180 likes | 322 Views
CS215 Project. Optimizing the Routing within Fisheye Scope for LANMAR. Dan Pei Mengqiu Wang March 20, 2001. Outline. Problem of the original LANMAR Optimized Link State Routing (OLSR) Simulation and Experiments Conclusion and Future work. Problem with the original LANMAR.
E N D
CS215 Project Optimizing the Routing within Fisheye Scope for LANMAR Dan Pei Mengqiu Wang March 20, 2001
Outline Problem of the original LANMAR Optimized Link State Routing (OLSR) • Simulation and Experiments Conclusion and Future work
Problem with the original LANMAR Routing scheme within a Fisheye scope -- every node periodically exchanges topology information only with its neighbors -- updates only contain entries for the nodes within its fisheye scope and the landmark nodes -- at the same time, nodes exchange a distance vector of all landmarks with neighbors -- as a result, each node will maintain accurate topology information about its neighborhood and a distance vector to all landmark nodes Problem: As the network density increases, it becomes inefficient -- the number of neighbors grows largely, then: !Storage -- Topology information kept by each node increases dramatically !Bandwidth -- information exchanged increases dramatically
Optimized Link state routing (OLSR) Link state routing -- has many advantages, like converging quickly and the shortest path routing -- the only problem is the unaffordable flooding overhead Multipoint Relay (MPR) -- Each node selects a subset of its neighbors as MPR * The node is called MPR selector of its MPR set -- Each node only sends out topology information about its MPR selectors -- A node forwards updates only if it is the sender’s MPR
11 retransmission to diffuse a message up to 3 hops 24 retransmission to diffuse a message up to 3 hops Retransmission node Retransmission node Optimized Link state routing (OLSR)
Optimized Link state routing (OLSR) Calculate MPR set -- Some definition MPR(X): Multipoint relay set of node X running this algorithm N(X): One hop neighbor set of node X N2(X): Two-hop neighbor set of node X, not including one-hop neighbors D(X, Y): The number of one-hop neighbors of node Y, which is a member of N(X) -- Algorithm 1. Start with an empty MPR(X) 2. For each node Y in N(X), calculate D(X, Y) 3. First select those nodes in N(X) which provide the “only” path to reach some node in N2(X) as MPRs 4. While there still exist some nodes in N2(X) not covered by MPR(X), do: 4.1 For each node Y in N(X), calculate the number of nodes in N2(X) not covered by MPR(X), but reachable through Y, say R(Y) 4.2 Select that node in N(X) which has the maximum R(Y). In case of a tie, select the node as MPR whose D(X,Y) is greater.
Simulation and Experiments Goal – compare the performance of OLSR with FSRL Simulation -- simulate OLSR in Glomosim environment -- combine with Landmark routing: replace the routing scheme within fisheye scope * Each node exchanges topology information with its neighbors * Such updates only include information about its MPR selectors * MPRs forwards these updates * Such forwarding is only limited within the fisheye scope of senders -- we call it MPRL ( MPR-Landmark) Performance metrics -- Deliveryrate: the ratio of packets totally sent to packets totally received -- End-to-End Delay: the time delay between a packet is sent out until it is received -- Bandwidth Cost: the total number of bytes are used in topology information exchanging
Simulation and Experiments • Experiment Parameters • Field range: 1000m X 1000m • Bandwidth: 2Mbits/s • Transmission range: 150 meter • Mobility: Random Waypoint + Reference Point Group Mobility, • 10s pause time • Traffic pattern: 10 randomly chosen pairs • CBR: 1packet/2second, 512 bytes/packet. • Simulation time: 10 minutes • Fisheye Scope: 2 hops • Landmark Interval: one update / 0.5s • Groups: 4 • FSRL parameter: one topology update / 2s • MPRL parameters: Hello interval: 0.5s, TC interval: 2s
Simulation and Experiments Experiment Set I: Fixed mobility at 2m/s, node number changes from 100~300. Experiment Set II: Fixed node number at 200, change mobility from 0m/s~10m/s
Experiment Set I (1) Bandwidth cost VS Topology density
Experiment Set I (2) Delivery rate VS Topology density
Experiment Set I (3) End-to-End delay VS Topology density
Experiment Set I : Analysis Bandwidth VS. Density -- Nscope: # neighbors within scope = *(2*Transmission Range)2 *density -- N1hop: # one-hop neighbor = *(Transmission Range)2 *density -- FSRL’s Topology Update Size = O(Nscope * N1hop) = O(density2) -- For MPRL: Hello Size = O(N1hop) = O(density) TC Size = O(#Mpr Selector) = O(N1Hop) = O(density) Delivery Rate VS. Density -- FSRL and MPRL take the same Landmark protocol(including election and drifters) -- What makes the difference: * Only symmetric connection can be used to route data packet in MPRL * Routing Control Packets compete with the data packets End-to-End delay VS. Density -- Routing accuracy contributes -- Routing Control Packets compete with the data packets
Experiment Set II (1) Bandwidth cost VS Mobility
Experiment Set II (2) Delivery rate VS Mobility
Experiment Set II (3) End-to-End delay VS Mobility
Experiment Set II : Analysis • Bandwidth Cost VS. Mobility • -- Neither of MPRL and FSRL changes greatly with mobility, but MPRL performs better • Delivery Rate VS. Mobility • -- Both drop with mobility increase( because routing accuracy decreases) • End-to-End Delay VS. Mobility • -- Neither of MPRL and FSRL changes greatly with mobility, but MPRL performs better • -- Only symmetric connection can be used to route data packet in MPRL
Conclusion and Future work Conclusion -- We implemented and improved OLSR, integrated with Landmark Routing. • -- Simulation results show that: • With high node density, MPRL performs much better in terms of Bandwidth cost, End-to-End Delay, and Delivery rate. • When mobility increase, MPRL performs better in Bandwidth cost, end-to-end delay. But only slightly better in Delivery rate. • Future work • Extensive simulation to find an optimal configuration for MPRL