EEL 5718 Computer Communications Routing Algorithms and Protocols

EEL 5718 Computer Communications Routing Algorithms and Protocols 1: 37 AM

Outline • Network service, topology and addressing • Routing algorithms – – – SPA Dynamic routing algorithms (DVR and LSR) Flooding Broadcast Multicast • IP in TCP/IP suite 1: 37 AM

Network Layer • Moving packets around t 1 t 0 Network 1: 37 AM

Peer-to-Peer Protocol • Messages Segments Transport layer Network service Network layer End system Data link layer a Data link layer Data link End system layer b Physical layer Network layer 1: 37 AM

Network Services • Service types – Connectionless: datagrams è Simple interactions between transport layer and network layer: a request and an indication è Network does not need to be informed – Connection-oriented: virtual circuits è Transport layer has to establish the connection before sending any data to network layer è Network layer must be informed • Choice of services – Best-effort, low-delay connectionless service, connection-oriented stream service, Qo. S service 1: 37 AM

Network Topology • LANs (a) (b) LAN 1 Bridge LAN 2 1: 37 AM

Gateway Organization Servers To internet or wide area network s s Backbone R R R S Departmental Server R S S R R s s s s s Figure 7. 6

Addressing • Two types of addresses – Physical address: MAC address – Logical address: IP address • Address Resolution Protocol (ARP) – Map the IP address to MAC address – ARP server is usually needed – Search IP address from MAC address: Reverse ARP • Domain Name Service (DNS) – Map service between IP address and alphanumeric address 1: 37 AM

Internetworking • Using border routers Interdomain level Border routers Autonomous system or domain LAN level 1: 37 AM Intradomain level Border routers Internet service provider

National service provider A (a) National service provider B NAP National service provider C (b) NAP RA RB Route server LAN RC

Routing • Task: Deliver packets from source to destination over a subnet – – Path finding Efficiency: how good a path is? Flow control or congestion control Routing • Design issues – Interfaces with transport layer and DLC layer – Framing: network protocol unit—packet è Packet 1: 37 AM networks

Information Needed • Topology: the layout of the subnet • Traffic • Cost factors – number of hops – delay: time factors (processing, queueing or propagation) – traffic – transmission capacity 1: 37 AM

Routing Algorithms • A routing algorithm: a decision algorithm of finding a path from a source to destination to efficiently deliver packets • Multiple paths: there may be many paths from a source to a destination, a routing algorithm is to find the BEST one! • Optimality: one’s meat is another’s poison! a criterion must be set before we talk about optimality 1: 37 AM

An example • A UF student – walk: does not cost anything but time (with additional benefit: exercise) – use a bike: you have to buy a bike – take a bus: cost a little (UF students get free ride? ) but less convenient – drive: need to buy/rent a car, PARKING – take a taxi: cost a lot but less time • Which “path” is optimal? • Solution? Depends: need to consider money, time, convenience etc 1: 37 AM

Routing Algorithm Evaluation • Correctness: loop-free and deadlock-free • Simplicity: the simpler, the better • Robustness or Adaptability: cope with changes of topology and traffic and with failures • Stability: a stable path may be better • Fairness: fair to all packets/users/paths/routers • Optimality: in terms of indices according to network design objective • Efficiency: the less overhead, the better (information vs. overhead/control messages) 1: 37 AM

Performance Indices • Average packet delay: need to be minimized • Network throughput (efficiency): need to be maximized • Tradeoff: Many times suboptimal routing algorithms may be the “best” when all things considered • Cost optimization: many factors are abstracted into a framework of costs, a routing algorithm is to find the optimal solution in terms of total cost from source to destination 1: 37 AM

Algorithm vs Protocol • An algorithm is the method finding a path • An protocol is the implementation of a routing algorithm, may involve interface designs, information collection, route maintenance/route repair, reaction to various changes etc • Interchangeably used in some literature 1: 37 AM

Algorithm Classification • Depending on where and when an algorithm is executed, may have many types of algorithms • Static (fixed) vs dynamic (adaptive) • centralized vs distributed 1: 37 AM

Static (fixed) Routing • Single permanent route for each source to destination pair • Determine routes using a least cost algorithm • Route fixed, at least until a change in network topology 1: 37 AM

Dynamic (adaptive) Routing • Used by almost all packet switching networks • Routing decisions change as conditions on the network change – Failure – Congestion • Requires info about network (topology, traffic etc) • Decisions are more complex • Tradeoff between quality of network info and overhead – Reacting too quickly can cause oscillation – Too slowly to be relevant 1: 37 AM

Centralized vs Distributed Routing • Centralized routing needs information about the subnet of interest in terms of topology and traffic, it is executed at node who “knows” the subnet • Distributed routing only uses local information, does not need to know the whole subnet, it is executed at any distributed node 1: 37 AM

Optimality Principle • Dynamic programming (Bellman, 1960, Rand Corporation) • The only theorem is stated without any assumptions! • The principle: In solving a problem, a series of stages need to go through, the optimal way is the on which from any stage on the way towards the final stage is also the optimal way from the stage to the final stage, i. e. , always optimize from the current stage to the final stage 1: 37 AM

Optimality Principle (cont) • The optimality principle for optimal path finding: The path from any node, say, I, on the optimal path to the destination, say, D, is also the optimal path from I to D • Many algorithms are inherently using this principle • Sink tree example: optimal in terms of hops, use destination node as the root of tree, then branch out, each node on the tree will give you the optimal path to the root (destination)! 1: 37 AM

Shortest Path Algorithm (SPA) • Objective: Find the path from the source to destination which minimizes the total path cost • Information needed: topology (every node knows all information about the network) and link cost from node to node • Modeling: a network can be modeled by a graph 1: 37 AM

SPA (cont) • Terminology – vertex: a node or a router – arc (edge): a link – weights: cost or link cost • Weights: hop (if weight=1), physical distance, bandwidth, traffic, communication cost, queueing length, delay or other factors • Two main SPA algorithms – Bellman-Ford – Dijkstra 1: 37 AM

Bellman-Ford Algorithm (1957, 1962) • The basic idea: if you could find a shorter path through your neighbor, use it! – If a node is on the shortest path between source S and destination D, then the path from the node to S must be the shortest and the path from the node to D must also be the shortest (Optimality Principle) • Also called Ford-Fulkerson algorithm • Notation: (n, D(v)): D(v) is the current path cost from v to the destination, n is the next node label along the currently computed path 1: 37 AM

Bellman-Ford Algorithm (1957, 1962) • Basic idea: if I have to go to the destination through a few alternatives, I will choose the best alternative • BF Algorithm – Step 1: Initialization-- With node d as the destination, set D(d)=0 and label all others (. , inf) – Step 2: Shortest path labeling of all nodes-- For each node v (not d), do the following: update D(v) by using the current value D(w) for each neighboring node w to calculate l(v, w)+D(w), i. e. , performing the operation: (n, D(v)) <-- (w 1, min {l(v, w)+D(w)}) where w 1 is the minimizer – Step 3: Repeat Step 2 until no further change 1: 37 AM

An Example • A sample network 1 2 5 6 2 3 4 1 2 3 2 4 1: 37 AM 1 3 5

Solution • Find shortest path to d=6 Iteration Node 1 Node 2 Node 3 Node 4 Node 5 Initial (-1, inf) (-1, inf) 1 1: 37 AM (-1, inf) (6, 1) (3, 3) (6, 2) 2 (3, 3) (4, 4) (6, 1) (3, 3) (6, 2) 3 (3, 3) (4, 4) (6, 1) (3, 3) (6, 2)

Shortest Path Tree • SP Tree 1 2 1 3 6 2 1: 37 AM 4 2 5 Figure 7. 29

Dijkstra Algorithm (1959) • Basic idea: If there a few alternative nodes passing packet to me from source, choose the best one to me! • SPA – D(v): the cost from source, say, 1 to node v – l(i, j): the link cost between node i and node j. If not directly connected, its cost is infinity (inf) – Step 1: Initialization-- Set N={1}. For each node v not in N, set D(v)=l(1, v). – Step 2: Optimize--find a node w not in N for which D(w) is the minimum and add w to N; Stop if all nodes are in N – Step 3: Update--update D(v) for all nodes not in N è D(v) 1: 37 AM <-- min{D(v), D(w)+l(w, v)} è Go to Step 2

An Example • The sample network Iteration N D 2 D 3 D 4 D 5 D 6 Initial {1} 3 2 5 inf 1 {1, 3} 3 2 4 inf 3 2 {1, 2, 3} 3 2 4 7 3 3 {1, 2, 3, 6} 3 2 4 5 3 4 {1, 2, 3, 4, 6} 3 2 4 5 3 5 {1, 2, 3, 4, 5, 6} 3 2 4 5 3 1: 37 AM

Dijkstra Algorithm (cont) • SP Tree 1 2 1 3 6 2 3 4 2 1: 37 AM 2 5

Dijkstra Algorithm (cont) • Shortest path tree from source to all destinations: each node will keep such a tree, or a routing table with next hop for each destination. Whenever a packet comes, the node checks the destination and forwards the packet to the corresponding “next hop node” • Example: Destination Next-hop node Total cost 1: 37 AM 2 2 3 3 4 5 6 3 3 2 4 5 3

Routing Tables • Routing table is used at each node for routing decision making – Virtual circuit è Incoming: the other end nodes of the link and VCIs è Outgoing: the other end nodes of the link and VCIs – Datagram è Destination è Next node IDs hop è Total cost to each destination 1: 37 AM

Routing Table Format • Node 1 Destination Next node 2 2 3 3 4 4 5 2 6 3 Node 2 Destination Next node 1 1 3 1 4 4 5 5 6 5 1: 37 AM Node 3 Destination Next node 1 1 2 4 4 4 5 6 6 6 Destination 1 2 3 5 6 Node 4 Next node 1 2 3 5 3 Node 6 Destination Next node 1 3 2 5 3 3 4 3 5 5 Node 5 Destination Next node 1 4 2 2 3 4 4 4 6 6

Asymmetric Costs • Link costs between two nodes may be different 1: 37 AM

Dynamic (Adaptive) Routing • Adapt to network topology changes and traffic changes • Handle node and link failures • Rely on gathered information from the network operation • Two important dynamic algorithms – Distance Vector Routing (DVR) – Link State Routing (LSR) 1: 37 AM

Distance Vector Routing (DVR) • Based on Bellman-Ford algorithm, first used in APARNET, RIP, early version of DECNET and Novell’s IPX • SPA, particularly, Bellman-Ford, uses the neighboring node’s cost to destination • Basic idea: Trust your neighbors! If your neighbor has a shorter path to the destination, use it 1: 37 AM

DVR (cont) • Each router maintains a routing table • Table entries: for each destination router, the preferred outgoing link (port) to use for that destination, the estimated cost to that destination • Each router will also have the link cost estimate to all its neighbors • If your neighbors tell you their routing tables and you also gather the link cost, then you could choose your best path to the destination 1: 37 AM

DVR (cont) • Example (Tanenbaum) 1: 37 AM

DVR (cont) • Table update: 1: 37 AM

DVR (cont) • J updates its table using neighbors’ routing info • J to A: JA=min{JA, JIA, JHA, JKA} =min{JA, JI+IA, JH+HA, JK+KA} =min{8, 10+24, 12+20, 6+21}=8 via outgoing link JA directly • J to B: JB=min{JAB, JIB, JHB, JKB} =min{JA+AB, JI+IB, JH+HB, JK+KB}=20 via outgoing link JA • etc. 1: 37 AM

Problem with DVR • Relying your neighbors may lead to serious problem: Romeo and Juliet’s tragedy • Security: a rogue router zeroes all costs to all destination, can be corrected by authentication! • Slow convergence: reacts quickly to good news, but slowly to bad news • Count-to-infinity 1: 37 AM

Good News Propagation 1: 37 AM

Count-to-infinity • A fails: 1: 37 AM

Solution to Count-to-infinity • Many ad hoc solutions • Split horizon algorithm: works the same way as DVR, except that the distance to X is not reported on the line that the packets for X are sent on (actually, it is reported as infinity) • Example: in previous example, C tells D the truth about its distance to A, but tells B its distance to B is infinity (because C obtains its update from B, or C finds the SPA from B), i. e. , a node will not send update packet to its neighbor which gives the node the optimal path 1: 37 AM

Link State Routing (LSR) • DVR relies on its neighbors, original implementation uses queueing length to estimate the link cost, does not consider the link BW (all lines were assumed to be 56 kbps!) • DVR is very slow to converge! • LSR attempts to learn (gather) information about the network topology and network traffic, then applies the SPA to find the optimal path. In this way, LSR could adapt to the network changes and speeds up the convergence, and solves the count-to-infinity problem 1: 37 AM

LSR--Five Steps • Discover its neighbors and learn their network addresses • Measure the cost to each of its neighbors • Construct a packet telling all it has just learned • Send this packet to all other routers • Compute the shortest path to every other router 1: 37 AM

Learning about the Neighbors • When a router is booted, it first learns all its neighbors by sending a special HELLO message on each point-to-point link • Upon receiving this message, a neighbor will send back a response message with network address • On a broadcast link (in LAN), the whole LINK/LAN can be regarded as one node (which is reasonable because whenever a packet is sent to the LAN, everybody hears) 1: 37 AM

Measuring the Link Cost • Commonly used link cost is the link delay • Sending a special ECHO message and the receiving neighbor is asked to respond immediately, the delay estimate will be half of Round-Trip-Time (RTT) • The choice: count the queueing delay or not (the load) • Using queueing delay may cause traffic oscillation, while ignoring queueing delay may congest a link 1: 37 AM

Building Link State Packets 1: 37 AM

Distributing the Link States • Trickiest part is to distribute the link state packets reliably • Basic algorithm: flooding or broadcast routing (discussed later) • Sequence number is used to keep track of link state packets (distinguish the duplicates), age is used to get rid of old copies • Refined more robust algorithms are possible 1: 37 AM

Computing the New Paths • When a router receives all updates, it will have rough picture of the network topology and traffic, then SPA can be used to find all new optimal paths and the routing table can be updated • Memory requirement: n routers, each of which has k neighbors will require kn table entries! 1: 37 AM

LSR Pros and Cons • Fast convergence and much more robust, it provides the foundation for many other algorithms (OSPF) and overcomes many problems DVR has, widely used in actual networks • Too many packets are generated for information gathering and dissemination, too much memory is required 1: 37 AM

Flooding • Basic flooding: sending a packet to all routers in the subnet • Keep flooding in check: check whether a packet has been flooded (forwarded), sequence number can be used, a list of sequence number may be kept by each source • Selective flooding: a sending router sends only on the lines which are going approximately in the right direction 1: 37 AM

Random Routing • Random routing has the simplicity and robustness of flooding with far less traffic load. • RR algorithm: a node selects only one outgoing link (not the link a packet arrives on) with certain probability to forward a copy of the broadcast packet. If the outgoing links are chosen with equal probability, it is round-robin scheme • The probability can be chosen according to link capacity, traffic situation etc, various schemes are possible (load balancing) 1: 38 AM

Hierarchical Routing (HR) • Overcome the memory problem of routing table and processing speed • Routers are divided into regions, a router only knows the routers in its region, but knows nothing about the internal structure of other regions • Multiple levels may be used: regions are formed into clusters, clusters into zones, etc. 1: 38 AM

HR Operations 1: 38 AM

Broadcast Routing • Sending a packet to all destinations simultaneously is broadcasting • Stupid one: send it to each node using the regular routing algorithm • Flooding: flood the packet over the subnet • Multidestination routing: the packet contains a list of destinations or a bit map indicating the desired destinations (HR may be used) 1: 38 AM

Broadcast Routing (cont) • Sink (spanning) tree broadcasting: find a spinning tree with the broadcast node as the root, forward packets along the tree * Spanning tree: a subset of the subnet containing all routers but no loops * Sink tree: a spanning tree on which each node has the minimum distance to the root in terms of hops * Sink tree can be found by SPA with hop count as cost with information provided by link state routing * Sink tree or spanning tree is not unique! 1: 38 AM

Broadcast Routing (cont) • Reverse path forwarding: When a broadcast packet arrives at a router, the router checks whether it arrives from the link that is normally used for sending packets to the source of the broadcast (the preferred path or shortest path to the broadcast source). If so, there is an excellent chance that the broadcast packet itself followed the optimal path and is the first copy to arrive at the router, the router will forward copies of it onto all lines except the line it arrives on. Otherwise, the packet is discarded 1: 38 AM

Broadcast Routing (cont) • Reverse path forwarding example (assuming that the spanning tree (b) provides the preferred paths) 1: 38 AM

Multicast Routing • Multicasting: Sending packets to a selected group of routers • MR: an algorithm that routes multicast packets • Group management: create and destroy groups, join and leave groups, information collection and dissemination • One MR algorithm: * Construct a spanning tree at each router * Prune the spanning tree for a multicasting group 1: 38 AM

Multicast Routing (cont) • Example 1: 38 AM

IP in TCP/IP Suite • IP: Internet Protocol • Provides a connectionless and best-effort service to transport layer • IPv 4 is commonly used, IPv 6 has been standardized • Packet is the PDU in IP layer – 20 bytes header – Variable payload 1: 38 AM

Packet Format • IPv 4 header 0 4 Version 8 IHL 16 Type of Service Identification Time to Live 19 24 31 Total Length Flags Protocol Fragment Offset Header Checksum Source IP Address Destination IP Address Options 1: 38 AM Padding

Packet Header • Various fields – – Version: current is 4, future is 6 Internet Header Length (IHL): 32 -bit unit To. S: specifies the priority Total length: length of the total packet in bytes è Max: 64 KB. In Ethernet, the max size is 1500 B – Identification, flags and fragment offset: fragmentation/assembly – TTL: time-to-live – Protocol: TCP, UDP, or ICMP – Header checksum: protect the header part, crucial data 1: 38 AM

Packet Header • Various fields (cont) – Address: IP address for source and destination – Options: for special features such as security level, route to be taken (source routing), time-stamp etc – Padding: adding zeros to make the packet as a multiple of 32 bits 1: 38 AM

IP Address • IP address = Network ID + Host ID Bit position: 0 1 2 3 Class A 0 Class B 1 0 Class C 1 1 0 Class D 1 1 1 0 Class E 1 1 1: 38 AM 8 16 Net ID 31 Host ID Net ID Multicast address Reserved for experiments Host ID

Address/ID • IP dotted-decimal notation – 32 bits is broken into 4 bytes, each byte is represented by a decimal number è 10000000 10000111 0100 00000101 128. 135. 68. 5 • Alphanumeric representation – Map between numeric number to meaningful words – DNS: domain name service provides the translation 1: 38 AM

IP Routing • LSR: link state routing is commonly used • OSPF: Open Shortest Path First • Rely on routing table – Next-hop – Network interface • Routing tables at each node are updated regularly (periodically) • Control messages are exchanges via ICMP: Internet Control Message Protocol 1: 38 AM

IP Routing • CIDR: Classless Interdomain Routing – Attempt to resolve the inflexibility of classful address space restriction (A, B, C) – A prefix is used • ARP: Address Resolution Protocol – Using IP address to find MAC address – Broadcast IP address and wait for response • RARP: Reverse Address Resolution Protocol – Using MAC address to determine IP address 1: 38 AM

IP Routing • Interior Routing: Inside AS (Autonomous System) or a domain sharing routing information – RIP: based on DVR (outdated) – LSR: used for quite sometime – OSPF: THE protocol • Exterior routing – BGP: Border Gateway Protocol è Internal, External • Internet Group Management Protocol (IGMP) 1: 38 AM

Reading • Textbook, Chapter 7, Chapter 8 • Tanenbaum, Chapter 5 • Stallings, Chapter 10 1: 38 AM