Improving network mobility in HAP networks
Transcription
Improving network mobility in HAP networks
First COST 297 – HAPCOS Workshop – York, UK Improving Network Mobility in HAP Networks Andrej Vilhar, Vilhar, Roman Novak Department of Communication Systems Jožef Stefan Institute, Ljubljana, Slovenia 26 - 27 October 2006 First COST 297 - HAPCOS Workshop - York, UK 1 Outline Issues related to network mobility (and proposed solutions) • route optimization • handoff efficiency Basic concepts of multihoming Simulation aspects • simulation scenario, assumptions, tools, methods etc. Performance analysis • route path optimality • handoff efficiency Conclusion 26 - 27 October 2006 First COST 297 - HAPCOS Workshop - York, UK 2 1 IP mobility: end-nodes Mobile IPv6 6 7 8 5 • uninterruptible service, route optimization 26 - 27 October 2006 First COST 297 - HAPCOS Workshop - York, UK 3 IP mobility: mobile networks Airplanes, ships, buses, trains etc. NEMO extensions of Mobile IPv6 • uninterruptible service, no route optimization 3 4 1st issue: How to achieve optimal route paths for mobile networks? 26 - 27 October 2006 First COST 297 - HAPCOS Workshop - York, UK 4 2 Public transport vehicles The movement is limited to certain area and/or can be quasi-deterministic. This implies route optimization problem could be alleviated by carefully selecting the HA’s location. If single HA is not sufficient, multiple HAs can be introduced. Possible solution for route optimization problem. 26 - 27 October 2006 First COST 297 - HAPCOS Workshop - York, UK 5 Handoff procedures When mobile network moves in a topology a set of procedures occur: During the Mobile router: • • • • detects the movement on a link layer, detects the movement on a network layer, forms a new care-of-address, informs home agent by sending binding update. Home agent: • confirms by sending binding acknowledgement. 26 - 27 October 2006 First COST 297 - HAPCOS Workshop - York, UK execution of the procedures, packets are being lost. 2nd issue: How to achieve smoother handoffs? 6 3 Multiple active egress interfaces By keeping multiple egress interfaces active, packets can be saved from loss, even when switching between different access networks. Possible solution for achieving smoother handoffs. 26 - 27 October 2006 First COST 297 - HAPCOS Workshop - York, UK 7 Multihoming A multihomed node is defined by [1] as a node with multiple addresses: • either due to multiple prefixes advertised in the access link • or because the node has multiple interfaces. According to such definition the mobile network with: • multiple HAs has several home addresses • multiple active egress interfaces has several addresses multihomed Besides achieving route path optimality and smoother handoffs, multihoming offers other benefits [1]: • Load sharing/load balancing, bi-casting, redundancy/fault-recovery. [1] T. Ernst, N. Montavont, R. Wakikawa, E. Paik, C. Ng, K. Kuladinithi, and T. Noel. (2005, October) Goals and benefits of multihoming. IETF Internet Draft. 26 - 27 October 2006 First COST 297 - HAPCOS Workshop - York, UK 8 4 Multihoming configurations To identify different possible multihoming configurations the following taxonomies were suggested [2][3]: • End-nodes: • Mobile networks: (iface, HoA, CoA) (MR, HA, MNP) To fully describe a certain NEMO configuration analyzed in our paper, we suggest the taxonomy: • (MR, HA, MNP, iface, CoA) [2] N. Montavont, R. Wakikawa, T. Ernst, C. Ng, and K. Kuladinithi. (2005, October) Analysis of multihoming in mobile IPv6. IETF Internet Draft. 26 - 27 October 2006 [3] C. Ng, E. Paik, and T. Ernst. (2005, October) Analysis of multihoming in network mobility support. IETF Internet Draft. First COST 297 - HAPCOS Workshop - York, UK 9 HAP important configurations (MR, HA, MNP, iface, iface, CoA) CoA) (1,1,1,N,1) – multiple interfaces, but only one active CoA – very basic, (1,1,N,N,N) – multiple advertised prefixes – allows end-nodes being multihomed, (1,N,1,N,N) – multiple HAs support single MR – despite the fact multihoming is transparent to end-nodes (single MNP), they do benefit from it, (N,1,1,N,N) – multiple MRs – one MR per train car (N,1,N,N,N) – multiple MNPs – better support for train modularity 26 - 27 October 2006 First COST 297 - HAPCOS Workshop - York, UK 10 5 Simulated configurations (MR, HA, MNP, iface, iface, CoA) CoA) Standard MIP configuration - (1,1,1,2,1): • used as a reference. Multihomed MIP configuration - (1,2,1,2,2): • two CoAs in parallel, • two HAs, • HAHA protocol. 26 - 27 October 2006 First COST 297 - HAPCOS Workshop - York, UK 11 Simulation scenario, assumptions and tools A train traveling from western to eastern part of Slovenia. HAP best choice LOS NLOS UMTS, WiMAX full coverage MR Empirical LOS data (two-state channel model) for this particular railway track was taken into account. OPNET Modeler simulation tool. 26 - 27 October 2006 First COST 297 - HAPCOS Workshop - York, UK 12 6 Simulation topology HA HA 26 - 27 October 2006 First COST 297 - HAPCOS Workshop - York, UK 13 Assessment of route path optimality If route path is optimal: • end-to-end delays lower higher throughput for TCP-based applications, • resource consumption reduced backhaul link utilization is lower. 26 - 27 October 2006 First COST 297 - HAPCOS Workshop - York, UK FTP file transfer from CN to MN 14 7 Assessment of route path optimality (CN in global network) network) LOS Visibility [LOS/NLOS], Backhaul link utilization [%] Visibility [LOS/NLOS], Throughput [Mbit/s] 1.2 HAGS 1 HAHAP 0.8 0.6 0.4 0.2 NLOS 40 LOS 30 HAHAP 20 10 HAGS 0 NLOS 0 500 550 600 650 700 750 800 500 550 600 Time [s] Throughput 26 - 27 October 2006 650 700 750 800 Time [s] Backhaul link utilization First COST 297 - HAPCOS Workshop - York, UK 15 Assessment of route path optimality (CN in HAP network) network) Visibility [LOS/NLOS], Backhaul link utilization [%] Visibility [LOS/NLOS], Throughput [Mbit/s] 4.8 LOS 4 HAHAP 3.2 2.4 HAGS 1.6 0.8 NLOS 80 LOS 60 HAGS 40 20 HAHAP 0 NLOS 0 500 550 600 650 700 750 800 500 550 Time [s] Throughput 600 650 700 750 800 Time [s] Backhaul link utilization Standard MIP route optimality depends on CN, HA and LOS Multihomed MIP routes are always optimal 26 - 27 October 2006 First COST 297 - HAPCOS Workshop - York, UK 16 8 Assessment of handoff efficiency If handoff is efficient: • less (none) packets are lost TCP-based flows are not appropriate for assessment due to built-in back-off algorithm Voice conversation between CN and MN 100 packets/s, 57 bytes/packet in both UL and DL 26 - 27 October 2006 First COST 297 - HAPCOS Workshop - York, UK 17 Assessment of handoff efficiency: efficiency: standard MIP vs. vs. multihomed MIP LOS NLOS HAP HA: UL: i-face TER DL: i-face TER 1 HA, 1 CoA (standard MIP) MR HAP HA L2 GS L2 GS HA (P): DL: i-face TER UL: i-face TER HAP HA: DL: i-face TER LOS NLOS 2 HA, 2 CoA (multihomed MIP) MR TER HAP HA L2 GS HA L2 GS HA (P): DL: i-face TER UL: i-face TER HAP HA: DL: i-face TER TER RS RS RA RA all packets lost BU HAP HA: DL: i-face HAP UL: packets lost BA UL: packets discarded by non-primary HA GS HA (P*): DL: i-face TER UL: i-face TER HAP HA (P): DL: i-face HAP P, BU BA BIU BIA GS HA: DL: i-face HAP HAP HA (P): DL: i-face HAP UL: i-face HAP HAP HA: DL: i-face HAP UL: i-face HAP all packets lost MR TER L2 HAP HA GS L2 MR GS HA TER L2 HAP HA GS HA: DL: packets lost HAP HA (P): DL: packets lost UL: i-face TER L2 RS RA all packets lost UL: packets discarded by non-primary HA P, BU GS HA (P): DL: i-face TER HAP HA (P*): DL: packets lost UL: i-face TER BIU BA BIA BU MR GS HA (P): DL: i-face TER HAP HA: DL: i-face TER UL: i-face TER HAP HA: DL: i-face TER UL: packets lost BA HAP HA L2 HAP HA: DL: i-face TER UL: i-face TER GS L2 TER RS RA all packets lost BU . . . . . . . . . . . . MR HAP HA L2 . . . GS HA L2 . . . GS HA (P): DL: i-face TER UL: i-face TER HAP HA: DL: i-face TER TER . . . . . . GS HA (P): DL: i-face TER UL: i-face TER HAP HA: DL: i-face TER . . . UL: i-face TER … optimal path UL: i-face TER … unoptimal path (P) … primary (P*) … old primary 26 - 27 October 2006 First COST 297 - HAPCOS Workshop - York, UK 18 9 Assessment of handoff efficiency: efficiency: standard MIP vs. vs. multihomed MIP 120 multihomed MIP 80 40 NLOS 15 Visibility [LOS/NLOS], Packet loss [packets] Visibility [LOS/NLOS], Packet loss [packets] standard MIP LOS 160 standard MIP LOS 12 9 6 3 NLOS multihomed MIP 0 300 400 500 600 700 800 0 300 400 Time [s] 500 600 700 800 Time [s] Packet loss on HAP interface (LOS to NLOS transition) Packet loss on terr. network interface (NLOS to LOS transition) Standard MIP: losses on both interfaces Multihomed MIP: no loss on terr. network interface, lower loss on HAP interface 26 - 27 October 2006 First COST 297 - HAPCOS Workshop - York, UK 19 Conclusions Multiple HAs can improve route optimality for quasi-deterministic movements. • The throughput increases, the load on backhaul link is reduced. Simultaneous usage of multiple access interfaces is beneficial in terms of handoff efficiency. Further challenges related to multihoming: • Handoff anticipation, link load balancing, efficient coordination between MRs. Other important challenges: • nested mobility, • hierarchical approach to mobility provision. 26 - 27 October 2006 First COST 297 - HAPCOS Workshop - York, UK 20 10 First COST 297 – HAPCOS Workshop – York, UK Thank you! andrej.v ilhar@ @ijs.si ndrej.vilhar 26 - 27 October 2006 First COST 297 - HAPCOS Workshop - York, UK 21 11