570 likes | 757 Views
Routing in Mesh Networks. Myung. J Lee, Chunhui Zhu Samsung Lab @ CUNY {lee, zhuc}@ccny.cuny.edu. Definitions of Mesh Network Features of Mesh Network Desirable Qualitative Properties (IETF) Quantitative Metrics Basic Classifications Proactive, Reactive and Hybrid
E N D
Routing in Mesh Networks Myung. J Lee, Chunhui Zhu Samsung Lab @ CUNY {lee, zhuc}@ccny.cuny.edu Myung Lee, Chunhui Zhu, Samsung
Definitions of Mesh Network Features of Mesh Network Desirable Qualitative Properties (IETF) Quantitative Metrics Basic Classifications Proactive, Reactive and Hybrid Single-path and Multi-path Flat and Hierarchical Basic Components of Routing Protocols Non-MANET Routing Protocols Geographical and Position-based Routing Directional Antenna Based Routing Power, Link Quality and other Cost-based Routing Multi-path Routing and Load Balancing Binary Tree Routing Virtual Backbone Based Routing Sensor Network Routing Layer 2 Data Forwarding Layer 2.5 Routing IETF & IRTF Activities Outline Myung Lee, Chunhui Zhu, Samsung
Mesh Network • General definition • Mobile Ad Hoc Networks (MANET) • Strict definition • Mesh routes (multiple routes) exist between each source-destination pairs • The application this group is considering • Some nodes are infrastructure nodes (e.g. APs) and others are client nodes (mobile user devices) [1] Myung Lee, Chunhui Zhu, Samsung
Features of Mesh Network • Infrastructure-less/Peer-to-Peer • Multi-hop – nodes need to relay packets for others • Random and dynamic topologies (e.g., mobility) • Bandwidth-constrained, variable capacity links • Lossy, unstable and asymmetric wireless links • Battery-powered nodes and sleeping nodes • MAC layer dependant (e.g. hidden terminal problem of 802.11) • Different antenna (omni-directional, directional) • Limited physical security • etc • All these affect the design of a routing protocol Myung Lee, Chunhui Zhu, Samsung
Desirable Qualitative Properties • Listed by IETF [2] • Distributed operation • Loop-freedom • Demand-based operation and/or Proactive operation • Security • "Sleep" period operation • Unidirectional link support Myung Lee, Chunhui Zhu, Samsung
Quantitative Metrics • IETF Metrics [2] • End-to-end data throughput and delay • Route Acquisition Time (on-demand routing) • Percentage Out-of-Order Delivery • Efficiency • Data Bit Efficiency: Data bits transmitted/data bit delivered • Control Overhead: Control bits transmitted/data bit delivered • Channel Access Efficiency: Control and data packets transmitted/data packet delivered • Other Protocol-Specific Metrics • Power consumption and network lifetime • Link quality and etc. Myung Lee, Chunhui Zhu, Samsung
Basic Classifications (1) • Proactive protocols • Maintaining route map of all nodes • Example protocols: DSDV[11], OLSR[12], FSR[13] • Reactive protocols • Adapting to the traffic pattern on a demand or need basis • Example protocols: DSR[14], AODV[15] • Hybrid protocols • Proactive for nearby nodes while reactive for distant nodes • Example protocols: ZRP[16] Myung Lee, Chunhui Zhu, Samsung
Basic Classifications (2) • Single Path • There exists only one route between a source-destination pair • Multiple Path • More than one route between a source-destination pair are created and maintained • Two types of Multi-path routing • Node disjoint routes • Link disjoint routes Myung Lee, Chunhui Zhu, Samsung
Basic Classifications (3) • Flat • All nodes are equivalent in routing functions • Hierarchical • The whole network is divided into multiple clusters • Cluster heads have more functions than regular nodes in routing • Suitable for large networks whose scalability are concerned Myung Lee, Chunhui Zhu, Samsung
Basic Components of Routing Protocols • Neighbor discovery and maintenance • Route discovery • Route selection • Route representation • Route maintenance • Failure response and route repair • Data forwarding Myung Lee, Chunhui Zhu, Samsung
1. Neighbor discovery and maintenance • Functions: • The base of most mesh network routing protocols • Link breakage detection • Calculating relay nodes for broadcast • Approaches: • 1-hop neighbor information • 2-hop Neighbor information (neighbor list in hello message) Myung Lee, Chunhui Zhu, Samsung
2. Route Discovery • Functions: • Finding the route(s) to the destination • Approaches: • Proactive • Broadcasting cost-to-all to neighbors (DSDV[11]) • Broadcastingcost-to-neighbor to all (OLSR[12] • A variation: Periodic update (aggregation possible) (GSR[17], FSR[13]) • Self-routing • Tree routing (binding address with topology) • Geographic-based routing Myung Lee, Chunhui Zhu, Samsung
2. Route Discovery (cont.) • Approaches: (cont.) • Reactive routing • Full-flooding RREQ (AODV[15], Expanding Ring Search) • Limited-Flooding RREQ (LAR[18]) • Probabilistic RREQ (ANT [29]) • Hybrid routing • Proactive for intra-cluster, reactive for inter-cluster (ZRP[16]) • Proactive for virtual backbone, reactive for end-devices (TTDD[19]) Myung Lee, Chunhui Zhu, Samsung
3. Route Selection • Functions: • Selecting the best route(s) to the destination • Approaches: • Proactive routing • Optimization based on distance vector (Bellman-Ford algorithm) • Optimization based on link state (Dijkstra’s algorithm to compute the shortest route) • Calculation based on tree topology (tree) • Calculation based on position information (GPSR[20]) Myung Lee, Chunhui Zhu, Samsung
3. Route Selection (cont.) • Approaches (cont.): • Reactive routing • Source behavior • First arrival RREP vs. selecting the best from multiple arrived RREPs • Destination behavior • Selective reply vs. reply to all RREQs • Intermediate nodes behavior • Filtering RREQs vs simple relaying Myung Lee, Chunhui Zhu, Samsung
4. Route Representation • Functions: • Storing the selected routes • Approaches: • Exact Route • Routing table • Interest entry (DD[21]) • Route in the packet (DSR[14]) • Binary tree (tree routing) • Route Guidance • Cost table (GRAd[22]) • Geographical information Myung Lee, Chunhui Zhu, Samsung
5. Route Maintenance • Functions: • Keeping the route available and fresh for desired time • Approaches: • Reactive • Control packet (AODVjr[23], “connection packet”) • Data packet • Proactive • Periodically Distance Vector update • Hybrid • Proactive for intra-cluster, reactive for inter-cluster (ZRP[16]) Myung Lee, Chunhui Zhu, Samsung
6. Failure Response & Route Repair • Functions: • Propagating the route failure information and recover the route • Failure Response: • Invalidate the corresponding route • Inform the source nodes who are using the broken route via a special control message (e.g. Route ERRor) • Route Repair: • Reactive • Rediscovery • Local repair • Route cache (DSR[14]) • Alternative path (multipath like TORA[24]) • Proactive • Update neighbor topology information to whole network upon failure (OLSR[12]) • Update neighbors with distance vector upon failure, then update whole network periodically (DSDV[11]) Myung Lee, Chunhui Zhu, Samsung
7. Data forwarding • Functions: • Forwarding data packets for other nodes using the discovered route • Approach: • Forward to next hop by routing table lookup or following the tree structure • Include routing information into data packets • Forward to certain direction (position-based or directional antenna based protocols) • Forward to cluster head or virtual backbone node (hierarchical protocols) • Simple data broadcast + receiver filtering Myung Lee, Chunhui Zhu, Samsung
Non-MANET Routing Protocols • Geographical and Position-based Routing • Directional Antenna Based Routing • Power, Link Quality and other Cost-based Routing • Multi-path Routing and Load Balancing • Binary Tree Routing • Virtual Backbone Based Routing • Sensor Network Routing • Layer 2 Data Forwarding (e.g. Spanning Tree Protocol) • Layer 2.5 Routing Myung Lee, Chunhui Zhu, Samsung
1. Geographical & Position-Based Routing • Why Position-based Routing? • Route packets without discovery. • Storage and bandwidth requirements increase slowly when the network grows • RREQs can be broadcast to certain direction instead of to the whole network • Enabling technologies • Absolute Position – GPS (outdoor only) • Relative Position – exchanging estimated distance information based on (indoor/outdoor) • TOA, AOA, and TDOA • Near field ranging algorithm • Signal strength based ranging algorithm Myung Lee, Chunhui Zhu, Samsung
1. Geographical & Position-Based Routing (cont.) • Location Service • Some for Some • Some for All • All for Some • All for All • Forwarding Strategies • Restricted Directional Flooding • Greedy Forwarding • Next-hop selection (minimum distance to the destination) • Recovery strategy when greedy algorithm fails • Hierarchical Approach • Virtual backbone/Infrastructure node Myung Lee, Chunhui Zhu, Samsung
Example: Location Aided Routing [18] • LAR – an on-demand source routing protocol • Route Discovery • The route request packet is forwarded to the rectangle request zone. • If a neighbor of S determines it is within the request zone, it forwards the route request further. • If a route reply packet is not received within the timeout, then the second route request is flooded through the network. Myung Lee, Chunhui Zhu, Samsung
Example: Location Aided Routing (cont) • The Expected Zone • – A circle area determined by the most recent location information on D. • The Request Zone • – A rectangle with S on one corner and the circle containing D in the other corner. • At time t0, Dest’s recorded location is (XD,YD), and velocity of Dest is Vavg • At current time t1, the expected zone is a circle with radius R = Vavg × (t1 – t0) Myung Lee, Chunhui Zhu, Samsung
2. Directional Antenna based Routing • Why directional antenna? • Improve network capacity by spatial reuse • Reduce interference • Extend range or Save power • Reduce the number of flooding packets • What to be considered? • Complexity (angle of the arrival packets and etc.) • Mobility handling / Antenna handoff • Deafness problem • Effective usage together with omni directional antenna Myung Lee, Chunhui Zhu, Samsung
Example: Directional DSR Myung Lee, Chunhui Zhu, Samsung
Example: Directional DSR (cont.) • Broadcast implemented through sweeping • RREQ transmitted sequentially on all N beams • Beam Handoffs (due to node mobility) handled through scanning • Send probe packets on recently used beams • Update neighbor cache based on replies to probes DiMAC – Directional MAC Myung Lee, Chunhui Zhu, Samsung
3. Power, Link Quality and other Cost-based Routing • Hop count • Less significant in MANETs, e.g. smaller hop-count does not necessarily mean smaller delay. • Link Quality • Less hops → longer per-hop distance → poorer link quality • Different link quality for forward and backward routes • Example: SSA[25] • Energy consumption is important • Many devices are battery powered • Especially true for wireless sensor networks • Shortened network lifetime if not well handled • Example: PARO[26] Myung Lee, Chunhui Zhu, Samsung
Example: SSA [25] • Signal Stability based Adaptive Routing • Selects routes based on the signal strength between nodes or a node’s location stability • Comprises two protocols: • Dynamic Routing Protocol (DRP) • Maintains the signal strength of neighboring nodes by periodic beacons among neighbors • Signal strength is recorded as a strong or weak channel • Static Routing Protocol (SRP) • processes the route search Myung Lee, Chunhui Zhu, Samsung
Route Discovery • The Query packets are forwarded only if they are received over strong channel and has never been processed before. • No packet can be sent over weak channel Myung Lee, Chunhui Zhu, Samsung
Route Discovery (cont.) • When there are only weak links can reach the destination • The source times out, it changes the PREF field in the header • Weak channels are acceptable Myung Lee, Chunhui Zhu, Samsung
Why Multi-path Routing? Robustness Higher packet delivery ratio Shorter route recovery time Energy and load balancing How multi-paths are formed? Node disjoint vs. Link disjoint Data forwarding strategies One route at a time (Round Robin) Multi-route at the same time Considerations Route maintenance cost Extra storage of multiple route entries Control traffic 4. Multi-path Routing and Load Balancing Myung Lee, Chunhui Zhu, Samsung
Example: Multipath Extension to DSR [27] • Multipath creation • The destination replies to a selected set of route queries. • Primary source route is the route taken by the first query reaching the destination. • Route queries that carry a source route that is link-wise disjoint from the primary source route are also replied – multiple routes created. • The source keeps all routes received on reply packets in its route cache; • Data Forwarding • When the primary route breaks, the shortest remaining alternate route is used. • This process continues until all routes break, then a fresh route discovery will be initiated. Myung Lee, Chunhui Zhu, Samsung
Example: Multipath Extension to DSR [27] • Improvement • multipath at all intermediate nodes (each node has a link-disjoint alternate path to the destination) Myung Lee, Chunhui Zhu, Samsung
5. Binary Tree Routing • Why Binary Tree Routing? • Good for networks with one portal to the wired network, e.g. home networks. • Binary tree structure binds addressing with routing • By checking the destination address of a given packet, every node knows how to route the packet • Routing decision is simple, either up to parent, or down to one of children • Low routing control overhead Myung Lee, Chunhui Zhu, Samsung
5. Binary Tree Routing (cont.) • Considerations: • Sub-optimal routes to peer nodes • Tree construction overhead • Mobility support • Fixed binding of address to the tree structure makes it a very difficult issue • when a node changes its connecting point to the tree, it has to change its address • Tree link breakage and repair • The scale of the network is limited by the address space • Load and energy balance Myung Lee, Chunhui Zhu, Samsung
Example: Cluster-Tree Routing [28] Cluster-tree addressing and self-routing Myung Lee, Chunhui Zhu, Samsung
6. Virtual Backbone Based Routing • Why Virtual Backbone Routing? • Virtual backbone nodes usually have more resources than client nodes, e.g. bandwidth, power, antenna gain and etc. • Virtual backbone nodes are usually static so that they can provide stable connections • Virtual backbone nodes sometimes form a dominant set over which can cover all network nodes – good for broadcast traffic • Considerations: • The selection and placement of the virtual backbone nodes • Virtual backbone maintenance overhead (e.g. virtual backbone reconfiguration when an existing backbone node fails) Myung Lee, Chunhui Zhu, Samsung
Example: Dynamic Backbone Algorithm [6][7] • A distributed protocol implements routing underneath the IP layer • At the IP layer, the network so-constructed behaves like an ordinary Ethernet subnet. • Different from MANET protocols (implemented at IP layer) • Maintaining a dynamic backbone for a subnet by periodically probing node connectivity • The backbone can be reconfigured in a known, fixed length of time (approximately 1 to 3 seconds) • Combining the features of ad-hoc networks with the control features of 802.11 infrastructure networks Myung Lee, Chunhui Zhu, Samsung
Example: Dynamic Backbone Algorithm (cont.) • Broadcast and multicast • The backbone approximates a spanning tree that connects all nodes and over which all broadcast traffic is routed. • Standard protocols such as ARP, RARP, DHCP, and other service discovery mechanisms can work in their normal fashion. • Unicast • Node connectivity information is sent over the backbone in order to develop unicast routing tables using Dijkstra’s Shortest Path Algorithm. • Unicast traffic is not constrained to use only backbone links, but is free to use shortest path routing. Myung Lee, Chunhui Zhu, Samsung
Source: Dennis Baker, James Hauser, “Mobile 802.11 Extended Service Sets using the Dynamic Backbone Subnet Architecture (DBS/802.11)”, Doc# 03/236 Myung Lee, Chunhui Zhu, Samsung
7. Sensor Network Routing • There could be only one sink (data collector) in the network; others are sensor nodes; • Sensor nodes could be very inexpensive and resource-limited; • Routing is done by using attributes instead of Node ID’s (e.g. temperature, light intensity and etc): • Directed Diffusion • Routing may be combined with collaborative signal processing • Data Fusion to reduce communication traffic Myung Lee, Chunhui Zhu, Samsung
Example: Directed Diffusion [21] interest reinforcement gradient high rate event low rate event Myung Lee, Chunhui Zhu, Samsung
8. Layer 2 Data Forwarding • Why Layer 2 Data Forwarding for 802.11s? • Hiding the multi-hop feature to higher layers (transparent to end nodes) • Network layer independent, e.g. IPv4, IPv6, AppleTalk and etc • Natural broadcast support • Good for protocols rely on broadcast, such as ARP, DHCP and service/device discovery • Roaming/mobility is easily supported • Considerations: • Broadcast even for unicast traffic when the destination is unknown • Address list grows proportionally with the number of the nodes in the network as with other proactive protocols Myung Lee, Chunhui Zhu, Samsung
Example: Spanning Tree Protocol • Spanning Tree • A tree connects all nodes in a network without loops • There is exactly one path from any one node to all other nodes • Spanning Tree Protocol (802.1D/1W) • A bridging technology and a link management protocol • Prevents undesirable loops in the network • Forcing redundant data paths into standby • Provides path redundancy • Activating the standby path Myung Lee, Chunhui Zhu, Samsung
Example : Spanning Tree Protocol (cont.) • Considerations: • Broadcast even for unicast traffic when the destination is unknown • Sub-optimal routes to peer nodes • Mobility support • Link change leads to global tree reconstruction Myung Lee, Chunhui Zhu, Samsung
Example : Spanning Tree Protocol (cont.) Spanning Tree Structure Myung Lee, Chunhui Zhu, Samsung
9. Layer 2.5 Routing • Why Layer 2.5 Routing? • Similar to Layer 2 Data forwarding • Major Approaches • Ad Hoc routing under IP • Ethernet emulation • Multihop support • Example: DBS[6][7], LUNAR[8], MCL (LQSR)[9] • Label switching • QoS support • Multi-protocol support • Mobility control Myung Lee, Chunhui Zhu, Samsung
Example 1: LUNAR [8] • Route discovery is triggered by ARP request • One-hop address resolution → multihop address resolution • New routes (L2.5 paths) are built on-demand • Source routing or L2.5 table driven • Routes are rediscovered every 3 seconds, route expires after 6 seconds • no hello, no repair Myung Lee, Chunhui Zhu, Samsung
Example 2: Label Switching • Route through L2.5 labels • Keep IP address consistent Myung Lee, Chunhui Zhu, Samsung