AN INTEGRATED ROUTING AND DISTRIBUTED SCHEDULING APPROACH FOR HYBRID IEEE E MESH NETWORKS FOR VEHICULAR BROADBAND COMMUNICATIONS Rahul Amin Electrical and Computer Engineering Clemson University M.S Thesis Defense, 11/25/2008 Advisor – Dr. Kuang-Ching Wang

Outline Background and Relevant Work Objective Network Model Routing and Scheduling Solution Analytical Model and Simulation studies Conclusions and Future Work

Background – WiMAX Operation Modes WiMAX – Worldwide Interoperability for Microwave Access – IEEE – IEEE e-2005 (Mobile WiMAX) extends mobility (handover) support Point-to-Multipoint (PMP) – Subscriber Stations (SS) connect via Base Station (BS) – BS in charge of coming up with a transmission schedule – Five Classes of Service defined to support QoS Mesh – SSs and BSs can communicate directly to other SSs and BSs – Scheduling solution can be centralized or distributed – QoS differentiation is on packet-by-packet basis

Background – PMP Mode Frame Structure Each frame divided into uplink and downlink sub-frame Uplink-to-downlink ratio adjusted according to traffic demand Sub-frame further divided into mini-slots to send data bursts

Background – Mesh Mode Frame Structure Each frame divided into control and data sub-frame No uplink or downlink sub-frame distinction Control sub-frame length depends on length of the mesh

Background – WiMAX Handover Support Standard Handover – Defined only for PMP mode in IEEE e-2005 – Each Mobile Station (MS) maintains an up-to-date CINR for target BSs – Either BS or MS can initiate the handover procedure – MS decides the target BS Two optional fast handover methods – Macro-Diversity Handover (Soft Handover) – Fast Base Station Switching (Hard Handover)

Background – Relevant Work WiMAX network architecture – Coverage: IEEE j workgroup (Relay Station Infrastructure) – Link-level performance: [Hartzog 2006] – Last mile Connectivity: [Bennett 2006], [Wongthavarawat 2003] – Mobile Ad Hoc Networks (MANET): [Sherman 2006], [Lebrun 2006] Scheduling schemes – PMP Mode: [Zhe 2007], [Jayaparvathy 2005], [Lera 2007], [Lin 2008] – Mesh Mode Centralized: [Wei 2005], [Chen 2006], [Qing 2006] Distributed: [Cao 2007]

Objective Challenge – WiMAX requires routing and scheduling packets for rapidly changing network topology due to fast moving vehicles Goals: – Develop Network Architecture – Routing and Scheduling solution – Throughput performance analysis

Network Model

Network Model Components Routing – Mobility-Aware Intra-Gateway Routing (MAIGR) – Ad Hoc Routing Protocols and Mobile IP Scheduling – PMP mode: between MS and MBS/MSS – Mesh mode: among MSS and MBS Handover – Scheduling options migrate from current BS to target BS

BS Protocol Architecture Base stations (MBS and MSS) support: – Dual radios (PMP mode, Mesh mode) – Routing: MAIGR MBS also supports: – Routing: IP and Mobile IP Mesh BS Mobile IP Home/foreign agents IP MAIGR PMP LINK + PHY Mesh SS Mesh LINK + PHY PMP LINK + PHY Mesh LINK + PHY MAIGR

MS Protocol Architecture Mobile Station (MS) supports: – Single radio (PMP mode) – Routing: IP – Transport layer protocols – Application specific protocols MS Application IP PMP LINK + PHY Transport (TCP, UDP)

Network Operation I.Base Station Setup 1. All MBSs use existing ad-hoc solutions to reach core network 2. All MSSs find closest path to MBS using MAIGR’s mesh topology discovery phase 3. All BSs (MSS and MBS) come up with initial mesh schedule using centralized mesh scheduler 4. All BSs maintain Mesh Link Capacity for PMP scheduler using the initially derived mesh schedule 5. All BSs use PMP scheduler to communicate with MS II.Mobile Station Setup 6.MS sets up routes to transmit/receive traffic to/from MBS using MAIGR 7.MS uses PMP scheduler to communicate with MSS/MBS III.Operational Phase 8. Depending on PMP traffic load at each BS, Mesh scheduler at each BS adapts its schedule 9. PMP scheduling options for a MS migrate to the target BS during the handover procedure 10. Mobile IP used by MBSs to assign care-of address to MS that switches routing zones

Base Station Route Setup I.Base Station Setup 1. All MBSs use existing ad-hoc solutions to reach core network 2. All MSSs find closest path to MBS using MAIGR’s mesh topology discovery phase 3. All BSs (MSS and MBS) come up with initial mesh schedule using centralized mesh scheduler 4. All BSs maintain Mesh Link Capacity for PMP scheduler using the initially derived mesh schedule 5. All BSs use PMP scheduler to communicate with MS II.Mobile Station Setup 6.MS sets up routes to transmit/receive traffic to/from MBS using MAIGR 7.MS uses PMP scheduler to communicate with MSS/MBS III.Operational Phase 8. Depending on PMP traffic load at each BS, Mesh scheduler at each BS adapts its schedule 9. PMP scheduling options for a MS migrate to the target BS during the handover procedure 10. Mobile IP used by MBSs to assign care-of address to MS that switches routing zones

Base Station Route Setup MSSMBS Zone 1Zone 2 MSSMBSMSS Core Network MSS MSS: Mesh SS MBS: Mesh BS MS: Mobile Station Traditional Ad Hoc Routing Solutions MSH_NCFG: Entry

Base Station Scheduler Setup I.Base Station Setup 1. All MBSs use existing ad-hoc solutions to reach core network 2. All MSSs find closest path to MBS using MAIGR’s mesh topology discovery phase 3. All BSs (MSS and MBS) come up with initial mesh schedule using centralized mesh scheduler 4. All BSs maintain Mesh Link Capacity for PMP scheduler using the initially derived mesh schedule 5. All BSs use PMP scheduler to communicate with MS II.Mobile Station Setup 6.MS sets up routes to transmit/receive traffic to/from MBS using MAIGR 7.MS uses PMP scheduler to communicate with MSS/MBS III.Operational Phase 8. Depending on PMP traffic load at each BS, Mesh scheduler at each BS adapts its schedule 9. PMP scheduling options for a MS migrate to the target BS during the handover procedure 10. Mobile IP used by MBSs to assign care-of address to MS that switches routing zones

Base Station Scheduler Setup Core Network MSSMBSMSS MSH_CSCH: Request MSH_CSCH: Request MSS: Mesh SS MBS: Mesh BS MS: Mobile Station MSH_CSCH: Grant x2x3x 2xx x xx xxx

Centralized Mesh Scheduling Solution Interference-aware solution achieving near-optimal throughput [Wei 2005, VTC] – Procedure: Transmission occurs in order of highest transmission demand Non-interfering links transmit simultaneously – Proposed changes: QoS considered – Constraints: Unidirectional traffic considered MSSMBSMSS x2x3x 2xx   Carrier Sense Range: 1 Tx RxTxRx

Mobile Station Setup I.Base Station Setup 1. All MBSs use existing ad-hoc solutions to reach core network 2. All MSSs find closest path to MBS using MAIGR’s mesh topology discovery phase 3. All BSs (MSS and MBS) come up with initial mesh schedule using centralized mesh scheduler 4. All BSs maintain Mesh Link Capacity for PMP scheduler using the initially derived mesh schedule 5. All BSs use PMP scheduler to communicate with MS II.Mobile Station Setup 6.MS sets up routes to transmit/receive traffic to/from MBS using MAIGR 7.MS uses PMP scheduler to communicate with MSS/MBS III.Operational Phase 8. Depending on PMP traffic load at each BS, Mesh scheduler at each BS adapts its schedule 9. PMP scheduling options for a MS migrate to the target BS during the handover procedure 10. Mobile IP used by MBSs to assign care-of address to MS that switches routing zones

Mobile Station Setup Core Network MSSMBSMSS MSS: Mesh SS MBS: Mesh BS MS: Mobile Station x2x3x 2xx x xx xxx MS DHCP REQ DHCP REP Service Flow: REQ Service Flow: REP

Operational Phase Scheduler Interaction I.Base Station Setup 1. All MBSs use existing ad-hoc solutions to reach core network 2. All MSSs find closest path to MBS using MAIGR’s mesh topology discovery phase 3. All BSs (MSS and MBS) come up with initial mesh schedule using centralized mesh scheduler 4. All BSs maintain Mesh Link Capacity for PMP scheduler using the initially derived mesh schedule 5. All BSs use PMP scheduler to communicate with MS II.Mobile Station Setup 6.MS sets up routes to transmit/receive traffic to/from MBS using MAIGR 7.MS uses PMP scheduler to communicate with MSS/MBS III.Operational Phase 8. Depending on PMP traffic load at each BS, Mesh scheduler at each BS adapts its schedule 9. PMP scheduling options for a MS migrate to the target BS during the handover procedure 10. Mobile IP used by MBSs to assign care-of address to MS that switches routing zones

PMP Mode Scheduler UGS ertPS rtPS nrtPS BE Enough Free Slots Available? New Flow Requests Active Connection Queue Mesh link limitation? Reject Flow Yes No Pending Connection Queue Pending Connection Queue Empty? Yes No Queue Size > Threshold? Yes PMP Scheduler Distributed Adaptation Centralized Scheduling Mesh Scheduler

Mesh Scheduler Distributed Adaptation (General Case) MSS MBSMSS x2x3x 2xx MSS MBSMSS 02x3x 2xx xxxxxx 0xxxx – Up to x slots can be borrowed in general

Mesh Scheduler Distributed Adaptation (Upper Bound) Case I: – Borrowing occurs from MSS away from MBS – 2x slots can be borrowed as an upper bound MSS MBSMSS x2x3x 2xx MSS MBSMSS 03x 2xx xxxxxx 3x00xxx

Mesh Scheduler Distributed Adaptation (Upper Bound) Case II: – Borrowing occurs from MSS closer to MBS – x/2 slots can be borrowed as upper bound MSS MBSMSS x2x3x 2xx MSS MBSMSS 1.5x 3x 2xx xxxxxx 1.5x0x x xx

Operational Phase Handover Scenario I.Base Station Setup 1. All MBSs use existing ad-hoc solutions to reach core network 2. All MSSs find closest path to MBS using MAIGR’s mesh topology discovery phase 3. All BSs (MSS and MBS) come up with initial mesh schedule using centralized mesh scheduler 4. All BSs maintain Mesh Link Capacity for PMP scheduler using the initially derived mesh schedule 5. All BSs use PMP scheduler to communicate with MS II.Mobile Station Setup 6.MS sets up routes to transmit/receive traffic to/from MBS using MAIGR 7.MS uses PMP scheduler to communicate with MSS/MBS III.Operational Phase 8. Depending on PMP traffic load at each BS, Mesh scheduler at each BS adapts its schedule 9. PMP scheduling options for a MS migrate to the target BS during the handover procedure 10. Mobile IP used by MBSs to assign care-of address to MS that switches routing zones

Operation Phase Handover Scenario MSSMBS Zone 1Zone 2 MSSMBSMSS MS MSS Tactical Network Gateway Tactical Network Gateway MSS MSS: Mesh SS MBS: Mesh BS MS: Mobile Station HO_IND Msg Care-of Address Mobile IP Msg

Analytical Model Throughput Comparison a)centralized only mesh scheduling algorithm b)centralized with distributed adaptation algorithm Topology considered – Symmetric Chain Topology – Single Intersection Topology Throughput bounds derived for – Dense Network – Sparse Network

Percentage Increase in Network Throughput per Routing Zone (Dense Network) MBS MSS 1’ MSS 1 MSS 2’ MSS 2 MSS 3 MSS 3’ Link 1Link 2Link 3 Link 1 ’ Link 2 ’ Link 3 ’ x2x3x 2xx Case I: Case II: x – q 3’ ParameterDescription qiqi Number of unused PMP data slots at base station i zTotal number of data slots per Mesh Frame n+n’Total number of MSS in routing zone kBorrowing MSS’s # of hops from MBS x – q 2’ x – q 1’ x – q 2 x – q 1 x – q 3

Percentage Increase in Network Throughput per Routing Zone (Sparse Network) MBS MSS 1’ MSS 1 MSS 2’ MSS 2 MSS 3 MSS 3’ Link 1Link 2Link 3 Link 1 ’ Link 2 ’ Link 3 ’ x2x3x 2xx 0 Case I: Case II: ParameterDescription qiqi Number of unused PMP data slots at base station i xPMP data slots per frame at Lender kBorrowing MSS’s # of hops from MBS x-q 2’ 0000

Simulation Studies Network Simulator ns-2 used – Based on NIST IEEE module (version pre-release2) Supports PMP mode, standard handover Scheduler functionality added Extension made to support: – TDD mesh links between base stations (emulated with wired links) – Dual radios at each base station – Chain topology presented – Carrier sensing range equal to communication range assumed – Analytical Model verified using simulation results

Mesh Mode Simulation Parameters ParameterValue Channel Bandwidth10 MHz Modulation SchemeOFDM_64QAM_3_4 Bits/Symbol856 Symbol/Slot1 Frame Duration10 ms Slots/Frame221 Control Sub-Frame7 * MSH_CTRL_LEN = 49 Data Sub-Frame221 – (7 * MSH_CTRL_LEN) = 172 Pending Queue Threshold0 Communication Range1 Carrier Sensing Range1 Maximum Routing Zone Rate18.91 Mbps (14.72 Mbps for Data)

Simulation Model – Dense Network Internet Sink Tactical Network Gateway MBS MSS 2 MSS 3 MSS 1 MSS 4 MSS 5 MSS MS * 0.64 Mbps UGS Flows per MS (16 Data Slots)

Simulation Results – Dense Network (Stationary) Centralized Scheduling Only Distributed Adaptation Scheduling CentralizedDistributed Throughput (Mbps)8.27 (13 flows)8.95 (14 flows) Pending Flows0/1/3/5/70/0/2/4/6

Simulation Model – Sparse Network Internet Sink Tactical Network Gateway MBS MSS 2 MSS 3 MSS 1 MSS 4 MSS 5 MSS 0 MS * 0.64 Mbps UGS Flow per MS (8 Data Slots)

Simulation Model – Sparse Network Internet Sink Tactical Network Gateway MBS MSS 2 MSS 3 MSS 1 MSS 4 MSS 5 MSS 0 MS * 0.64 Mbps UGS Flow per MS (8 Data Slots)

Simulation Results – Sparse Network (Stationary) Centralized Throughput (Mbps) Distributed Throughput (Mbps) % Increase 5 MSs1.92 (3 flows)3.19 (5 flows)66 10 MSs1.92 (3 flows)6.38 (10 flows)232

Conclusions The network architecture simplifies scheduling re- computation by only involving base-stations as opposed to a complete ad-hoc solution Distributed adaptation does not increase the throughput significantly under dense network conditions (~8.2 %) due to the inability of neighboring BSs to lend slots Distributed adaptation increases throughput significantly under sparse network conditions (~ 232 %)

Future Work Different topologies such as grid topology need to be studied Borrowing from 1-hop neighbor only is not optimal and studies needs to be expanded to determine the performance improvement if borrowing from multi-hop neighbors is allowed

Thank You

Backup - Routing Partitions

Mobility-Aware Intra-Gateway Routing Mesh Topology discovery – Each Mesh BS sends a flooding message at periodic intervals – Each Mesh SS records the next hop towards the closest Mesh BS Packet Forwarding from Mesh BS to MS – At serving BS, the next hop is the connection identifier (CID) of MS’s data connection established during network entry or handover – At non-serving BS, the next hop is a mesh link CID which is obtained in one of two ways When receiving a packet from neighboring BS forwarded by an unknown MS, record the neighboring BS as next hop to the unknown MS When notified of a handover of currently associated MS, record target BS as next hop to MS Packet Forwarding from MS to Mesh BS – At MS, the next hop is always the CID of currently serving BS’s data connection – At subsequent Mesh SS’s, the next hop is the mesh link CID towards the closest Mesh BS obtained during Mesh Topology discovery phase

Scheduling Solution PMP Mode Scheduling – Order determined according to service classes (UGS, ertPS, rtPS, nrtPS, BE) – Same service classes are serviced in FIFO order Mesh Mode Scheduling – Centralized interference-aware solution Transmission occurs in order of highest transmission demand [Wei] – Distributed Adaptation Need-based borrowing

Mesh Mode Scheduler Initial schedule using centralized algorithm – Upon network deployment – Every time BS is added to the mesh – Uses average traffic load information if available, else assumes uniform traffic distribution Distributed adaptation copes with changes in traffic loads – Whenever pending queue length exceeds predefined threshold – Slots borrowed from immediate (1-hop) neighbors – Three-way handshake technique used

Distributed Adaptation

Centralized Algorithm