Weather Disruption-Tolerant Self-Optimising Millimeter Mesh Networks: Architecture, Routing Protocols, Performance

Transcription

1 Weather Disruption-Tolerant Self-Optimising Millimeter Mesh Networks: Architecture, Routing Protocols, Performance James P.G. Sterbenz, Abdul Jabbar, Justin Rohrer, Egemen Çetinkaya, Bharatwajan Raman, Victor Frost Department of Electrical Engineering & Computer Science Information Technology & Telecommunications Research Center The University of Kansas April Sterbenz

2 WDTN Mesh Networks Abstract With growing demand for high-speed access to mobile handheld devices, there is a significant cost benefit in deploying fixed wireless-mesh networks for backhaul access. However, enabling reliable broadband access over high-frequency radios (such as millimeter-wave networks) posses a fundamental challenge due to weather disruptions in general and rain attenuation in particular. In this paper, we present an analysis of the impact of precipitation on millimeter-wave mesh networks based on radar measurements of real storms in the Midwest US. Furthermore, we compare two novel algorithms that use physical-layer information to optimize routing at the network layer: P-WARP (Predictive Weather-Assisted Routing Protocol) and XL-OSPF (Cross-Layered Open Shortest Path First). Finally, we present simulation studies to compare the performance of the proposed protocols and evaluate the dependability of the end-user service during weather disruptions. 22 April 2009 Weather Disruption-Tolerant Mesh Networks 2

3 WDTN Mesh Networks Outline Introduction and motivation Millimeter-wave mesh networks Impact of weather disruptions WDTN algorithms Simulation model and performance analysis 22 April 2009 Weather Disruption-Tolerant Mesh Networks 3

4 Motivation Wireless vs. Fiber Optic Links Wireless data access for untethered network access supported by wired backbone fiber optic cables provide high-speed reliable connections increasingly wireless access to an optical core (MAN+WAN ) 22 April 2009 Weather Disruption-Tolerant Mesh Networks 4

5 Motivation Wireless vs. Fiber Optic Links Wireless data access for untethered network access supported by wired backbone fiber optic cables provide high-speed reliable connections increasingly wireless access to an optical core (MAN+WAN ) Deployment barriers to fiber extremely expensive: ~ $100K/mile lack of extensive market: rural regions or construction and regulatory issues: metropolitan cities barrier to new service providers expensive to lease competitor s capacity for backhaul 22 April 2009 Weather Disruption-Tolerant Mesh Networks 5

6 Motivation Broadband Wireless Links Wireless alternative to traditional fiber-optic links point-to-point wired link replacement in MANs and WANs mesh between base stations and cell towers Potential use backhaul of 3G (and eventually 4G) data Internet expansion to new areas, such as rural alternative to add capacity in congested areas, such as cities 22 April 2009 Weather Disruption-Tolerant Mesh Networks 6

7 Motivation Millimeter-Wave Wireless Links Millimeter-wave communication links exploit recently available commercial radios frequency band: GHz lightly licensed by FCC in US higher data rate potential than microwave links very-high data rate: 1 10 Gb/s potential replacement for 1/10 GbE and OC-14/192 significantly cheaper to deploy than fiber 22 April 2009 Weather Disruption-Tolerant Mesh Networks 7

8 Motivation Millimeter-Wave Wireless Links Disadvantages of millimeter-wave links highly susceptible to weather significant rain-based attenuation unreliable high-speed links do not meet carrier requirements for backhaul and distribution reliability, delay, 50 ms restoration 22 April 2009 Weather Disruption-Tolerant Mesh Networks 8

9 Motivation Weather Disruption-Tolerant Routing Problem: slow recovery from rain-based attenuation many frames lost before detection no inherent restoration equivalent to SONET/SDH APS Proposed solution: compensate at network layer new routing mechanisms predictive nearly-instantaneous reactive exploit weather radar information permits short-term prediction weather dynamics longer time scale than network control 22 April 2009 Weather Disruption-Tolerant Mesh Networks 9

10 WDTN Mesh Networks Millimeter-Wave Mesh Networks Introduction and motivation Millimeter-wave mesh networks Impact of weather disruptions WDTN algorithms Simulation model and performance analysis 22 April 2009 Weather Disruption-Tolerant Mesh Networks 10

11 Millimeter-Wave Mesh Networks Architecture Mesh architecture high degree of connectivity G CO/POP 22 April 2009 Weather Disruption-Tolerant Mesh Networks 11

12 Millimeter-Wave Mesh Networks Architecture Mesh architecture high degree of connectivity alternate diverse paths mm wave link to CO/POP alternate mm links G CO/POP 22 April 2009 Weather Disruption-Tolerant Mesh Networks 12

13 Millimeter-Wave Mesh Networks Architecture Mesh architecture high degree of connectivity alternate diverse paths severely attenuated mm wave alternate mm links alternate lower-freq. RF fiber bypass (competitor) Proposed solution route around failures before they occur avoid high error links CO/POP G 22 April 2009 Weather Disruption-Tolerant Mesh Networks 13

14 Millimeter-Wave Mesh Networks Open Research Questions Impact of weather on mesh network correlated link failures, link availability Actual weather pattern storm frequency and intensity geographic coverage with respect to mesh network Feasibility is this approach feasible? field measurements on deployed infrastructure 22 April 2009 Weather Disruption-Tolerant Mesh Networks 14

15 WDTN Mesh Networks Impact of weather disruptions Introduction and motivation Millimeter-Wave Mesh Networks Impact of weather disruptions WDTN algorithms Simulation model and performance analysis 22 April 2009 Weather Disruption-Tolerant Mesh Networks 15

16 Impact of Weather Disruptions Link Impact of weather on individual links ITU-R P.530 and Crane models relates link attenuation to a homogenous rain rate long term statistics of precipitation probabilities cannot predict availability easily of a particular link does not address correlated link failures of a mesh network not useful for tolerance of specific events on specific links 22 April 2009 Weather Disruption-Tolerant Mesh Networks 16

17 Impact of Weather Disruptions Approach Evaluate impact of real storms on actual mesh links collect data on observed storms at a given location translate storm data to link attenuation using geometric model of actual storm characterise link behavior and availability during an event 22 April 2009 Weather Disruption-Tolerant Mesh Networks 17

18 Impact of Weather Disruptions Geometric Storm Modeling Storm model is abstraction of precipitation intensity cells modelled as ellipses moving along a trajectory links modelled as line segments effective attenuation calculated based on link & cell overlap C green C green red red R 2 R 3 A red yello w B A R 1 R s yellow B red R 4 22 April 2009 Weather Disruption-Tolerant Mesh Networks 18

19 Impact of Weather Disruptions Actual Observations Radar reflectivity data from national weather service Evaluate effect of real storms on potential networks Collected and analyzed data ~ 30 storms over the Midwest U.S 3 hypothetical mesh networks in the same region 22 April 2009 Weather Disruption-Tolerant Mesh Networks 19

20 Impact of Weather Disruptions Observed Storm 1 Rain Distribution Millimeter-wave grid location N, W Storm observed at: 20:39:26Z 30 Sep April 2009 Weather Disruption-Tolerant Mesh Networks 20

21 Impact of Weather Disruptions Observed Storm 2 Rain Distribution Millimeter-wave grid location N, W Storm observed at: 05:04:11Z 22 Apr April 2009 Weather Disruption-Tolerant Mesh Networks 21

22 Impact of Weather Disruptions Link Characterization Channel error rate for links of a 4 4 grid 10 sec intervals few links severely degraded large number either partially degraded or normal 22 April 2009 Weather Disruption-Tolerant Mesh Networks 22

23 Impact of Weather Disruptions Link Availability Link availability after FEC Reed Solomon (204,188) Link states three states 1: BER < : < BER < : BER > Modelled as Markov process 22 April 2009 Weather Disruption-Tolerant Mesh Networks 23

24 Impact of Weather Disruptions Link State Characteristics Link state characteristics state transition probability matrix state probabilities: probability of being in each state at given time State State Probability April 2009 Weather Disruption-Tolerant Mesh Networks 24

25 WDTN Mesh Networks WDTN Algorithms Introduction and motivation Millimeter-wave mesh networks Impact of weather disruptions WDTN algorithms Simulation model and performance analysis 22 April 2009 Weather Disruption-Tolerant Mesh Networks 25

26 Weather Disruption-Tolerant Routing Alternative Algorithm Types Reactive frames in transit are lost time to converge on new route Predictive time to detect frame loss > 50 ms restoration use weather radar to react instantaneously re-route traffic before link failure use weather radar to predict the link condition advance warning in the order of minutes is sufficient 22 April 2009 Weather Disruption-Tolerant Mesh Networks 26

27 Weather Disruption-Tolerant Routing Predictive Routing Concept April 2009 Weather Disruption-Tolerant Mesh Networks 27

28 Weather Disruption-Tolerant Routing New Algorithms P-WARP: predictive weather-assisted routing prot. uses weather-radar data to forecast future link conditions precipitation modelled as {none/light, moderate, heavy} effective BER calculated and used to adjust link cost XL-OSPF: cross-layered OSPF uses radar to instantaneously estimate attenuation conventional OSPF but without dead interval detection 22 April 2009 Weather Disruption-Tolerant Mesh Networks 28

29 Weather Disruption-Tolerant Routing P-WARP Predicting link conditions predictive routing algorithm using weather radar data effective BER is calculated from radar reflectivity data modelled in real-time using the geometric model processing done at a core or a subset of nodes edge nodes with external (Internet) access Cost metric per link cost calculated as C ij = P BER ij γ C ij = cost of link i j P = average frame size BER ij = predicted BER γ = sensitivity factor 22 April 2009 Weather Disruption-Tolerant Mesh Networks 29

30 Weather Disruption-Tolerant Routing P-WARP Link status updates WLSUs: weather-based link status updates contains the predicted cost of all links (incremental) WLSUs are flooded in to the network by core nodes single update for all links generated only when one or more link costs change significantly reduces protocol overhead 22 April 2009 Weather Disruption-Tolerant Mesh Networks 30

31 Weather Disruption-Tolerant Routing P-WARP Route recomputation nodes recompute routes using shortest paths first algorithm individual nodes do not generate separate LSAs network reroutes traffic ahead of the disruption weather predictions in the order of seconds are sufficient route reconvergence is sub second Route sensitivity route flaps avoided using thresholds and hysteresis changes in BER below BER thresh are ignored H thresh is the minimum change in cost that triggers an update 22 April 2009 Weather Disruption-Tolerant Mesh Networks 31

32 Weather Disruption-Tolerant Routing XL-OSPF Standard OSPF dynamic link state algorithm Reacts too slowly causing end-to-end packet loss link costs do not reflect physical link status dead interval needed to detect failed links route convergence is slow 22 April 2009 Weather Disruption-Tolerant Mesh Networks 32

33 Weather Disruption-Tolerant Routing XL-OSPF Cross-layered OSPF cost metric is proportional to bit error rate several mechanisms exist to achieve this reactively e.g. packet error estimation at the receiver could use current weather data to calculate link BER Route computation nodes aware of the quality of all their links LSAs from other nodes give the complete network status shortest paths calculated based on link metric reroute traffic reactively 22 April 2009 Weather Disruption-Tolerant Mesh Networks 33

34 Weather Disruption-Tolerant Routing XL-OSPF Cost metric (same as P-WARP) per link cost calculated as C ij = cost of link ij, P = average frame size BER ij = predicted BER, γ = sensitivity factor Compared to P-WARP differs in the mechanism to calculate link costs reactive instead of predictive C ij = P BER ij γ higher overhead, generates per link LSAs 22 April 2009 Weather Disruption-Tolerant Mesh Networks 34

35 WDTN Mesh Networks Simulation Model and Performance Analysis Introduction and motivation Millimeter-wave mesh networks Impact of weather disruptions WDTN algorithms Simulation model and performance analysis 22 April 2009 Weather Disruption-Tolerant Mesh Networks 35

36 ns-2 Simulations Simulation Model and Parameters 16 node mesh: 4 4 grid two corner sink nodes connected to Internet (0, 15) other 14 nodes generate traffic to randomly chosen sink 2.4 Mb/s CBR over UDP Several synthetic storms we will look at one example Several actual storms we will look at one representative example 22 April 2009 Weather Disruption-Tolerant Mesh Networks 36

37 Synthetic storm outer ellipse: ~ Simulations Sample Storms four inner ellipses: ~ 5 10 Actual storm storm 1 from slide April 2009 Weather Disruption-Tolerant Mesh Networks 37

38 Simulations Synthetic Storm Overlaid on Lawrence, KS D 2 low intensity rain heavy intensity rain D 1 22 April 2009 Weather Disruption-Tolerant Mesh Networks 38

39 Synthetic Storm Performance Analysis: Packet Loss node out 40s OSPF dead interval 10s XL-OSPF HELLO interval link out 22 April 2009 Weather Disruption-Tolerant Mesh Networks 39

40 WDTN Routing Performance Analysis: Packet Loss 100% packet delivery ration with no precipitation Link attenuation: P-WARP reroutes predicatively with no loss XL-OSPF reroutes with minimal loss after HELLO interval conventional OSPF must wait for loss detection and recovery Node out attenuation of all links to given node transit traffic rerouted but packets sourced and sinked lost until one link returns end-to-end recovery necessary 22 April 2009 Weather Disruption-Tolerant Mesh Networks 40

41 Synthetic Storm Performance Analysis: Cumulative Loss 10s OSPF HELLO interval 40s OSPF dead interval 22 April 2009 Weather Disruption-Tolerant Mesh Networks 41

42 WDTN Routing Performance Analysis: Cumulative Loss Cumulative loss statistics track and compare overall availability during storm event P-WARP slightly better than XL-OSPF P-WARP and OSPF significantly better than standard OSPF static shows worst-case baseline 22 April 2009 Weather Disruption-Tolerant Mesh Networks 42

43 Simulations Observed Storm in Northeast Kansas Millimeter-wave grid location N, W Storm observed at: 20:39:26Z 30 Sep April 2009 Weather Disruption-Tolerant Mesh Networks 43

44 Observed Storm Performance Analysis: Packet Loss node out 40s OSPF dead interval link out 40s OSPF dead interval 10s XL-OSPF HELLO interval 22 April 2009 Weather Disruption-Tolerant Mesh Networks 44

45 Observed Storm Performance Analysis: Cumulative Loss 10s OSPF HELLO interval 40s OSPF dead interval 22 April 2009 Weather Disruption-Tolerant Mesh Networks 45

46 Observed Storm Performance Analysis: Delay 10s OSPF HELLO interval 40s OSPF dead interval 22 April 2009 Weather Disruption-Tolerant Mesh Networks 46

47 WDTN Routing Performance Analysis: Delay Delay proportional to number of hops rerouting adds some delay but not significant within a metro-area mesh Congestion avoidance overprovisioning of mesh essential to avoid congestion simulation studies based on past precipitation events drives network engineering for a given network 22 April 2009 Weather Disruption-Tolerant Mesh Networks 47

48 WDTN Routing Availability Protocol Availability Availability 10s OSPF HELLO interval synthetic storm observed storm P-WARP XL-OSPF 40s OSPF dead interval OSPF Static Availability during storm presence in grid neighbourhood overall availability much higher majority of time no storms are present 22 April 2009 Weather Disruption-Tolerant Mesh Networks 48

49 WDTN Mesh Networks Conclusions Overcomes a fundamental limitation millimeter wave links in the presence of weather events Demonstrates a resilient network architecture P-WARP slightly better than XL-OSPF difference important for loss-sensitive traffic XL-OSPF significantly better than conventional OSPF still affected by node outage (storm cell over tower) Real case study based on actual radar measurements Potential solution for data and Internet access in rural areas and metropolitan cities 22 April 2009 Weather Disruption-Tolerant Mesh Networks 49

50 WDTN Mesh Networks Future Work Model additional topologies hexagonal-packed cellular networks Model additional storm types hurricanes, nor easters with thundersnow, tropical cyclones monsoon rains Model link alternatives alternative lower-rate links fiber bypass 22 April 2009 Weather Disruption-Tolerant Mesh Networks 50

51 WDTN Mesh Networks Future Work: Alternative Wireless Links Mesh architecture high degree of connectivity alternate diverse paths severely attenuated mm wave alternate mm links alternate lower-freq. RF fiber bypass (competitor) Proposed solution route around failures before they occur avoid high error links CO/POP G 22 April 2009 Weather Disruption-Tolerant Mesh Networks 51

52 WDTN Mesh Networks Future Work: Alternative Fiber Bypass Mesh architecture high degree of connectivity alternate diverse paths severely attenuated mm wave alternate mm links alternate lower-freq. RF fiber bypass (competitor) Proposed solution route around failures before they occur avoid high error links CO/POP G 22 April 2009 Weather Disruption-Tolerant Mesh Networks 52

53 Weather Disruption-Tolerant Nets Publications Abdul Jabbar, Justin P. Rohrer, Andrew Oberthaler, Egemen Çetinkaya, Victor S. Frost, and James P.G. Sterbenz, Performance Comparison of Weather Disruption-Tolerant Cross-Layer Routing Algorithms, Proceedings of 28th IEEE Conference on Computer Communications (INFOCOM 2009 ), Rio de Janeiro, April 2009 [thanks to Merkouris Karaliopoulos for presenting] Abdul Jabbar, Bharatwajan Raman, Victor S. Frost, and James P.G. Sterbenz, Weather Disruption-Tolerant Self-Optimising Millimeter Mesh Networks, Third IFIP/IEEE Workshop on Self-Organizing Systems (IWSOS 2008 ), Vienna/Wein, December 2009, LNCS 5343, pp April 2009 Weather Disruption-Tolerant Mesh Networks 53

54 End 22 April 2009 Weather Disruption-Tolerant Mesh Networks 54