MobiArch 2008 MobiArch 2008 Shall we apply paging technologies to Shall we apply paging technologies to Proxy Mobile IPv6 ? Jong-Hyouk Lee, Tai-Myoung Chung (Sungkyunkwan University, Korea) J H k L T i M Ch (S k k U i it K ) Sangheon Pack (Korea University, Korea) Sri Gundavelli (Cisco, USA) 1
Proxy Mobile IPv6 y � The general way for IPv6 based mobility � host-based mobility protocols � Mobile IPv6 (RFC 3775), Hierarchical Mobile IPv6 (RFC 4140) � depending on the mobility stack installed on mobile hosts � depending on the mobility stack installed on mobile hosts � sending binding update messages and maintaining binding information � signaling concentrated to mobile hosts � The new trend in IPv6 based mobility � network-based mobility protocol k b d bili l � � Proxy Mobile IPv6 (PMIPv6, RFC 5213, published: August 2008) � no need for installing the mobility stack on mobile hosts � no need for installing the mobility stack on mobile hosts � ordinary hosts can hand off between different subnets � sending binding update messages and maintaining binding information are done by newly introduced mobility entities are done by newly introduced mobility entities 2
Problem statement � The newly proposed mobility entities in PMIPv6 � Mobile Access Gateway (MAG) � Mobile Access Gateway (MAG) � sending proxy binding update messages on behalf of a mobile host � Local Mobility Anchor (LMA) � maintaining all binding information for mobile hosts in its domain � maintaining all data traffic for mobile hosts in its domain � can be a performance bottleneck � can be a performance bottleneck LMA CN MAG 2 MAG 8 MAG 8 MAG 9 MAG MAG 1 MAG 3 MAG 5 MAG 7 MAG 4 MAG 6 MH MH 6 MH MH 3
Problem statement � Scalability issues in PMIPv6 � need to reduce binding update messages focusing to LMA � need to increase the number of supporting mobile hosts � need to optimize mobility management cost d t ti i bilit t t � � Paging technologies � can be a candidate solution for solving scalability issues f g y 4
Paging extension for PMIPv6 g g � Design considerations � support for unmodified mobile hosts � reduction in mobility signaling cost � avoidance of paging processing at a single point id f i i t i l i t � � Paging algorithm � Paging algorithm � fixed algorithm � hierarchical algorithm g � last-location algorithm � dynamic algorithm 5
Paging extension for PMIPv6 g g � Paging architecture • router based paging CN : paging state is distributed among MAGs • multicast group multicast group : MAGs have the same multicast group • time-based state LMA : paging state changes based on binding p g g g g Localized l d mobility domain MAG 8 MAG 9 MAG 2 MAG 7 MAG 1 MAG 3 Paging Area 3 MAG 5 Paging Area 1 Paging Area 2 MAG 4 MAG 6 6 6
Paging extension for PMIPv6 g g � Paging message sequence Paging Area Paging Area 1 Paging Area 2 Paging Area MAG 3 MAG 4 LMA CN MH MAG 1 MAG 2 Booting Detection 1 PBU 1. PBU 2. PBAck 3. Tunnel 1 for MH 4. Packets for MH Idle mode Idle mode 5. Idle Indication 5. Idle Indication Handoff Detection 6. Packets for MH 7. Paging request 8. PBU 9. PBAck 10. Tunnel 2 for MH Idle mode Idle mode Handoff Detection Handoff Detection 11. PBU 12. PBAck 7 13. Tunnel 3 for MH
Performance evaluation � System model � layered hexagonal network model � L-level paging area is consisted of 3L(L+1) + 1 � fluid-flow mobility model � for calculating cell (subnet) crossing and paging area crossing rates � for calculating cell (subnet) crossing and paging area crossing rates � Signaling Costs • PMIPv6 without paging p g g • PMIPv6 with paging p g g 8
Performance evaluation � Effect of paging level on the signaling cost v = 5, p = 0.005, s = 5% 5 0 005 5% v = 5, p = 0.01, s = 5% 5 0 01 5% PMIPv6 PMIPv6 16000 8000 PMIPv6 + paging PMIPv6 + paging 14000 7000 12000 6000 signaling cost signaling cost 10000 5000 8000 4000 The The 6000 3000 2000 4000 1000 2000 0 0 2 3 4 5 6 7 8 9 10 2 3 4 5 6 7 8 9 10 (a) The paging level (L) (b) The paging level (L) •The signaling cost for the paging extension is generally smaller than the basic PMIP6. •However, the paging extension consumes more cost when there are low-velocity mobile hosts in the above 7 paging level. p g g •The size of paging areas should carefully designed. 9
Performance evaluation � Effect of velocity on the signaling cost 4 4 4 4 L = 8, p = 0.005, s = 5% L 8 p 0 005 s 5% L L = 8, p = 0.01, s = 5% 8 p 0 01 s 5% x 10 x 10 PMIPv6 PMIPv6 5 PMIPv6 + paging PMIPv6 + paging 2.5 4.5 4 4 2 signaling cost signaling cost 3.5 3 1.5 2.5 The The 2 1 1.5 1 0.5 0 5 0.5 0 0 10 20 30 40 50 60 70 80 10 20 30 40 50 60 70 80 (a) The velocity of an MH (v) (b) The velocity of an MH (v) •When the velocity is lower than 10 m/s, the basic PMIP6 shows better performance. •However, as the velocity increases, the paging extension requires lower signaling cost. 10
Performance evaluation � Effect of paging level and velocity on the signaling cost concentrated to the LMA L p = 0.005 p = 0.005, s = 5% PMIPv6, v = 5 PMIPv6 + paging, v = 5 7000 7000 PMIPv6, v = 10 PMIPv6 + paging, v = 10 d to the LMA d to the LMA PMIPv6, v = 20 PMIPv6 + paging, v = 20 PMIPv6, v = 40 PMIPv6 + paging, v = 40 6000 6000 gnaling cost concentrated gnaling cost concentrated 5000 5000 4000 4000 3000 3000 2000 2000 The si The si 1000 1000 19 61 127 2 (19) 4 (61) 6 (127) (a) The number of cells (b) The paging level (the number of cells) p = 0.005, s = 10% p = 0.005, s = 20% PMIPv6 + paging, v = 5 PMIPv6 + paging, v = 5 7000 7000 PMIPv6 + paging, v = 10 PMIPv6 + paging, v = 10 ed to the LMA ed to the LMA PMIPv6 + paging, v = 20 PMIPv6 + paging, v = 20 PMIPv6 + paging, v = 40 PMIPv6 + paging, v = 40 6000 6000 5000 5000 5000 5000 signaling cost concentrate signaling cost concentrate 4000 4000 3000 3000 2000 2000 The s The s 1000 1000 11 2 (19) 4 (61) 6 (127) 2 (19) 4 (61) 6 (127) (c) The paging level (the number of cells) (d) The paging level (the number of cells)
Conclusions Conclusions � Proxy Mobile IPv6 is a centralized mobility architecture � posing heavy burden on the LMA � bring network traffic bottleneck at the LMA � Proposed paging extension is a decentralized architecture � P d i t i i d t li d hit t � distributing processing load among the MAGs � solving the bottleneck problem and scalability issues � solving the bottleneck problem and scalability issues 12
Questions ? Questions ? hurryon@gmail.com or jhlee@imtl.skku.ac.kr 13
Recommend
More recommend
