320 likes | 343 Views
Explore the world of routing protocols, interfaces, tables, and autonomous systems. Learn about OSPF, RIP, and how routing decisions are made within large network systems using advanced algorithms and protocols.
E N D
What Are Routers? • Routers are an intermediate system at the network layer that is used to connect networks together based on a common network layer protocol
Main routing components include: • Routing interface • Routing protocol • Routing table
What Are Routing Interfaces? • A routing interface is an interface over which IP packets are forwarded
What Are Routing Protocols? • A routing protocol is a set of messages that routers use to determine the appropriate path to forward data OSPF RIP • Designed for small to medium-size networks • Uses a routing table • Easier to configure and manage • Does not scale well • Designed for large to very large networks • Uses a link-state database • Complex to configure and manage • Operates efficiently in large networks
What Are Routing Tables? • A routing table is a series of entries called routes that contain information about the location of the network IDs in the internetwork
Interior Routing Protocols • The internet is a connection of many different networks. We have seen how to navigate around this big system • How do we make our routing decisions within a network?
Autonomous Systems • Owned and administered by a single organization • Frequently a backbone system connecting a number of LANs (e.g., University) • Usually connected to the Internet at one or more points • Use routers to connect the LANS
Autonomous System LAN 1 LAN 2 LAN 3 LAN 4 LAN 5 LAN 6 Backbone link
Large Autonomous System • The Autonomous System of a large organization could include many routers, and cover a wide geographical area • Telstra’s AS includes about 400 routing nodes
Routing Problem in AS • Much of the traffic generated within the AS would be addressed to stations within the AS • Traffic to or from the Internet needs to reach the correct LAN • If it is a WAN, it might be used by the Internet to conduct packets over long distances
Routing Problem in AS • Any of these situations reduce to the following problem. What is the “best” way to get a packet from router A to router B? • Router A is the first router to receive the packet • Router B is the router connected to the LAN where the destination station is connected or is the router which passes the packet onto its next hop
Routing Problem in AS • We assign a weight to each link • To avoid congestion, or balance load, the weight is a function of the utilization of the link, eg u2 • Usually, traffic will be different in each direction, in the same link • The same link must then have two different weights for the two directions
Weight Function • Why not use a weight function which is simply equal to the utilization? • An extremely simple example will show why - two routers and two links
Weight Function • Total traffic is 100 Mbps, bandwidth of each link is 100 Mbps • Might split traffic 50-50, or, say, 75-25 • Sutilisation = 0.5 + 0.5 = 1 (first case) • Sutilisation = 0.75 + 0.25 = 1 (second case) • Cannot discriminate between these cases • Cannot tell us which is the “best”
Weight Function • However, if weight is u2 • Su2 = 0.52 + 0.52 = 0.25 + 0.25 = 0.5 (first case) • Su2 = 0.752 + 0.252 = 0.5625 + 0.0625 = 0.625 (second case) • Prefer first case, ie 50-50 (80% of 0.625) • Higher powers, eg u4, give stronger discrimination
Weight Function u4 • Su4 = 0.54 + 0.54 = 0.0625 + 0.0625 = 0.125 (first case) • Su4 = 0.754 + 0.254 = 0.3164 + 0.0039 = 0.3203 (second case) • Prefer first case, ie 50-50 • First case gives total weight 39% of second case
One Solution • Based on Bellman-Ford (distance vector) algorithm • Neighboring routers send each other their routing tables
RIP Protocol • Routing Information Protocol (RIP) uses weight of 1 for each link min hop count • Max hop count allowed is only 16 - after that, RIP assumes destination is unreachable • Uses elapsed time as cost function
Dijkstra’s (Link State) Algorithm • Every router has a map of the complete network • All link costs are available to every router • Router computes best route, independently of other routers • Routers use Dijkstra’s algorithm for this computation
Dijkstra’s Algorithm • Algorithm to find lowest cost from router ‘s’ to all other routers in the network • Algorithm proceeds by looking at routers directly connected to router “p” • Step 1. Set RC(s) = 0, mark this RC as permanent (route cost to router “s”). Set RC to all other routers = . They are temporary. Set p = s.
Dijkstra’s Algorithm • Step 2. For all routers, “x”, connected to router p, and which have temporary RCs, update the RC as follows RC = min(present RC, RC(p) + c(p,x)) where c(p,x) = cost of link from p to x • Step 3. Find minimum of all temporary RCs. Let this be RC(i)
Dijkstra’s Algorithm • Step 4. Make this RC permanent. Set p = i • Step 5. If all RC are permanent, stop. Otherwise, return to step 2
Example Network 4 2 3 1 3 7 6 3 3 1 6 5 4 4 5 4
Two Way Traffic • In practice, we will find different traffic flow in the two directions through one link • If the link cost is a measure of flow, speed or congestion, it should be different in the two directions • Thus a single link, as shown in diagrams so far, should be replaced with two directed links
Network with Directed Links 4 1 2 2 1 4 5 1 3 2 4 7 3 6 4 5 3 c(1,2) c(2,1)
Effect of Routing Tables on Traffic Pattern • A single change in a routing table produces a change in the traffic pattern • This produces a change in the link costs • Which may produce a change in a routing table, and so on • The system may settle down, or may continue in a “limit cycle”
Possibility of alternate routes • Routing table normally defines one route for any source/destination pair • Can achieve better utilisation of network if two (or more) routes are used • Effect is greater if network is small
OSPF • The most popular link state protocol is OSPF (Open Shortest Path First) • Developed by the IETF • Capable of calculating more than one “best” path
Similar Protocols • Similar protocols are Intermediate Host to Intermediate Host (IS-IS) • IS-IS is designed for OSI networks • Extended Interior Gateway Routing Protocol (EIGRP) • Developed by Cisco before OSPF came out