Routing is the process of selecting paths in a network along which to send network traffic.

Transcription

1 Routing is the process of selecting paths in a network along which to send network traffic. In packet switching networks, routing directs packet forwarding, the transit of logically addressed packets from their source toward their ultimate destination through intermediate nodes, typically hardware devices called routers, bridges, gateways, or switches. Generalpurpose computers can also forward packets and perform routing, though they are not specialized hardware and may suffer from limited performance. The routing process usually directs forwarding on the basis of routing tables which maintain a record of the routes to various network destinations. Thus, constructing routing tables, which are held in the router's memory, is very important for efficient routing. In a practice known as static routing (or non-adaptive routing), small networks may use manually configured routing tables. Larger networks have complex topologies that can change rapidly, making the manual construction of routing tables unfeasible. Adaptive routing, or dynamic routing, attempts to solve this problem by constructing routing tables automatically, based on information carried by routing protocols, and allowing the network to act nearly autonomously in avoiding network failures and blockages. A routing protocol is a protocol that specifies: 1. how routers communicate with each other, 2. disseminating information that enables them to select routes between any two nodes on a computer network, 3. the choice of the route being done by routing algorithms. Each router has a priori knowledge only of networks attached to it directly. A routing protocol shares this information first among immediate neighbors, and then throughout the network. This way, routers gain knowledge of the topology of the network. The term routing protocol may refer specifically to one operating at layer three of the OSI model, which similarly disseminates topology information between routers. Although there are many types of routing protocols, three major classes are in widespread use on IP networks: Interior gateway routing via link state routing protocols, such as OSPF and IS-IS Interior gateway routing via path vector or distance vector protocols, such as RIP, IGRP and EIGRP Exterior gateway routing. BGP v4 is the routing protocol used by the public Internet. The specific characteristics of routing protocols include 1. the manner in which they either prevent routing loops from forming or break them up if they do 2. the manner in which they select preferred routes, using information about hop costs 3. the time they take to converge 4. how well they scale up 5. many other factors

2 Routed versus routing protocols A routed protocol can be routed by a router, i.e., it can be forwarded from one router to another. A routing protocol sends and receives packets containing routing information to and from other routers. In some cases, routing protocols can themselves run over routed protocols: for example, BGP runs over TCP which runs over IP; care is taken in the implementation of such systems not to create a circular dependency between the routing and routed protocols. That a routing protocol runs over particular transport mechanism does not mean that the routing protocol is not of layer three. Adaptive routing dominates the Internet. However, the configuration of the routing protocols often requires a skilled touch; networking technology has not developed to the point of the complete automation of routing. Distance Vector Algorithms A distance-vector routing protocol requires that a router informs its neighbors of topology changes periodically. Compared to link-state protocols, which require a router to inform all the nodes in a network of topology changes, distance-vector routing protocols have less computational complexity and message overhead. The term distance vector refers to the fact that the protocol manipulates vectors (arrays) of distances to other nodes in the network. Routers using distance vector protocol do not have knowledge of the entire path to a destination. Instead DV uses two methods: Direction in which or interface to which a packet should be forwarded. Distance from its destination. Examples of distance-vector routing protocols include RIPv1 and RIPv2 and IGRP. EGP and BGP are not pure distance-vector routing protocols because a distance-vector protocol calculates routes based only on link costs whereas in BGP, for example, the local route preference value takes priority over the link cost. Distance vector algorithms use the Bellman-Ford algorithm. This approach assigns a number, the cost, to each of the links between each node in the network. Nodes will send information from point A to point B via the path that results in the lowest total cost (i.e. the sum of the costs of the links between the nodes used). The algorithm operates in a very simple manner. When a node first starts, it only knows of its immediate neighbors, and the direct cost involved in reaching them. (This information, the list of destinations, the total cost to each, and the next hop to send data to get there, makes up the routing table, or distance table.) Each node, on a regular basis, sends to each neighbor its own current idea of the total cost to get to all the destinations it knows of. The neighboring node(s) examine this information, and compare it to what they already 'know'; anything which represents an improvement on what they already have, they insert in their own routing

3 table(s). Over time, all the nodes in the network will discover the best next hop for all destinations, and the best total cost. When one of the nodes involved goes down, those nodes which used it as their next hop for certain destinations discard those entries, and create new routing-table information. They then pass this information to all adjacent nodes, which then repeat the process. Eventually all the nodes in the network receive the updated information, and will then discover new paths to all the destinations which they can still "reach". Distance Vector means that Routers are advertised as vector of distance and Direction. Direction is simply next hop address and exit interface and Distance means hop count. Routers using distance vector protocol do not have knowledge of the entire path to a destination. Instead DV uses two methods: Direction in which router or exit interface a packet should be forwarded. Distance from its destination. In distance vector routing, the least cost route between any two nodes is the route with minimum distance. In this protocol, as the name implies, each node maintains a vector (table) of minimum distance to every node. As the name suggests the DV protocol is based on calculating the direction and distance to any link in a network. The cost of reaching a destination is calculated using various route metrics. RIP uses the hop count of the destination whereas IGRP takes into account other information such as node delay and available bandwidth. Updates are performed periodically in a distance-vector protocol where all or part of a router's routing table is sent to all its neighbors that are configured to use the same distance-vector routing protocol. RIP supports cross-platform distance vector routing whereas IGRP is a Cisco Systems proprietary distance vector routing protocol. Once a router has this information it is able to amend its own routing table to reflect the changes and then inform its neighbors of the changes. This process has been described as routing by rumor because routers are relying on the information they receive from other routers and cannot determine if the information is actually valid and true. There are a number of features which can be used to help with instability and inaccurate routing information. Limitations Count-to-infinity problem The Bellman-Ford algorithm does not prevent routing loops from happening and suffers from the count-to-infinity problem. The core of the count-to-infinity problem is that if A tells B that it has a path somewhere, there is no way for B to know if the path has B as a part of it. To see the problem clearly, imagine a subnet connected like as A-B-C-D-E-F, and let the metric between the routers be "number of jumps". Now suppose that A is taken offline. In the vectorupdate-process B notices that the route to A, which was distance 1, is down - B does not receive the vector update from A. The problem is, B also gets an update from C, and C is still not aware of the fact that A is down - so it tells B that A is only two jumps from C (C to B to A), which is false. This slowly propagates through the network until it reaches infinity (in which case the algorithm corrects itself, due to the "Relax property" of Bellman-Ford). The addition

4 of a hold time (refusing route updates for a few minutes after a route retraction) avoids loop formation in virtually all cases, but causes a significant increase in convergence times. Link-state algorithms When applying link-state algorithms, each node uses as its fundamental data a map of the network in the form of a graph. To produce this, each node floods the entire network with information about what other nodes it can connect to, and each node then independently assembles this information into a map. Using this map, each router then independently determines the least-cost path from itself to every other node using a standard shortest paths algorithm such as Dijkstra's algorithm. The result is a tree rooted at the current node such that the path through the tree from the root to any other node is the least-cost path to that node. This tree then serves to construct the routing table, which specifies the best next hop to get from the current node to any other node. Examples of link-state routing protocols include OSPF and IS-IS. This contrasts with distance-vector routing protocols, which work by having each node share its routing table with its neighbors. In a link-state protocol the only information passed between nodes is connectivity related. Link state algorithms are sometimes characterized informally as each router 'telling the world about its neighbors'. Process Distributing maps The first main stage in the link-state algorithm is to give a map of the network to every node. This is done with several simple subsidiary steps. 1. Determining the neighbors of each node First, each node needs to determine what other ports it is connected to, over fully working links; it does this using a simple reachability protocol which it runs separately with each of its directly connected neighbors. 2. Distributing the information for the map Next, each node periodically and in case of connectivity changes makes up a short message, the link-state advertisement, which: Identifies the node which is producing it. Identifies all the other nodes (either routers or networks) to which it is directly connected. Includes a sequence number, which increases every time the source node makes up a new version of the message. This message is then flooded throughout the network. As a necessary precursor, each node in the network remembers, for every other node in the network, the sequence number of the

5 last link-state message which it received from that node. With that in hand, the method used is simple. Starting with the node which originally produced the message, it sends a copy to all of its neighbors. When a link-state advertisement is received at a node, the node looks up the sequence number it has stored for the source of that link-state message. If this message is newer (i.e. has a higher sequence number), it is saved, and a copy is sent in turn to each of that node's neighbors. This procedure rapidly gets a copy of the latest version of each node's link-state advertisement to every node in the network. Networks running link state algorithms can also be segmented into hierarchies which limit the scope of route changes. These features mean that link state algorithms scale better to larger networks. 3. Creating the map Finally, with the complete set of link-state advertisements (one from each node in the network) in hand, it is obviously easy to produce the graph for the map of the network. The algorithm simply iterates over the collection of link-state advertisements; for each one, it makes links on the map of the network, from the node which sent that message, to all the nodes which that message indicates are neighbors of the sending node. No link is considered to have been correctly reported unless the two ends agree; i.e. if one node reports that it is connected to another, but the other node does not report that it is connected to the first, there is a problem, and the link is not included on the map. Notes about this stage The link-state message giving information about the neighbors is recomputed, and then flooded throughout the network, whenever there is a change in the connectivity between the node and its neighbors, e.g. when a link fails. Any such change will be detected by the reachability protocol which each node runs with its neighbors. Calculating the routing table As initially mentioned, the second main stage in the link-state algorithm is to produce routing tables, by inspecting the maps. This is again done with several steps. 1. Calculating the shortest paths Each node independently runs an algorithm over the map to determine the shortest path from itself to every other node in the network; generally some variant of Dijkstra's algorithm is used. This is based around a link cost across each path which includes available bandwidth among other things. Basically, a node maintains two data structures: a tree containing nodes which are "done",

6 and a list of candidates. The algorithm starts with both structures empty; it then adds to the first one the node itself. The variant of a Greedy Algorithm then repetitively does the following: All nodes which are connected to the node just added to the tree (excepting any nodes which are already in either the tree or the candidate list) are added to the second (candidate) list. Each node in the candidate list is compared to each of the nodes already in the tree. The candidate node which is closest to any of the nodes already in the tree is itself moved into the tree and attached to the appropriate neighbor node. When a node is moved from the candidate list into the tree, it is removed from the candidate list and is not considered in subsequent iterations of the algorithm. The above two steps are repeated as long as there aren't any nodes left in the candidate list. (When there are none, all the nodes in the network will have been added to the tree.) This procedure ends with the tree containing all the nodes in the network, with the node on which the algorithm is running as the root of the tree. The shortest path from that node to any other node is indicated by the list of nodes one traverses to get from the root of the tree, to the desired node in the tree. [edit] Filling the routing table With the shortest paths in hand, filling in the routing table is trivial. For any given destination node, the best path for that destination is the node which is the first step from the root node, down the branch in the shortest-path tree which leads toward the desired destination node. To create the routing table, it is only necessary to walk the tree, remembering the identity of the node at the head of each branch, and filling in the routing table entry for each node one comes across with that identity. Path vector protocol Distance vector and link state routing are both intra-domain routing protocols. They are used inside an autonomous system, but not between autonomous systems. Both of these routing protocols become intractable in large networks and cannot be used in Inter-domain routing. Distance vector routing is subject to instability if there are more than a few hops in the domain. Link state routing needs huge amount of resources to calculate routing tables. It also creates heavy traffic due to flooding. Path vector routing is used for inter-domain routing. It is similar to distance vector routing. In path vector routing we assume there is one node (there can be many) in each autonomous system which acts on behalf of the entire autonomous system. This node is called the speaker node. The speaker node creates a routing table and advertises it to neighboring speaker nodes in neighboring autonomous systems. The idea is the same as distance vector routing except that only speaker nodes in each autonomous system can communicate with each other. The speaker node advertises the path, not the metric of the nodes, in its autonomous system or other autonomous systems. Path vector routing is discussed in RFC 1322; the path vector routing algorithm is somewhat similar to the distance vector algorithm in

7 the sense that each border router advertises the destinations it can reach to its neighboring router. However, instead of advertising networks in terms of a destination and the distance to that destination, networks are advertised as destination addresses and path descriptions to reach those destinations. A route is defined as a pairing between a destination and the attributes of the path to that destination, thus the name, path vector routing, where the routers receive a vector that contains paths to a set of destinations. The path, expressed in terms of the domains (or confederations) traversed so far, is carried in a special path attribute that records the sequence of routing domains through which the reachability information has passed. BGP is an example of a path vector protocol. In BGP the routing table maintains the autonomous systems that are traversed in order to reach the destination system. Exterior Gateway Protocol (EGP) does not use path vectors. Comparison of routing algorithms Distance-vector routing protocols are simple and efficient in small networks and require little, if any, management. However, traditional distance-vector algorithms have poor convergence properties due to the count-to-infinity problem. This has led to the development of more complex but more scalable algorithms for use in large networks. Interior routing mostly uses link-state routing protocols such as OSPF and IS- IS. A more recent development is that of loop-free distance-vector protocols (e.g., EIGRP). Loopfree distance-vector protocols are as robust and manageable as naive distance-vector protocols, but avoid counting to infinity, and have good worst-case convergence times. Path selection Path selection involves applying a routing metric to multiple routes, in order to select (or predict) the best route. In the case of computer networking, the metric is computed by a routing algorithm, and can cover such information as bandwidth, network delay, hop count, path cost, load, MTU, reliability, and communication cost. The routing table stores only the best possible routes, while link-state or topological databases may store all other information as well. Because a routing metric is specific to a given routing protocol, multi-protocol routers must use some external heuristic in order to select between routes learned from different routing protocols. Cisco's routers, for example, attribute a value known as the administrative distance to each route, where smaller administrative distances indicate routes learned from a supposedly more reliable protocol. A local network administrator, in special cases, can set up host-specific routes to a particular machine which provides more control over network usage, permits testing and better overall security. This can come in handy when required to debug network connections or routing tables.

8 IGRP (Interior Gateway Routing Protocol) EIGRP (Enhanced Interior Gateway Routing Protocol) OSPF (Open Shortest Path First) RIP (Routing Information Protocol) IS-IS (Intermediate System to Intermediate System)

9 Network Layer Routing and Switching switch a network node that forwards packets towards a destination depending on a locally significant connection identifier over a fixed path. Fixed path is called a virtual circuit and is set up by a signaling protocol (switched virtual circuit (SVC)). connection a logical association of two endpoints. Only needs to be referenced not identified by to and from information. A data unit sent on connection 22 can only flow between two endpoints where it is established; no need to specify more. As long as there is no confusion in the network, connection identifiers can be reused and therefore have what is called local significance only. Packets on SVCs are often checked for errors hop by hop and are resent as necessary from node to node with the originator playing no role. Packet switching networks (PSN) offer guaranteed delivery and are also reliable in the sense that certain performance guarantees, in terms of bandwidth, delay, etc., can be enforced on the connection because packets always follow the same path through the network. router a network node that independently forwards packets towards a destination based on a globally unique address over a dynamic path that can change from packet to packet but usually is fairly stable over time. Packets are seldom checked for errors hop by hop and are only resent from host to host (originator plays a key role). Offers best effort delivery (but is usually error free) and is also considered unreliable in the sense that certain performance guarantees in terms of bandwidth, delay, etc., cannot be enforced from end-to-end because packets follow different paths. Connection-oriented and Connectionless Networks (Switched vs Router based networks) The signaling protocol messages used on a switched network to set up SVCs are themselves routed between switches in a connectionless manner using globally unique addresses (GUA). These call setup messages must be routed because obviously there are no connection paths to follow yet. Every switched network that offers SVCs must also be a connectionless router based network as well. Router based networks are simplier than switched networks since they handle everything like a signaling protocol message. Internet routers have to keep up with a list of every possible reachable destination in the world. Switches only have to keep track of local associations of two end points currently established. Quality of Service (QoS) In spite of the movement to converge all types of information onto the Internet, no functional inter-domain QoS mechanism exists. QoS is at heart a queue management mechanism and only by applying these strategies across an entire routing domain will QoS result in any route optimization at all. No ISP can impose its own QoS methodology on any other.

10 What does QoS mean? Requires a broad definition with six parameters. Definition: The ability of an application to specify required values of certain parameters to the network, values without which the application will not be able to function properly. The network either agrees to provide these parameters or not. 1. minimum bandwidth (1.5 Mbps, 155 Mbps, 1 Gbps) 2. maximum delay (50 milliseconds round-trip delay, 150 ms delay) 3. Jitter (delay variation) (10% of maximum delay, 5 ms variation) 4. Information loss (error effects) (1 in 10,000 packets undelivered) 5. Security (all data steams encrypted and authenticated) Bandwidth is usually first and foremost because for a long time it was the only QoS parameter that could be delivered with any degree of consistency. Jitter how much the end-to-end network latency varies from time-to-time due to effects such as network queuing and link failures, which cause alternate routes to be used. Voice and Video streams cannot realistically resend information and must deal with errors in some other way. Availability (global quality) and Reliability (local network quality) are related. An unreliable IP network means that the network cannot be relied upon to deliver any QoS parameter at all. Service providers will gather the typical values of the characteristics for voice, video, and other data applications and bundle them as a class of service (CoS) appropriate for that traffic flow and charge more for it.