Tutorial Mobility Modeling for Design and Analysis of Ad Hoc ...

Reviews
UNIVERSITY OF SOUTHERN CALIFORNIA Tutorial Mobility Modeling for Design and Analysis of Ad Hoc Wireless Networks Ahmed Helmy Electrical Engineering Department University of Southern California helmy@usc.edu Webpage: http://ceng.usc.edu/~helmy Wireless Networking Lab: http://nile.usc.edu UNIVERSITY OF SOUTHERN CALIFORNIA Outline • – – Ad Hoc Networks & Mobility Classification (15 min.) Synthetic and Trace-based Mobility Models The Need for Systematic Mobility Framework • – Survey of the Major Mobility Models (30 min.) Random models - Group mobility models – Vehicular (Manhattan/Freeway) models - Obstacle models • – – Characterizing Mobility (30 min.) Mobility Dimensions (spatial and temporal dependency, geographic restrictions) Mobility Metrics (spatio-temporal correlations, path and link duration) Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Outline (contd.) • – – Mobility-centric framework to analyze ad hoc networks (40 min.) The IMPORTANT mobility framework Case Studies: BRICS, PATHS, MAID • – – Trace-based mobility modeling (30 min.) Analyzing wireless network measurements and traces Survey-based and observation techniques • – – Mobility simulation and analysis tools (20 min.) Available software packages and tools Resources and related projects Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Wireless Mobile Ad hoc Networks (MANETs) • A Mobile Ad hoc Network (MANET) is a collection of mobile devices forming a multi-hop wireless network with minimal (or no) infrastructure • To evaluate/study adhoc networks mobility and traffic patterns are two significant factors affecting protocol performance. • Wireless network performance evaluation uses: – Mobility Patterns: usually, uniformly and randomly chosen destinations (random waypoint model) – Traffic Patterns: usually, uniformly and randomly chosen communicating nodes with long-lived connections • Impact of mobility on wireless networks and ad hoc routing protocols is significant Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Example Ad hoc Networks Mobile devices (laptop, PDAs) Vehicular Networks on Highways Hybrid urban ad hoc network (vehicular, pedestrian, hot spots,…) Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Classification of Mobility and Mobility Models Static (e.g., sensor networks) Uncontrolled Mobility Mobility Mobile Hybrid Controlled Mobility Unpredictable Mobility Predictable Mobility Hybrid I- Based on Controllability Hybrid Synthetic II- Based on Model Construction Model Trace-based Usage pattern Movement Pattern Hybrid Hybrid Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Mobility Dimensions & Classification of Synthetic Uncontrolled Mobility Models * F. Bai, A. Helmy, "A Survey of Mobility Modeling and Analysis in Wireles Adhoc Networks", Book Chapter in the book "Wireless Ad Hoc and Sensor Networks”, Kluwer Academic Publishers, June 2004. Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA I. Random Waypoint (RWP) Model 1. A node chooses a random destination anywhere in the network field 2. The node moves towards that destination with a velocity chosen randomly from [0, Vmax] 3. After reaching the destination, the node stops for a duration defined by the “pause time” parameter. 4. This procedure is repeated until the simulation ends – Parameters: Pause time T, max velocity Vmax – Comments: • • Speed decay problem, non-uniform node distribution Variants: random walk, random direction, smooth random, ... Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Random Way Point: Basics Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Random Way Point: Example Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA -1- RWP leads to non-uniform distribution of nodes due to bias towards the center of the area, due to non-uniform direction selection. To remedy this the “random direction” mobility model can be chosen. -2- Average speed decays over time due to nodes getting „stuck‟ at low speeds Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA II. Random (RWK) Walk Model • Similar to RWP but – – – – Nodes change their speed/direction every time slot New direction  is chosen randomly between (0,2] New speed chosen from uniform (or Gaussian) distribution When node reaches boundary it bounces back with (-) Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Random Walk Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA III. Reference Point Group Mobility (RPGM) • • • • Nodes are divided into groups Each group has a leader The leader‟s mobility follows random way point The members of the group follow the leader‟s mobility closely, with some deviation • Examples: – Group tours, conferences, museum visits – Emergency crews, rescue teams – Military divisions/platoons Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Group Mobility: Single Group Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Group Mobility: Multiple Groups Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA IV. Obstacle/Pathway Model • Obstacles/bldgs map • Nodes move on pathways between obstacles • Nodes may enter/exit buildings • Pathways constructed by computing Voronoi graph (i.e., pathways equidistant to nearby buildings) • Obstacles affect communication – Nodes on opposite sides (or in/outside) of a building cannot communicate Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA V. Related Real-world Mobility Scenarios • Pedestrian Mobility – University or business campuses – Usually mixes group and RWP models, with obstacles and pathways • Vehicular Mobility – Urban streets (Manhattan-like) – Freeways – Restricted to streets, involves driving rules Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Urban Street Streets - Manhattan Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Freeway Map Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Motivation • Randomized models (e.g., random waypoint) do not capture – (I) Existence of geographic restriction (obstacles) – (II) Temporal dependence of node movement Mobility (correlation over history) Space – (III) Spatial dependence (correlation) of movement among nodes Temporal Correlation Geographic Restriction Spatial Correlation • A systematic framework is needed to investigate the impact of various mobility models on the performance of different routing protocols for MANETs • This study attempts to answer – – – – What are key characteristics of the mobility space? Which metrics can compare mobility models in a meaningful way? Whether mobility matters? To what degree? If the answer is yes, why? How? Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA IMPORTANT: A framework to systematically analyze the "Impact of Mobility on Performance Of RouTing in Ad-hoc NeTworks" Fan Bai, Narayanan Sadagopan, Ahmed Helmy {fbai, nsadagop, helmy}@usc.edu website “http://nile.usc.edu/important” * F. Bai, N. Sadagopan, A. Helmy, "IMPORTANT: A framework to systematically analyze the Impact of Mobility on Performance of RouTing protocols for Adhoc NeTworks", IEEE INFOCOM, pp. 825-835, April 2003. * F. Bai, N. Sadagopan, A. Helmy, “The IMPORTANT Framework for Analyzing the Impact of Mobility on Performance of Routing for Ad Hoc Networks”AdHoc Networks Journal Elsevier Science, Vol. 1, Issue 4, pp. 383-403, November 2003. * F. Bai, A. Helmy, "The IMPORTANT Framework for Analyzing and Modeling the Impact of Mobility in Wireless Adhoc Networks", Book Chapter in the book "Wireless Ad Hoc and Sensor Networks”, Kluwer Academic Publishers, June 2004. UNIVERSITY OF SOUTHERN CALIFORNIA Framework Goals (Questions to Answer) • Whether mobility matters? and How much does it matter? – Rich set of mobility models that capture characteristics of different types of movement – Protocol independent metrics such as mobility metrics and connectivity graph metrics to capture the above characteristics • Why? – Analysis process to relate performance with a specific characteristic of mobility via connectivity metrics • How? – Systematic process to study the performance of protocol mechanistic building blocks (BRICS) across various mobility characteristics Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA The IMPORTANT Framework Overview Mobility Models Connectivity Graph Building Block Analysis Routing Protocol Performance DSR AODV DSDV GPSR GLS ZRP Random Waypoint Group Mobility Freeway Mobility Manhattan Mobility Contraction/Expansion Hybrid Trace-driven Mobility Metrics Relative Speed Spatial Dependence Temporal Dependence Node Degree/Clustering Connectivity Metrics Link Duration Path Duration Encounter Ratio Performance Metrics Flooding Caching Error Detection Error Notification Error Handling Throughput Overhead Success rate Wasted Bandwidth Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Mobility Metrics • Relative Speed (mobility metric I) – The magnitude of relative speed of two nodes, average over all neighborhood pairs and all time  1 T N N  R S   | v (i, t )  v ( j , t ) | P t 0 i 1 j 1 j i if dist(( xi , yi ), ( x j , y j ))  2 R • Spatial Dependence (mobility metric II) – The value of extent of similarity of the velocities/dir of two nodes that are not too far apart, average over all neighborhood pairs and all time     1 T N N min( v (i, t ), v ( j, t )) v (i, t )  v ( j, t ) Dspatial        P t 0 i 1 j 1 max( v (i, t ), v ( j, t )) | v (i, t ) || v ( j, t ) | j i if dist(( xi , yi ), ( x j , y j ))  2 R For example, RWP model, Vmax=30m/s, RS=12.6m/s, Dspatial=0.03 Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Connectivity Graph Metrics • Average link duration (connectivity metric I) – The value of link duration, average over all nodes pairs 1 N N L D   LD(i, j ) if there is a link between i and j P i 1 j 1 j i – Link/Path duration distributions (PATHS study) Protocol Performance Metrics • Throughput: delivery ratio • Overhead: number of routing control packets sent Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Mobility Models Summary Application Random Waypoint Model Group Mobility Model Freeway Mobility Model Manhattan Mobility Model Spatial Dependence Geographic Restriction General (uncorrelated straight lines) No No Conventions, Campus Metropolitan Traffic/Vehicular Urban Traffic/Vehicular Yes No Yes Yes No Yes Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Parameterized Mobility Models • Random Waypoint Model (RWP) – Each node chooses a random destination and moves towards it with a random velocity chosen from [0, Vmax]. After reaching the destination, the node stops for a duration defined by the “pause time” parameter. This procedure is repeated until simulation ends Parameters: Pause time T, max velocity Vmax Each group has a logical center (group leader) that determines the group’s motion behavior Each nodes within group has a speed and direction that is derived by randomly deviating from that of the group leader – – – • Reference Point Group Model (RPGM) member member Leader – –   | Vm em ber(t ) |  | Vleader(t ) |  random()  SDR Vmax  m em ber(t )  leader(t )  random()  ADR   max Parameters: Angle Deviation Ratio(ADR) and Speed Deviation Ratio(SDR), number of groups, max velocity Vmax. In our study, ADR=SDR=0.1 In our study, we use two scenarios: Single Group (SG) and Multiple Group (MG) Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Parameterized Mobility Models • Freeway Model (FW) – Each mobile node is restricted to its lane on the freeway – The velocity of mobile node is temporally dependent on its previous velocity – If two mobile nodes on the same freeway lane are within the Safety Distance (SD), the velocity of the following node cannot exceed the velocity of preceding node – Parameter: Map layout, Vmax Map for FW • Manhattan Model (MH) – Similar to Freeway model, but it allows node to make turns at each corner of street – Parameter: Map layout, Vmax Map for MH Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Experiment I: Analysis of mobility characteristics • IMPORTANT mobility tool – integrated with NS-2 (released Jan ‟04, Aug „05) – http://nile.usc.edu/important • Simulation done using our mobility generator and analyzer • • • • • • Number of nodes(N) = 40, Simulation Time(T) = 900 sec Area = 1000m x 1000m Vmax set to 1,5,10,20,30,40,50,60 m/sec across simulations RWP, pause time T=0 SG/MG, ADR=0.1, SDR=0.1 FW/MH, map layout in the previous slide Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Mobility metrics • Objective: – validate whether proposed mobility models span the mobility space we explore • Relative speed – For same Vmax, MH/FW is higher than RWP, which is higher than SG/MG Relative Speed • Spatial dependence – For SG/MG, strong degree of spatial dependence – For RWP/FW/MH, no obvious spatial dependence is observed Ahmed Helmy - USC Spatial Dependence UNIVERSITY OF SOUTHERN CALIFORNIA Connectivity Graph Metrics • Link duration – For same Vmax, SG/MG is higher than RWP, which is higher than FW, which is higher than MH Link duration • Summary – Freeway and Manhattan model exhibits a high relative speed – Spatial Dependence for group mobility is high, while it is low for random waypoint and other models – Link Duration for group mobility is higher than Freeway, Manhattan and random waypoint Path duration - Similar observations for Path duration Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Experiment II: Protocol Performance across Mobility Models Simulations done in ns-2: • Routing protocols: DSR, AODV, DSDV • Same set of mobility trace files used in experiment1 • Traffic pattern consists of source-destination pairs chosen at random • 20 source, 30 connections, CBR traffic • Data rate is 4packets/sec (low data rate to avoid congestion) • For each mobility trace file, we vary traffic patterns and run the simulation for 3 times Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Results and Observations • Performance of routing protocols may vary drastically across mobility patterns (Example for DSR) Throughput Routing Overhead • There is a difference of 40% for throughput and an order of magnitude difference for routing overhead across mobility models! Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Which Protocol Has the Highest Throughput ? • We observe that using different mobility models may alter the ranking of protocols in terms of the throughput! Random Waypoint : DSR Ahmed Helmy - USC Manhattan : AODV ! UNIVERSITY OF SOUTHERN CALIFORNIA Which Protocol Has the Lowest Overhead ? • We observe that using different mobility models may alter the ranking of protocols in terms of the routing overhead! RPGM(single group) : DSR Manhattan : DSDV • Recall: Whether mobility impacts protocol performance? • Conclusion: Mobility DOES matter, significantly, in evaluation of protocol performance and in comparison of various protocols! Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Putting the Pieces Together • Why does mobility affect protocol performance? • We observe a very clear trend between mobility metric, connectivity and performance – With similar average spatial dependency • Relative Speed increases Link Duration decreases Routing Overhead increases and throughput decreases – With similar average relative speed • Spatial Dependence increase Link Duration increasesThroughput increases and routing overhead decreases • Conclusion: Mobility Metrics influence Connectivity Metrics which in turn influence protocol performance metrics ! Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Relative Velocity Putting the Pieces Together Link Duration Throughput Spatial Dependence Path Duration Overhead Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Mechanistic Building Blocks (BRICS) • How does mobility affect the protocol performance? • Approach: * – The protocol is decomposed into its constituent mechanistic, parameterized building block, each implements a well-defined functionality – Various protocols choose different parameter settings for the same building block. For a specific mobility scenario, the building block with different parameters behaves differently, affecting the performance of the protocol • We are interested in the contribution of building blocks to the overall performance in the face of mobility • Case study: – Reactive protocols (e.g., DSR and AODV) * F. Bai, N. Sadagopan, A. Helmy, "BRICS: A Building-block approach for analyzing RoutIng protoCols in Ad Hoc Networks - A Case Study of Reactive Routing Protocols", IEEE International Conference on Communications (ICC), June 2004. Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Building Block Diagram for reactive protocols DSR Local Inquiry & Global Flooding Link Monitoring Error Notification Cache Management AODV Expanding Ring Search & Global Flooding Link Monitoring Error Broadcast Cache Management Salvaging (a) Generalization of Flooding (b) Generalization of Flooding Localized Rediscovery Generalization of Error Handling Route Setup Route Request Flooding Add Route Cache Caching Route Reply Range of Flooding Caching Style Expiration Timer Localized/Non-localized method Route Maintenance Error Detection Link Breaks Route Invalidate Error Handling Notify Error Notification Notify (c) Detection Method Handling Mode Ahmed Helmy - USC Recipient UNIVERSITY OF SOUTHERN CALIFORNIA How useful is caching? DSR AODV • • In RW, FW and MH model, most of route replies come from the cache, rather than destination (>80% for DSR, >60% for AODV in most cases) The difference in the route replies coming from cache between DSR and AODV is greater than 20% for all mobility models, maybe because of caching mode Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Is aggressive caching always good? DSR • • The invalid cached routes increase from RPGM to RW to FW to MH mobility models Aggressive Caching may have adverse effect at high mobility scenarios! Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Conclusions • Mobility patterns are very IMPORTANT in evaluating performance of ad hoc networks • A rich set of mobility models is needed for a good evaluation framework. • Richness of those models should be evaluated using quantitative mobility metrics. • Observation – In the previous study only „average‟ link duration was considered. – Are we missing something by looking only at averages? – Next: We conduct the PATHS study to investigate statistics and distribution of link and path duration. Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA PATHS: Analysis of PATH Duration Statistics and their Impact on Reactive MANET Routing Protocols Fan Bai, Narayanan Sadagopan, Bhaskar Krishnamachari, Ahmed Helmy {fbai, nsadagop, brksihna, helmy}@usc.edu * F. Bai, N. Sadagopan, B. Krishnamachari, A. Helmy, "Modeling Path Duration Distributions in MANETs and their Impact on Routing Performance", IEEE Journal on Selected Areas in Communications (JSAC), Special Issue on Quality of Service in Variable Topology Networks, Vol. 22, No. 7, pp. 1357-1373, Sept 2004. •N. Sadagopan, F. Bai, B. Krishnamachari, A. Helmy, "PATHS: analysis of PATH duration Statistics and their impact on reactive MANET routing protocols", ACM MobiHoc, pp. 245-256, June 2003. UNIVERSITY OF SOUTHERN CALIFORNIA Motivation and Goal • Mobility affects connectivity (i.e., links), and in turn protocol mechanisms and performance • It is essential to understanding effects of mobility on Protocol Mechanisms link and path characteristics Performance Mobility Connectivity (Throughput, • In this study: Overhead) – Closer look at the mobility effects on connectivity metrics (statistics of link duration (LD) and path duration (PD)) – Develop approximate expressions for LD & PD distributions (Is it really exponential? When is it exponential?) – Develop first order models for Tput & Overhead as f(PD) Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Connectivity Metrics • Link Duration (LD): – For nodes i,j, the duration of link i-j is the longest interval in which i & j are directly connected – LD(i,j,t1)=t2-t1 • iff t, t1  t  t2,   > 0 : X(i,j,t)=1,X(i,j,t1-)=0, X(i,j,t2+)=0 • Path Duration (PD): – Duration of path P={n1,n2,…,nk} is the longest interval in which all k-1 links exist Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Simulation Scenarios in NS-2 • Path duration computed for the shortest path, at the graph and protocol levels, until it breaks. • Used the IMPORTANT mobility tool: – nile.usc.edu/important • Mobility Parameters – Vmax = 1,5,10,20,30,40,50,60 m/s, – RPGM: 4 groups (RPGM4), Speed/Angle Deviation Ratio=0.1 • 40 nodes, in 1000mx1000m area • Radio range (R)=50,100,150,200,250m • Simulation time 900sec Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Link Duration (LD) PDFs • At low speeds (Vmax < 10m/s) link duration has multi-modal distribution for FW and RPGM4 – In FW due to geographic restriction of the map • Nodes moving in same direction have high link duration • Nodes moving in opposite directions have low link duration – In RPGM4 due to correlated node movement • Nodes in same group have high link duration • Nodes in different groups have low link duration • At higher speeds (Vmax > 10m/s) link duration does not exhibit multi-modal distribution • Link duration distribution is NOT exponential Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Nodes moving in opposite directions FW model Vmax=5m/s R=250m RPGM w/ 4 groups Vmax=5m/s R=250m Nodes in different groups Nodes in the same group Nodes moving in the same direction/lane Multi-modal Distribution of Link Duration for Freeway model at low speeds Multi-modal Distribution of Link Duration for RPGM4 model at low speeds Link Duration (LD) distribution at low speeds < 10m/s Ahmed Helmy - USC UNIVERSITY OF RW SOUTHERN CALIFORNIA RPGM (4 groups) FW Vmax=30m/s R=250m Link Duration at high speeds > 10m/s Not Exponential !! Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Path Duration (PD) PDFs • At low speeds (Vmax < 10m/s) and for short paths (h2) path duration has multi-modal for FW and RPGM4 • At higher speeds (Vmax > 10m/s) and longer path length (h2) path duration can be reasonably approximated using exponential distribution for RW, FW, MH, RPGM4. Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Nodes moving in opposite directions Nodes in different groups RPGM4 Vmax=5m/s h=2 hops R=250m FW Vmax=5m/s h=1 hop R=250m Nodes moving in the same direction Nodes in the same group Multi-modal Distribution of Path Duration Multi-modal Distribution of Path Duration for Freeway model at low speeds, low hops for RPGM4 model at low speeds, low hops Path Duration (PD) distribution for short paths at low speeds < 10m/s Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA RW h=2 RPGM4 h=4 100 FW h=4 Vmax=30m/s R=250m Path Duration (PD) distribution for long paths ( 2 hops) at high speeds (> 10m/s) Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Exponential Model for Path Duration (PD) • Let path be the parameter for exponential PD distribution: – PD PDF f(x)= path e- path x – As path increases average PD decreases (and vice versa) • Intuitive qualitative analysis: – – – – PD=f(V,h,R); V is relative velocity, h is path hops & R is radio range As V increases, average PD decreases, i.e., path increases As h increases, average PD decreases, i.e., path increases As R increases, average PD increases, i.e., path decreases • Validate intuition through simulations Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Exponential Model for PD But, PD PDF f(x)= path e- path x Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA 0.5 0.1 Probability Exponential PD Probability RW h=2 Exponential 0.4 0.3 0.2 0.1 PD FW h=4 0.05 0 0 10 20 30 40 50 Path Duration (sec) 0 0 10 Path Duration (sec) 20 - Correlation: 94.1-99.8% - Goodness-of-fit Test RW FW RPGM K-S test 0.04-0.065 0.045-0.085 0.09-0.12 Probability 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 2 Exponential D= 0.048 PD Vmax=30m/s R=250m FW h=4 4 6 8 10 12 14 16 18 20 22 24 26 28 30 Cumulative Distribution Function (CDF) Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Effect of Path Duration (PD) on Performance: Case Study for DSR • PD observed to have significant effect on performance • (I) Throughput: First order model – T: simulation time, D: data transferred, Tflow: data transfer time, Trepair: total path repair time, trepair: av. path repair time, f: path break frequency Throughput  D T  1 T  T flow  Trepair  T flow  trepair. f .T  T flow  trepair. .T PD trepair D Throughput  (1  )  (1  ).rate PD T flow PD trepair  T T flow trepair (1  ) PD 1 ) PD Throughput ( Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Effect of PD on Performance (contd.) • (II) Overhead: First order model – Number of DSR route requests= PD – p: non-propagating cache hit ratio, N: number of nodes  T Overhead 1 PD • Evaluation through NS-2 simulations for DSR Throughput Overhead Random Waypoint (RW) Freeway (FW) -0.9165 -0.9597 0.9753 0.9812 Manhattan (MH) -0.9132 0.9978 Pearson coefficient of correlation () with 1 PD – RPGM exhibits low , due to relatively low path changes/route requests Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Conclusions • Detailed statistical analysis of link and path duration for multiple mobility models (RW,FW,MH,RPGM4): – Link Duration: multi-modal FW and RPGM4 at low speeds – Path Duration PDF: • Multi-modal FW and RPGM4 at low speeds and hop count • Exponential-like at high speeds & med/high hop count for all models • Developed parametrized exponential model for PD PDF, as function of relative velocity V, hop count h and radio range R • Proposed simple analytical models for throughput & overhead that show strong correlation with reciprocal of average PD • Open Issues: – Can we prove this mathematically? Yes – Is it general for random and correlated mobility? Yes Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Case Studies Utilizing Mobility Modeling Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Case Study on Effects of Mobility on the Grid Location Service (GLS) • Group mobility: - prolongs protocol convergence - incurs max overhead - incurs max query failure rate * Subtle Coupling between – (1) Mobility – (2) The Grid Topology – (3) Protocol Mechanisms 100 90 80 70 60 50 40 30 20 10 0 100 90 Percentage Failed Queries Percentage Overhead Manhattan Freeway Group Mobility RWP Models Manhattan Freeway Group Mobility RWP 80 70 60 50 40 30 20 10 0 Model * C. Shete, S. Sawhney, S. Herwadka, V. Mehandru, A. Helmy, "Analysis of the Effects of Mobility on the Grid Location Service in Ad Hoc Networks", IEEE ICC, June 2004. UNIVERSITY OF SOUTHERN CALIFORNIA Case Study on Geo-routing across Mobility Models • Depending on beacon frequency location info may be out of date • Nodes chosen by geographic routing may move out of range before next beacon update. • Increasing beacon updates does not always help! • Using simple mobility prediction achieved up to 37% saving in wasted bandwidth, 27% delivery rate 700 w/o MP w/o NLP w/ MP(NLP+DLP) 1 GPSR D e liv e ry R a te (% ) 0.9 0.8 0.7 0.6 0.5 0.4 0.3 GPSR with prediction 600 N u mb e r o f p a cke t d ro p s 500 400 300 200 GPSR with prediction 1 1.5 3 Beacon Interval (sec) 6 0.2 0.1 0 10 w/o MP w/o NLP w/ MP(NLP+DLP) GPSR 20 30 Max Node Speed (m/sec) 40 50 100 0 0.250.5 (FWY) * D. Son, A. Helmy, B. Krishnamachari, "The Effect of Mobility-induced Location Errors on Geographic Routing in Ad Hoc Networks: Analysis and Improvement using Mobility Prediction", IEEE WCNC, March 2004, and IEEE Transactions on Mobile Computing, Special Issue on Mobile Sensor Networks, 3rd quarter 2004 (to appear). UNIVERSITY OF SOUTHERN CALIFORNIA Contraction, Expansion and Hybrid Models • May be useful for sensor networks • Contraction models show „improved‟ performance (e.g., Tput, link duration) with increased velocity Expansion Contraction Hybrid * Y. Lu, H. Lin, Y. Gu, A. Helmy, "Towards Mobility-Rich Performance Analysis of Routing Protocols in Ad Hoc Networks: Using Contraction, Expansion and Hybrid Models", IEEE ICC, June 2004. UNIVERSITY OF SOUTHERN CALIFORNIA MAID Case Study: Utilizing Mobility • MAID: Mobility Assisted Information Diffusion • May be used for: resource discovery, routing, node location applications • MAID uses „encounter‟ history to create time (or age) gradients towards the target/destination • MAID uses (and depends on) mobility to diffuse information, hence its performance may be quite sensitive to mobility degree and patterns • Unlike conventional adhoc routing, link/path duration may not be the proper metrics to analyze • The „Age gradient tree‟ and its characteristics determine MAID‟s performance * F. Bai, A. Helmy, "Impact of Mobility on Mobility-Assisted Information Diffusion (MAID) Protocols", USC Technical Rreport, July 2005. [Submitted for review] Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Time: t1 Location: x1,y1 S A Time: t3 Location: x3,y3 E D B C Time: t2 Location: x2,y2 Time: t4 Location: x4,y4 F Basic Operation of MAID: Encounter history, search and age gradient tree Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA MAID protocol phases and metrics • Cold cache (initial, transient, phase): – Encounter cache is empty – More encounters „warm up‟ the cache by increasing the entries • Warm cache (steady state phase) : – Average encounter ratio reaches ~30% of network nodes – Age gradient trees are established • Metrics: – Warm up time – Average path length to a destination – Cost of search to establish the route to the destination Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Warm Up Phase The Warm Up Time depends heavily on the Mobility model and the Velocity Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Steady State Phase Steady State Performance depends only on the Mobility model but NOT on the Velocity - These metrics reflect the structure of the age-gradient trees (AGTs). - Hence, MAID leads to stable characteristics of the AGTs. Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Spatio-Temporal Correlations in the AGT RWK 400 nodes 3000mx3000m area Radio range 250m V=10m/s RWP RPGM (80grps) Ahmed Helmy - USC MH UNIVERSITY OF SOUTHERN CALIFORNIA RWK RWP V=30m/s RPGM (80grps) Ahmed Helmy - USC MH UNIVERSITY OF SOUTHERN CALIFORNIA RWK RWP V=50m/s RPGM (80grps) Ahmed Helmy - USC MH UNIVERSITY OF SOUTHERN CALIFORNIA Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA On-going and Future Work • Extend the IMPORTANT mobility tool: – URL: http://nile.usc.edu/important • Trace-based mobility models – Pedestrians on campus • Usage pattern (WLAN traces) – USC, MIT, UCSD, Dartmouth,… nile.usc.edu/MobiLib • Student tracing (survey, observe) – Vehicular mobility • Transportation literature – Parametrized hybrid models • Integrate Weighted Group mobility with Pathway/Obstacle Model • Derive the parameters based on the traces Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Trace-based Mobility Modeling Total Population: ~ 25,000 students Wireless Users: ~6000 students Access Points: ~400 Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA IMPACT: Investigation of Mobile-user Patterns Across University Campuses using WLAN Trace Analysis* • Classes of future wireless ad hoc and sensor networks will be attached to humans • What kinds of correlations exist between wireless users? • Analyze measurements of wireless networks – Understand Wireless Users Behavior (individual and group) – Develop models to understand associations and friendship • Study of relationships and user behavior based on measurements of various University WLANs * W. Hsu, A. Helmy, “IMPACT: Investigation of Mobile-user Patterns Across University Campuses using WLAN Trace Analysis”, USC TR, July „05 (Under Submission) Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Statistics of Studied Traces - Four major campuses - Month long traces studied - Total users in the study: over 12,000 users - Total Access Points in the study: over 1,300 Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Observations: On-line Time On-off behavior is very common for wireless users. This seems especially true for small handheld devices. There are clear categories of heavy and light users, the distribution of which is skewed and heavily depends on the campus. Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Observations: Visited Access Points (APs) [percentage of visited APs] •Individual users access only a very small portion of APs in the network, less than 35% in all campuses. The long-term mobility of users is highly skewed in terms of time associated with each AP. On average a user spends more than 95% of time at its top five most visited APs. Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Observations: Visited APs •The majority of users experience low mobility while using the network. This is even true for portable devices such as PDAs. The actual handoff statistics depend heavily on the environment. Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Observations: Similarity Index •We observe clear repetitive patterns of association in wireless network users. Typically, user association patterns show the strongest repetitive pattern at time gap of one day/one week. Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Observations: Encounters •In all the traces, the MNs encounter a relatively small fraction of the user population; below 40% in most cases and never reaching above 60% in any case. Except for UCSD trace, on average a MN only encounters 1.88%-5.94% of the whole population. The number of total encounters for the users follows a BiPareto distribution, the parameters of which depends on the campus. Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Encounter-graphs • Definition – When 2 nodes access the same AP at the same time we call this an „encounter‟ – The encounter graph has all the mobile nodes as vertices and its edges link all those vertices that encounter each other Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Regular Graph - High path length - High clustering 1 0.8 0.6 0.4 0.2 Small World Graph: Low path length, High clustering Random Graph - Low path length, - Low clustering Clustering Path Length 0 0.0001 0.001 0.01 0.1 1 probability of re-wiring (p) - In Small Worlds, a few short cuts contract the diameter (i.e., path length) of a regular graph to resemble diameter of a random graph without affecting the graph structure (i.e., clustering) UNIVERSITY OF SOUTHERN CALIFORNIA Encounter-graphs and Friendship • Encounters link most of the MNs together in a connected graph: – Albeit each MN encounters only with small portion of the population. – The encounter graph is a SmallWorld graph – Even for short time period (1 day) its clustering coefficent, average path length, and connectivity are all close to those for longer traces. • Friendship between MNs is highly asymmetric. – The distribution for the friendship index is exponential for all the traces, regardless of the friendship definition (based on time, encouner, or location). – Among all node pairs there are less than 5% with friendship index larger than 0.01, and less than 1% with friendship index larger than 0.4. Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Encounter-graphs using Friends •Top-ranked friends tend to form cliques and low-ranked friends are the key to provide random links and reduce the degree of separation in encounter graph. Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Encounter-based Information Diffusion •Encounters patterns are rich enough to support information diffusion. Specifically, information can be delivered to more than 94% of users within two days. The reachability and average delay do not decrease significantly until at least ~40% of nodes are selfish. Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Vision: Building Community-wide Wireless/Mobility Library • Library of measurements from WLANs, mobility and associations from potential wireless societies (e.g., universities, vehicular nets) • Library of realistic models of user behavior (e.g., mobility, traffic, friendship, encounter models, … ) • Library of benchmarks and guidelines for simulation and evaluation • How much insight can we get by analyzing the traces? • Can we use the insight to „design‟ protocols of the future (not only for evaluation)? • Currently 20+ major universities willing to share their traces • …. more to come: http://nile.usc.edu/MobiLib • If you have traces: helmy@usc.edu ! Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Network Usage vs. Mobility Wireless Network (WLAN) Usage Traces • Collect measurements of network access patterns for WLAN users at various locations/buildings on campus Draw map and join the buildings via shortest pathways to approximate user movement routes Estimate transition probability from one location to another at a given time slot KOH 1 • • Number of Access Points (AP) Number of Buildings with APs Number of Registered Users Number of Users in Trace • 200 44 5250 4576 JEP 2 LVL 3 6 5 PED OHE C. Jr. 7 TOMMY 4 Tracers trap MAC addresses accessing the WLAN - Building level granularity Wireless Network Coverage Map at USC - main campus Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA USC MAP WITH OBSERVATION LOCATIONS Observation Location (OL) Type (1) KOH Computational/Residential (2) JEP Residential/Library route (3) LVL Library (4) TOMMY Center of Campus (5) PED Classes (6) OHE Classes (7) Carl's Jr. Cafeteria 6 KOH 1 JEP 2 LVL 3 PED 5 OHE TOMMY C. Jr. 7 4 At1 Bt1 Ct1 At2 10:00-10:15 Bt2 Ct2 10:45-11:00 10:15-10:30 Statistics about recorded mobility traces used in this study 10:30-10:45 Trace Period Feb 25 - April 25 Number of Persons in the trace Number of Observers 60 Number of Groups Observed Total Observation Hours 220 Number of Subgroups 6389 1758 2382 Partial Recorded Data and example Observation Location (OL) Date Time Group Size Direction of Group Subgroup Size(s) Direction of Subgroup Olin Hall (OHE) OL6 12-Mar 10:03 AM 3 SE 2 SE 1 NE Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Distributions 1000 #of people 800 observed at 600 400 200 LVL JEP KOH OHE 0 TOMMY PED Observations vs. WLAN traces # of access S4 time slots S1 Series1 Series2 Series3 Series4 Observation Location CARLS JR. Observation traces • • WLAN access traces Observation traces exhibit drastically different trends than WLAN traces The two traces include different parts of the student population – – – WLAN users tend to cluster around base stations WLAN users exhibit on-off behavior (sit-down, turn on laptop, access wireless network, turn off, then move). Seldom did users access the WLAN when mobile Observation traces trace actual mobility instead of network access patterns • Mobility models based on network access traces may not reflect actual mobility of the users Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Survey based: Weighted Way Point (WWP) Model 0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 Classroom classroom Off-campus Other area 61-120 121-240 on campus pause time (m) probability Library cafeteria 0-30 31-60 > 240 0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 0-30 31-60 61-120 Other 0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 Library probability 121-240 > 240 probability 0-30 31-60 61-120 pause time (m) 121-240 > 240 pause time (m) Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Mobility Simulation Tools • The Network Simulator (NS-2) (USC/ISI, UCB, Xerox Parc) [wireless extensions CMU/Rice] – www.isi.edu/nsnam • The GloMoSim Simulator (UCLA)/QualNet (Commercial) • The IMPORTANT Mobility Tool (USC) – nile.usc.edu/important • The Obstacle Mobility simulator (UCSB) – moment.cs.ucsb.edu/mobility • The CORSIM Simulator • OPNET (commercial) Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA IMPORTANT • Includes: – Mobility generator tools for FWY, MH, RPGM, RWP, RWK (future release), City Section (future rel.) – Acts as a pre-processing phase for simulations, currently supports NS-2 formats (can extend to other formats) – Analysis tools for mobility metrics (link duration, path duration) and protocol performance [future rel.] (throughput, overhead, age gradient tree chars) – Acts as post-processing phase of simulations – nile.usc.edu/important Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Manhattan IMPORTANT Freeway Group Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA CORSIM (Corridor Traffic Simulator) • Simulates vehicles on highways/streets • Micro-level traffic simulator – Simulates intersections, traffic lights, turns, etc. – Simulates various types of cars (trucks, regular) – Used mainly in transportation literature (and recently for vehicular networks) – Does not incorporate communication or protocols – Developed through FHWA (federal highway administration) http://ops.fhwa.dot.gov – Need to buy license Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA CORSIM Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA – Protocol design for controlled mobility scenarios – Examples: DakNet, Message Ferries, Info Station! On-going and Future Directions Utilizing mobility – Mobility-Assisted protocols – A. Helmy, "Mobility-Assisted Resolution of Queries in Large-Scale Mobile Sensor Networks (MARQ)", Computer Networks Journal, Vol. 43, Issue 4, pp. 437-458, Nov03 – Context-aware Networks • Mobility-aware protocols: self-configuring, mobility-adaptive protocols • Socially-aware networks: security, trust, friendship, associations, small worlds contact 1 R 2 7 contact contact 3 C R 5 4 R – Experiments: • Boundless Classroom [Next generation education paradigm?!] • E-buddy system for campus security/safety Ahmed Helmy - USC 6 R Route UNIVERSITY OF SOUTHERN CALIFORNIA Related Links and Resources • Delay Tolerant Networks (DTNs) – Research group: www.dtnrg.org • Vehicular/Transportation Networks – PATH project/center: www.path.berkeley.edu – METRANS center: www.metrans.org (USC, CSULB) • Mesh Networks – Microsoft research http://research.microsoft.com/mesh/ – MIT RoofNet http://pdos.csail.mit.edu/roofnet – GATech Message Ferries http://www.cc.gatech.edu/fac/Mostafa.Ammar/ferrying.html – UIUC, Rice, …. and others (chk nile.usc.edu/MobiLib) Ahmed Helmy - USC UNIVERSITY OF SOUTHERN CALIFORNIA Thank You ! • Ahmed Helmy • Webpage: ceng.usc.edu/~helmy • Lab: nile.usc.edu • IMPORTANT Mobility tool: nile.usc.edu/important • Wireless/Mobile Network Library of Traces: nile.usc.edu/MobiLib Ahmed Helmy - USC

Related docs
Ad-hoc Simulation
Views: 19  |  Downloads: 0
Tutorial on Ad-Hoc Networking_3_
Views: 8  |  Downloads: 1
Tutorial on Ad-Hoc Networking_2_
Views: 2  |  Downloads: 1
Data Modeling And Analysis
Views: 345  |  Downloads: 58
The Alchemy Tutorial
Views: 85  |  Downloads: 6
Taguchi Design Tutorial
Views: 72  |  Downloads: 7
Other docs by techmaster
2006 Tax Form 8398
Views: 68  |  Downloads: 0
IRS Publication _1544
Views: 76  |  Downloads: 1
2006 Tax Form 4461
Views: 81  |  Downloads: 3
IRS Publication _3202
Views: 102  |  Downloads: 1
2006 Tax Form 8316
Views: 56  |  Downloads: 0
IRS Publication _1460
Views: 137  |  Downloads: 1
Small Business Loan Application Checklist
Views: 283  |  Downloads: 13
hadith_2
Views: 263  |  Downloads: 8
2006 Tax Form 3520A
Views: 76  |  Downloads: 3
IRS Publication _3809
Views: 123  |  Downloads: 1
2006 IRS Form 1040-PR
Views: 440  |  Downloads: 3
2006 Tax Form 8586
Views: 41  |  Downloads: 0
2006 Tax Form 6045
Views: 71  |  Downloads: 1
2006 Tax Form 8596A
Views: 62  |  Downloads: 0
2006 Tax Form 8027
Views: 120  |  Downloads: 0