow2 does. Delay (Flow1) ≥Delay(Flow2) Flow3 relayed by a congested MP (MP-5). Delay (Flow3) ≥Delay(Flow4) Some flows might not be able to meet their delay requirements while other flows reach their destinations with time to spare!

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 21 End-to-end delay is highly dependant on –number of hops –congestion condition at its relay MPs These QoS issues will be left unsolved even IEEE e technique is applied to mesh networks. Proposed approaches –Prioritizing mechanism based on number of hops –Prioritizing mechanism based on congestion condition at relay MPs –Prioritizing mechanism based on time stamp Delay based Priority Control Schemes

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 22 Prioritizing Mechanism based on Number of Hops Larger number of hops Longer end-to-end delay. MPs give priority control among flows based on number of hops they traverse. Flow1 Flow Frame2: Flow2 Frame1: Flow1 Routing Table at MP-2 FlowTTL Flow11 Flow22 The number of hops for Frame2 is larger than one of Frame1 for both from itself to the destination and end-to-end hops. Frame2 is sent first with a higher priority

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 23 Priority Control Scheme based on Congestion Condition at Relay MPs Congestion condition of MP is measured by the mean queue lengths of its buffers. Each MP broadcasts its congestion condition periodically. When a MP receives queue length information from its neighbour MP, it updates the information by adding own queue length to the previous value. Each MP knows the cumulative queue length each path. Frame on a path with longer cumulative queue length is sent first.

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 24 The cumulative queue length for frame 2 is larger than one of frame 1 at MP-2. Frame2 is sent first with a higher priority Queue length [Kbyte] 30 Queue length [Kbyte] 50 Flow1 Flow DestinationCumulative queue length [Kbyte] Frame2: Flow2 Frame1: Flow1 Priority Control Scheme based on Congestion Condition at Relay MPs (cont’) Routing Table at MP-2

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 25 Priority Control Scheme based on Time-stamp Source MP originates a request frame and sends to the destination. Request frame has a delay field. Each relay MP updates the delay field  delay += sojourn time in its buffer Destination MP it creates a report frame  copy the delay field of the request frame to the reply frame Relay MPs learn the end-to-end delay from the reply frame. At relay nodes, a frame with the longest delay characteristic is sent first.

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 26 Reply: Dst:1, Delay: 70 msec Request: Dst:4, Delay: 20 msec DestinationExpected delay [msec] 140 Reply: Dst:1, Delay: 70 msec Flow1 Flow Frame2:Dst:4 Frame1: Dst: 1 Request: Dst:4, Delay: 20 msec Request: Dst:4, Delay: 60 msec Request: Dst:4, Delay: 70 msec DestinationExpected delay [msec] The expected delay for frame 2 is larger than one of frame 1. Frame2 is sent first with a higher priority Priority Control Scheme based on Time-stamp (cont’)

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 27 Multiple-radio Support

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 28 IF0 IF1 C B D A Autonomous Channel Assignment (CA) Premise –Multiple channels are available –Hybrid equipment of single or multiple interfaces/radios Proposal –Mergence of isolated network by sharing common channel –Distribute traffic over all channels by CA Basic principle for CA –Least used channel first Features of proposed scheme –Auto configuration of the mesh network –Minimizing co-channel interference –Pairwise channel assignment Avoiding hidden terminal / deafness problem Do not require quick channel switching

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 29 CA Model One channel is used as common channel Common channel is statically configured or dynamically selected Each MP –monitors the common channel. –knows how many interfaces it has. –detects whether its neighbor has multiple interfaces (by beacon). Any pair of MPs with multiple interfaces –distributes traffic over different channels. MAP1 MP1 MAP2MAP3MAP4 MP2 MAP5MAP MP With two interfaces With single interface Ch-1 Ch-2 Ch-3

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 30 Determination of Common Channel – dynamic common channel adjustment – MPs may start at different time with different channels. Different channels may cause the mesh network isolated. Mesh Portal, or a MP (ex. MP with biggest ID), manages common channel. Common channel is managed by periodic beacon. MP8MP7MP9MP3MP4MP10MP2 MP1 MP6MP5MP8MP7MP9MP3MP4 MP10 MP2 MP1 MP6MP5 Mesh Portal

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 31 Determination of Common Channel (cont’) A MP increases CAS before transmitting the beacon if –it is GW (it sets GW_CAS in CAS_FLAG) –otherwise if it is the temporary manager (it sets TEMP_CAS in CAS_FLAG) When a MP receives a beacon, it updates CAS information if –it is not the manager of CAS If the new common channel is different from before, the MP –notify its neighbors by triggering a beacon over the old common channel with the new common channel information –switch to the new common channel A MP, without a route to GW, periodically scans all the channels. If it receives a beacon with GW_CAS set, it notifies its neighbor & switches to new common channel CASCAS Flag CAS manager New common channel Multi-channel capability New beacon element

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 32 Pairwise CA Only for a link, both ends support multi-channel capability –learn multi-channel capability by exchanging the beacons. –record the multi-channel capability in the link entry. Channel usage collection –An MP learns the load of the channels it is using. –Each MP exchanges the channel load information. A MP learns the average channel load of other channels. MAP1 MP1 MAP2 MAP3MAP4 MP2 MAP5 MAP2 Load of Ch-1, Ch-2 Load of Ch-3 Direct knowledge: Indirect knowledge: Ch-1 Ch-2 Ch-3

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 33 Two Kinds of CA Global CA –Initiated by Mesh Portal, from Mesh Portal to leaf MPs. –GlobalCA message controls the CA operation. –In each step, an MP assigns channels for all its links. –Each MP notifies its downstream MPs to continue Global CA. Local CA –Only a single step, initiated by either end of a link. –Locally adjust the channel based on channel load. Global CA Local CA

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 34 A Step in Global CA If an MP does not support multi-channel, –it attributes all the neighbors to the common channel group. If an MP supports multi-channel, –Single interface neighbors: Common channel group. –Multiple interface neighbors: fresh CAS -> CA is already finished by the upstream MPs. Otherwise -> Channel selection and Per-link CA operation. MP1MP2MP3 MP4 MP5MP6 MP7 MP MP with multiple interfaces MP with single interface Data channel 2 Data channel 1 Common channel

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 35 Per-link CA Operation 1.Select a channel and send CAReq 3-1If receiving a CARep with Success –Change the channel 3-2If Failure and a suggested channel –If the new channel is usable, try the new channel by sending new CAReq 3-3If Failure and no suggestion –Stick to the old channel 2-1If specified channel is acceptable –Send a CARep with Success –Change the channel 2-2Otherwise, if it has a different channel usable –Send a CARep with Failure and contain a suggested channel 2-3Otherwise –Send a CARep with Failure without suggestion InitiatorResponder (on receiving a CAReq) Global CA Local CA

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 36 Channel Selection & Post CA Operation If a data interface with older CAS exists –Select a new channel different from the already assigned channels for the other interfaces and assign it for the interface. Otherwise, all the interfaces are used up –Select a channel among the already assigned ones. Copy the assigned channel to the link entry. Copy the assigned channel to the interface entry. Update the association between the link and the interface. Channel selection for the link Post CA operation – if CA for the link is really successful

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide Simulation Model ◆ ns-2 simulator ◆ Transmission rate : 54 Mbps ◆ Transmission range : 85 m ◆ Traffic type : Constant Bit Rate (CBR) ◆ Routing : OLSR ◆ Layer2 : IEEE802.11a ◆ Codec : G.711 ◆ Payload size : 400 Bytes ◆ Interval : 50 ms Common channel 0 Data channel 1 Data channel 2 Data channel 3 GW

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 38 Fixed Phone Delivery ratio :97 % Delay :100 ms Mobile Phone Delivery ratio :94% Delay :150 ms Class A Class B

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 39 Security - Authentication and Link Establishment Mechanisms over Mesh Networks-

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 40 MP-5MP-3MP-4MP-1 (1) Beacon (2) Association (3) Authentication ＋ Key Derivation AS (Authentication Server) (1) The newly arrived MP (MP-5) receives beacon frames from neighbor MPs. (2) MP-5 associates to an MP (MP-2). (3)- MP-2 forwards an authentication request from MP-5 to AS. - If MP-5 is accepted to be connected, AS creates a key to MP-5. Authentication of Newly Arrived MP MP-2 Secure link between MP-5 and MP-2 is established.

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 41 YES NO YES Link Establishment between Neighboring MPs MP-5MP-3MP-4MP-1 (1) Beacon (4) Authentication, PTK creation AS (Authentication Server) MP-2 (2) Authenticator selection (5) PTK creation (3) Connection An MP receives a beacon frame from non-connected neighbor MP Have a route to AS ？ Act as supplicantAct as authenticator Authenticator selection (Ex: comparing number of hops to AS) Selected as authenticator ？ Flow chart of authenticator selection

doc.: IEEE /0386r3 Submission Bing Zhang, et al. July 2005 Slide 42 Conclusions A proactive routing protocol –Accommodating the legacy STA associated with Mesh AP –Optimizing the amount of link-state information –Supporting unicast and broadcast Enhancement of the proposed OLSR-based routing –Selecting the strong link with RSSI –Performing the flow-based load balance QoS support –Delay based priority control Prioritizing mechanism based on number of hops, congestion and time-stamp, to extend the IEEE802.e Multiple radios –Channel assignment with an autonomous scheme Security –Establishing the secure link for mesh networks