Multicast Research Multicast Routing ns-2 for Multicast Research - PDF document

Multicast Research • Multicast Routing ns-2 for Multicast Research – Dense Mode, Sparse Mode – Source Tree, Shared Tree • Reliable Multicast Polly Huang – Whiteboard, File Transfer AT&T Labs Research huang@catarina.usc.edu http://netweb.usc.edu/huang 12 August, 1999 1 2 Reliable Multicast Issues in Reliable Multicast • Loss Detection – ack or nack Data • Request and Repair R S Ack – who (everyone or certain ones) Data R – how (unicast, multicast, or subcast) R Data Ack Ack Ack S Data Ack R Ack Data R Data R 3 4 SRM SRM (Scalable Reliable Multicast) • Loss Detection Data Session R – nack Request R Session Session – packets not in sequence S Session Request R – periodic ‘ Session ’ messages Session Data Data R • Request and Repair Session R – few of those who lose the packet – few of those who have the requested packet Timers to avoid duplicate Requests/Recoveries – requesters and repairers multicast requests and repair 5 6 1

Session Messages Request/Repair • Periodic • Requester – The closest to the data source • Multicast • Repairer • Message Contents – The closest to the requester – last packet seen – timestamp • Each member knows the delays to every other member. 7 8 Timers SRM Scripts • Deterministic Timer • Chain Topology – string topology • Star Topology • Random Timer • Y Topology – star topology • Adaptive SRM • Both Timers – tree topology • Adaptive Timer – adapting weights of deterministic and random timers 9 10 Scalability Simulation Results (1) Recovery Delay • Delay – Recovery delay 30 25 Delay in RTT • Overhead 20 15 – Number of duplicate requests/repairs 10 5 0 20 40 60 80 Session Size 11 12 2

Simulation Results (2) Simulation Results (3) Duplicate Requests Duplicate Repairs 4 15 Number of Requests Number of Repairs 3 10 2 5 1 0 0 20 40 60 80 20 40 60 80 Session Size Session Size 13 14 Multicast Routing • set ns [new Simulator] Multicast in ns-2 • $ns multicast • $ns mrtproto <type> • <type> CtrMcast, DM, ST, BST 15 16 CtrMcast CtrMcast Configuration • Centralized Multicast • set mproto CtrMcast • No control overhead or delay • set mrthandle [$ns mrtproto $mproto {}] • Support 2 tree types • $ns at 0.5 "$mrthandle switch-treetype $group" – Shared tree (default) – Source tree (switch tree type) • set rp [list $n0] • $mrthandle set_c_rp $rp 17 18 3

DM DM Configuration • Dense Mode Multicast • DM set PruneTimeout $timeout • Support 2 dense mode types • DM set CacheMissMode <mode> – PIM-like • <mode>: pimdm or dvmrp – DVMRP-like • $ns mrtproto DM “” 19 20 ST ST Configuration • Sparse Mode Multicast • ST set RP_($group) $n0 • Shared Tree Only • $ns_ mrtproto ST "" 21 22 BST BST Configuration • Sparse Mode Multicast • BST set RP_($group) $n0 • Bi-directional Shared Tree • $ns_ mrtproto BST "" 23 24 4

SRM Agent Types • set srm [new Agent/SRM] SRM in ns-2 • set srm [new Agent/SRM/Deterministic] • set srm [new Agent/SRM/Probabilistic] • set srm [new Agent/SRM/Adaptive] 25 26 Session Member Session Source • set src [new Application/Traffic/CBR] • $srm set dst_ $group • $src set packet_size_ $packetSize • $srm set fid_ $fid • $src attach-agent $srm • $srm log [open srmEvents.tr w] • $srm trace [open srmStats.tr w] • $srm set tg_ $src • $ns attach-agent $n0 $srm • $srm set app_fid_ 0 • $ns at 1.0 “$srm start” • $srm set packetSize_ $packetSize • $ns at 2.0 "$srm start-source" 27 28 Questions • SRM Data Throughput Exercise II: SRM – data bytes / total bytes transmitted • Compare Throughput using DM vs. ST – DV throughput / ST throughput 29 30 5

Key Steps Simulation Description (1) • Topology • Simulation Script – binary tree • Post-processing Script – 15 nodes, starting from 0 – 10MB bandwidth – 20ms delay • Multicast Routing – DM vs. ST 31 32 Simulation Description (2) Simulation Description (3) • Sessions • Member – 2 sessions – 5 members/session – CBR sources – SRM agents, – sources randomly selected – members randomly selected – start at time 2s – start at time 1s • Terminate at 15s 33 34 Post-Processing Questions • trace-all output • SRM Data Throughput • r entries – data bytes / total bytes transmitted • calculate data bytes and total bytes • Compare Throughput using DM vs. ST – DV throughput / ST throughput 35 36 6