ICE 1332/0715 Mobile Computing (Summer, 2008)

Transcription

1 ICE 1332/0715 Mobile Computing (Summer, 2008) Ns-2 Laboratory Prof. Chansu Yu In-Class Lab: Mobile IP 5 nodes 2 wired nodes, W(0) and W(1) 2 mobile agents, HA and FA A mobile host, MH Traffic: W(0) => MH W(0) W(1) HA MH Mobility MH moves toward FA Then, moves back to HA Routing path changes to W(0) W(1) FA MH and then W(0) W(1) HA MH again 2 (1,2) (2,2) (650,600) 1

2 In-Class Lab: Scenario Traffic A wired node W(0), with address is sending TCP packets to a MH associated to its HA. Mobility MH starts moving towards FA, gets disconnected from the HA (~ 113s), then connects to the FA later on (~139s). While moving back towards the HA, the MH gets disconnected again (~ 213s) then reconnects to the HA s network (~ 239s). 3 (1,2) $ns_ at "$MH setdest " $ns_ at "$MH setdest " (2,2) (650,600) In-Class Lab: Nodes set W(0) [$ns_ node 0.0.0] set W(1) [$ns_ node 0.1.0] $ns_ node-config -mobileip ON \ -wiredrouting ON \ -adhocrouting DSDV \ -mactype Mac/802_11 \ -anttype Antenna/OmniAntenna \ -proptype Propagation/TwoRayGround \ -phytype Phy/WirelessPhy \ -channeltype Channel/WirelessChannel \... Wired nodes, W(0) and W(1) Mobile agents, HA and FA set HA [$ns_ node 1.0.0] set FA [$ns_ node 2.0.0] set MH [$ns_ node 1.0.1] set HAaddress [AddrParams addr2id [$HA nodeaddr]] [$MH set regagent_] set home_agent_ $HAaddress 4 Mobile node, MH (Ch of ns-2 manual) 2

3 Hierarchical Addresses (Ch. 32 of ns-2 manual) In this topology we are using hierarchical addresses: $ns_ node-config -addresstype hierarchical Composed of 3 domains (0.x.x to which the wired nodes belong, 1.x.x where HA and MH belong, and 2.x.x where the FA belongs), therefore: AddrParams set domain_num_ 3 The first of the 3 domains has 2 clusters: 0.0.x and 0.1.x. The second domain (1.x.x) has only 1 cluster (i.e. 1.0.x), and the last domain also has 1 cluster (i.e. 2.0.x), therefore: lappend cluster_num Finally, cluster 0.0.x has 1 node; cluster 0.1.x has 1 node; cluster 1.0.x has 2 nodes; and cluster 2.0.x has 1 node. Therefore: lappend eilastlevel In-Class Lab: Wired Links All wired links are 5Mbps and have 2ms delays (Ch.6 Links: Simple Links, ns-2 manual) $ns_ duplex-link $W(0) $W(1) 5Mb 2ms DropTail $ns_ duplex-link $W(1) $HA 5Mb 2ms DropTail $ns_ duplex-link $W(1) $FA 5Mb 2ms DropTail We can control how these wired links look like in NAM by setting: $ns_ duplex-link-op $W(0) $W(1) orient down $ns_ duplex-link-op $W(1) $HA orient left-down $ns_ duplex-link-op $W(1) $FA orient right-down HA W(0) W(1) FA 6 3

4 In-Class Lab: Mobile IP Download infra.tcl fil-tcp.awk Run ns infra.tcl Source files ~ns/mip.{cc,h}, ~ns/mip-reg.cc, ~ns/tcl/lib/ns-mip.tcl, and ~ns/tcl/lib/ns-wireless-mip.tcl 7 Lab Report Mobility: none Traffic: S1 and S2 can send traffic to sink either 1448-byte packets every second 724-bytes packets every second Or, TCP traffic Compare throughput from S1 and S2 with various combinations of traffic (0,0) (100,0) (200,0) 8 4

5 Simulation Study Simulation study usually based on random numbers (seed) Results vary depending on the selection of seed number One single simulation run does not tell us the true performance metrics Methodology Run multiple simulation runs with different seed numbers Take the average Is it enough? Observed packet delays are 2.9, 3.0 and 4.1 seconds Observed packet delays are 0.0, 3.0 and 6.0 seconds Averages are 3.0 seconds in both cases but we are more confident about this value 9 in the first case. Confidence Interval & Level How much are we confident that the TRUE value lies within a certain value interval? Confidence level Confidence interval Example: Average packet delay is within [2.9, 3.1] seconds with probability 95%. Confidence level: 95% Confidence interval: [2.9, 3.1] seconds or ±3.3% error (=0.1/3.0) 10 5

6 95% Confidence Intervals Write as: µ = x ±1. 96σ x Method Get the measurements Calculate the mean and standard deviation If all the observed measurements are within the CI, we can say that We are 95% confident that the true population mean is between 11 99% Confidence Intervals Write as: µ = x ±1. 96σ x Method Get the measurements Calculate the mean and standard deviation If all the observed measurements are within the CI, we can say that We are 99% confident that the true population mean is between 12 6

7 Using the Sample Std.Dev. Since we do not know our true population standard deviation to calculate the standard error, we must substitute the sample standard deviation in the standard error formula The provides us with an estimate of our standard error; thus, our confidence intervals can only be approximate Can be substituted because of the Central Limit Theorem, that a n greater than 30 will give us a nearly normal distribution 13 At Least When confidence interval/level is not doable, You should run multiple runs of the same simulation scenarios with different seeds to obtain an average performance Different seeds??? Same traffic (#connections), but different pairs Same traffic between the same pairs, but different mobility pattern 14 7