25-Routing-OSPF-1-Introduction - OSPF – Introduction The...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: OSPF – Introduction The IETF Routing Master Part 1 (C) Herbert Haas 2005/03/11 1 “Dijkstra probably hates me” Linus Torvalds in kernel/sched.c 2 "Open Shortest Path First" Official (IETF) successor of RIP RIP is slow RIP is unreliable RIP produces too much routing traffic RIP only allows 15 hop routes Yes, RIP is bad Voodoo... OSPF is a link-state routing protocol Inherently fast convergence Designed for large networks Designed to be reliable (C) Herbert Haas 2005/03/11 3 OSPF was developed by IETF to replace RIP. In general link-state routing protocols have some advantages over distance vector, like faster convergence, support for lager networks. Some other features of OSPF include the usage of areas, which makes possible a hierarchical network topologies classless behavior,there are no such a problem like in RIP with discontiguous subnets. OSPF also supports VLSM and authentication. 3 OSPF Background OSPF is the IGP recommended by the IETF "Open" means "not proprietary" Dijkstra's Shortest Path First algorithm is used to find the best path OSPF's father: John Moy Version 1: RFC 1131 Version 2: RFC 2328 (244 pages !!!) And a lot of additional OSPF related RFCs available... (C) Herbert Haas 2005/03/11 4 The Internet Engineering Task Force (IETF) strictly recommends to use OSPF for Interior Gateway routing (i. e. within an AS) instead of RIP or other protocols. Integrated IS-IS is an alternative routing protocol but not explicitly recommended by the IETF. Note that IS-IS has been standardized by the ISO world. Both (Integrated) IS-IS and OSPF use Dijkstra's famous Shortest Path First (SPF) algorithm to determine all best paths for a given topology. OSPF version 2 has been specified in RFC 2328. Note that there are a lots of additional RFCs around OSPF. Use http://www.rfc-editor.org/rfcsearch.html to find them all. 4 Dijkstra's SPF Algorithm Used in graph theory Very efficient Calculates all paths to all destinations at once Creates a (loop-free) tree with local router as source See SPF section for more details (C) Herbert Haas 2005/03/11 Edsger W. Dijkstra (1930-2002) 5 The Dijkstra's SPF algorithm is generally used in graph theory and was not invented especially for IP routing. The most interesting point on the SPF algorithm is its efficiency. SPF is capable to calculate all paths to all destinations at once. The result of the SPF algorithm is a loop-less tree with the local router as source. 5 OSPF Ideas Metric: "Cost" = 108/BW (in bit/s) Therefore easily configurable per interface OSPF Routers exchange real topology information Stored in dedicated topology databases Now Routers have a "roadmap" Instead of signposts (RIP) Incremental updates NO updates when there is NO topology change (C) Herbert Haas 2005/03/11 6 In the Cisco IOS implementation starting with 11.2, the cost is calculated automatically by the simple formula 10,000,000/BW. Here the bandwidth parameters on a routers interface are used, thus it is especially important to configure it on the serial interfaces. In other OSPF implementations cost must be configured manually for each of the interfaces. OSPF—and other link state protocols—exchange true topology information which is stored in a dedicated database by each router. This database acts like a "roadmap" and allows a router to determine all best routes. Note that once OSPF got the toplogy database there is no need to exchange further routing traffic—unless the topology changes. In this case only incremental updates are made. 6 What is Topology Information? The smallest topological unit is simply the information element ROUTER-LINK-ROUTER So the question is: Which router is linked to which other routers? R2 R1 R5 (C) Herbert Haas Link Database: R3 R4 = R1– R2 R1– R5 R2– R3 R2– R4 R4– R5 The Link Database exactly describes the roadmap 2005/03/11 7 Obviously the dots are routers and the links between the routers are actually networks. The basic idea of OSPF and the topology table is that simple. OSPF is actually much more complicated. There are 5 types of networks defined in OSPF: pointto-point networks, broadcast networks, non-broadcast multi-access networks, point-tomultipoint networks, and virtual links. Furthermore it is reasonable to divide the topology into multiple "areas" to increase performance ("divide and conquer"). These are the reasons why OSPF is a rather complex protocol. This is explained later. 7 OSPF Routing Updates The routing updates are actually link state updates Parts of link state database are exchanged Instead of parts of routing table (RIP) Applying the SPF algorithm on the link state database, each router can create routing table entries by its own (C) Herbert Haas 2005/03/11 8 The Links State Updates LSUs are sent in a special packets – Link State Advertisments LSAs. There are several types of LSAs, depending on what kind of information is sent and which router originated it. 8 OSPF Protocol All OSPF messages are carried within the IP payload ("raw IP") Protocol number 89 Error recovery and session management is covered by OSPF itself Multicast address "All OSPF routers" (C) Herbert Haas 2005/03/11 9 LSUs are encapsulated in IP packet directly, unlike RIP where we have an additional UDP overhead. IP is not reliable by itself, but OSPF updates are transmited reliable using Link State Acknolegements LSAck. There are 2 multicast addresses which are reserved for OSPF, – for all OSPF routers and for designated and back designated OSPF routers. 9 LSA Flooding LSA's are small packets, forwarded by each router without much modifications through the whole OSPF area (!) Much faster than RIP updates RIP must receive, examine, create, and send Convergence time Detection time + LSA flooding + 5 seconds before computing the topology table = "a few seconds" (C) Herbert Haas 2005/03/11 10 When the router gets a new information in its link state database it should send this information to all adjacent routers – flood. The packets are small, only the changes are sent and not the whole database. All other routers do the same, receive new information, update link state database, flood changes to others. 10 OSPF Overview Large networks: "Divide and conquer" into areas LSA-procedures inside each area But distance-vector updates between areas Additional complexity because of performance optimizations Limit number of adjacencies in a multi-access network OSPF Limit scope of flooding through "Areas" Deal with stub areas efficiently Learn external routes efficiently Realized through different LSA types Fast convergence, almost no routing traffic in absence of topology changes (C) Herbert Haas 2005/03/11 11 Perfomance is very important with OSPF, to run SPF algorithm a CPU resources are required, to store a link state database an additional memory, compared to RIP we need much more router‘s resources. Some additional improvements were made to OSPF in order to improve performance. Areas were introduced to limit the flooding of LSAs, Stub Areas to minimize a link state database and routing tables. Several types of LSAs were implemented: Type 1 – Router LSA Type 2 – Network LSA Type 3 – Network Summary LSA Type 4 – ASBR Summary LAS Type 5 – AS External LSA Type 6 – Group Membership LSA Type 7 – NSSA External LSA and others 11 ...
View Full Document

This note was uploaded on 02/13/2010 for the course ENGINEERIN Network taught by Professor Khaled during the Spring '10 term at Abilene Christian University.

Ask a homework question - tutors are online