DD2490 p Link state routing and OSPF. Olof Hagsand KTH/CSC

Transcription

1 DD490 p4 00 Link state routing and OSPF Olof Hagsand KTH/CSC

2 Literature RFC 3: Section except.. Section 3 (areas), but only last two paragraphs of 3.5

3 Link state routing Each router spreads information about its links to its neighbours. This information is flooded to every router in the routing domain so that every router has knowledge of the entire network topology. Using Dijkstra's algorithm, the shortest path to each prefix in the network is calculated

4 Comparison with Distance vector Link state uses a distributed database model Distance vector uses a distributed processing model Link state pros: More functionality due to distribution of original data, no dependency on intermediate routers Easier to troubleshoot Fast convergence: when the network changes, new routes are computed quickly Less bandwidth consuming Distance vector pros: Less complex easier to implement and administrate Needs less memory

5 Comparison with IS IS Both are link state protocols IS IS has a longer history from Digital via OSI OSPF is newer and developed in IETF Area difference OSPF defines area boundaries between interfaces IS IS defines area boundaries between nodes IS IS areas leads to simpler configuration Protocol dependency IS IS can run many protocols (IPv6, CLNP) OSPF only IPv4, ( OSPFv3 supports IPv6) OSPF is implemented on more platforms and more deployed IS IS often popular among backbone networks

6 Original OSPF requirements A more descriptive routing metric Equal cost multipath Two level routing scheme: areas Separate internal and external routes Multiple best paths: load balance Routing hierarchy Link metric: External routes Security Cryptographic authentication

7 Basic OSPF. The hello protocol Is there anybody out there? Detection of neighboring routers Election of designated routers. The exchange protocol Exchange database between neighbours 3. Reliable flooding When links change/age send: update to neighbours and flood recursively. 4. Shortest path calculation Dijkstra's algorithm Compute shortest path tree to all destinations

8 Dijkstra's shortest path computation From the link state database, compute a shortest path delivery tree using a permanent set S and a tentative set Q: ) Define the root of the tree: the router ) Assign a cost of 0 to this node and make it the first permanent node. 3) Examine each neighbor node of the last permanent node. 4) Assign a cumulative cost to each node and make it tentative. 5) Among the list of tentative nodes: Find the node with the smallest cumulative cost and make it permanent. If a node can be reached from more than one direction, select the direction with the smallest cumulative cost. 6) Repeat steps 3 to 5 until every node is permanent.

9 Dijkstra pseudo code fun ct ion Dijkstra(G, w, s) for e ach vertex v in V[G] // Initializations d[v] := infinity previous[v] := undefined d[s] := 0 S := empty set // S: Permanent set Q := set of all vertices // Q: Tentative set whi le Q is not an empty set // The algorithm itself u := Extract_Min(Q) S := S union {u} for ea ch edge (u,v) outgoing from u if d[v] > d[u] + w(u,v) // Relax (u,v) d[v] := d[u] + w(u,v) previous[v] := u (from Wikipedia)

10 Example network N A B N5 D N4 N N3 3 C 3 N6 E N7 5 N F Metric (may be assymmetric) N9 N0

11 Corresponding link state database N A B N N3 3 N5 D N9 0 N4 C 3 N6 N6 s Designated Router E N7 0 N F N0 5

12 Dijkstra s algorithm computed N A B N5 D N9 Final shortest path delivery tree from A N N3 3 0 N4 C 3 N6 E N7 N F N0 5

13 OSPF Encapsulation OSPF runs directly on IP Needs its own reliable protocol No port numbers Need to run as root raw sockets No checksum The flooding protocol Computes its own checksum or digest Since it runs on IP (IS IS runs on the link level) OSPF messages can be routed tunneled or routed by some other protocol

14 OSPF header Version # Type Packet length Router ID Area ID Checksum AuType Authentication Authentication IP OSPF common header OSPF specific header

15 OSPF common header Version for IPv4 3 for IPv6 Router ID Type First configured interface or highest loopback IP Area ID Backbone: Checksum Hello DD database description Link State Request Link State Update NULL Link State Acknowledge Simple password Cryptographic authentication Standard IP checksum AUtype

16 Cryptographic authentication Key ID Auth Data Len Cryptographic sequence number Authentication field (see figure) Checksum field set to zero Create message digest from complete packet Eg, MD5 Append digest to packet Set key (if many keys are used), seq#, and digest length. IP OSPF common header OSPF specific header Digest

17 OSPF Adjacency How does a router know who its neighbours are? By sending each other Hello packets Multicast to AllSPFRouters ( ) Sent every 0 seconds Hello interval Three failed Hello attempts result in a link failure report. Router dead interval

18 The Hello packet Network Mask HelloInterval Options Rtr Pri RouterDeadInterval Designated Router Backup Designated Router Neighbor IP OSPF common header Hello

19 Adjacency on a broadcast network N squared problem: too many adjacencies if the network is fully meshed: n(n )/ We elect one router to represent the network We elect one router to take over in case of failure Designated router (DR) Backup designated router (BDR) OSPF communication From a router to the DR and BDR The DR sends messages to other neighbours Multicast: AllDRouters ( ) Multicast: AllSPFRouters ( ) Data traffic still forwarded directly!

20 Electing designated router Election algorithm First router always DR Second router always BDR Only in case of failure change DR/BDR Elect DR and BDR from router priority If equal prio => Highest router ID Routerprio is 0 => can never be DR/BDR Why is a BDR necessary? For fast failover if DR fails The BDR runs in parallell with the DR

21 Variants of multi point networks Broadcast networks The link layer is broadcast capable Non Broadcast Multiple Access (NBMA) Any two routers can communicate but no broadcast X.5, ATM, FR Neighbour detection via configuration Point to Multipoint Not all routers can communicate Packet radio, cloud of point to point links No DR/BDR, treated more like many point to point protocols.

22 Database exchange When two routers has established adjacency, the databases need to get synchronized. First: mutually send summary to each other's databases. The Database description packet includes a list of Link state headers Then: Request explicitly database entries Only database headers not actual entries. Link State requests Last: Send database entries Link state updates

23 Link State Advertisements LSAs are the elements of the distributed database A router describes its environment in the form of networks that it is connected to Fundamental task in OSPF: Also called LSPs (Link State packets) Distribute the LSAs to all nodes in a reliable way Then, each node can compute Dijkstra on the same database

24 Reliable flooding Every router spreads its LSAs to all its peers All routers forward the LSAs to its other peers That is, all information about its own links LSAs are acknowledged When a link changes, a new instance of the LSA is distributed Periodic updates every 30 minutes Flood a new instance

25 Reliable flooding example The originator floods an LSA over the network An LSA update of same instance is taken as an implicit ACK

26 Flooding in a transit network From peer to DR and BDR From DR to all others If the BDR does not hear an update from the DR, it assumes the DR has crashed and takes over DR BDR

27 LSA header Every LSA has a common header The rest is different depending on LSA type LSA headers appear in DD, LS update and LS ack LS age Options LS type Link State ID Advertising Router LS sequence number LS checksum length LSA hdr IP OSPF common header LSA LSA body LSA... LSA n

28 The type field. Router LSA Transit, stub, and point to point connections between routers. Network LSA Originated by DR. Contains list of routers connected to shared medium. 3. Network Summary LSA 4. ASBR Summary LSA 5. AS External LSA 6. Group Membership LSA (MOSPF) 7. NSSA external information LSA. External attributes LSA

29 The link state ID field Different for different types: Type Router LSA: Router ID Type Network LSA: IP address of DR...

30 The age field An LSA is valid up to 30 minutes When an LSA reaches 30 minutes, the originating router makes a new instance If not refreshed, the LSA will be deleted after one hour (MaxAge) Premature aging is used to flush LSAs from the database New instance: increment sequence number Age == MaxAge is the same as delete! If deleted by any router that router floods the LSAs with MaxAge To ensure that all deletes the LSA at the same time

31 Sequence number An originating router typically increments the sequence every 30 minutes when age has expired Larger sequence => more recent LSA instance But how do you define larger if sequence number may wrap? circular lollipop linear

32 Sequence numbers Original ARPANET: Circular OSPFv: Lollipop OSPFv: Linear Initial sequence number: 0x Max sequence number: 0x7fffffff When an LSA sequence number reaches Max, the router must delete the LSA By flooding of a prematured aged LSA And then reintroduce the LSA But sequence number is 3 bits, if router updates sequence # every 5 seconds it takes 600 years to wrap around!

33 Metric The metric is dependent on LSA and is not in the common header The metric is a scalar It can mean anything: hops,, delay, load,... Metrics are asymmetric CISCO's default metric is: 0^ / <linkbw> Eg 0Mb eth has metric 0 E (serial Mbps) has metric 50 Juniper does not have this

34 LSA type : Router LSA A list of links that a true router is connected to Link to a Point to point network Link to Transit networks Link to Stub networks Broadcast, NBMA or point to multipoint No other router Virtual link Tunnel to other router Used in error cases and to keep the backbone connected LSA hdr Router LSA hdr Link # Link # Link #n

35 Router LSA: Stub network RFC 3, fig a Stub network: only one entry point Represented by IP address and network mask Note that the database arrow is not bidirectional Physical network LSA (database) view RT7 RT7 N3 N3 RT7's router LSA (part of a LS update) Packet view LSA hdr Router LSA hdr Stub Link: N3

36 Router LSA: Point to point For unnumbered interfaces, Ia and Ib are omitted Ia and Ib does not need to be on common subnet OSPF obscurity: RT points to Ib! In practice (OSPF + JunOS): both addresses in same subnet. Ia Ib RT RT RT Ia Ib RT RT's router LSA LSA hdr Router LSA hdr Router link:rt Stub Link: Ib RT's router LSA LSA hdr Router LSA hdr Router link:rt Stub Link: Ia

37 Router LSA: Transit network The link points to a transit network's IP address The address of the designated router RT3 RT3 RT4 RT4 RT4 RT6 N RT5 N RT3's router LSA LSA hdr Router LSA hdr RT6 Transit link:n

38 Router LSA: Virtual link The link points to a remote router connected by an IP network Similar to point to point, but remote peer is not physically connected Used to keep the backbone connected We will talk more about virtual links in the area section

39 LSA Type : Network LSA Links of a transit network distributed from a designated router The designated router distributes the information on behalf of the connected routers Metric on entry to network but zero cost to leave Example: (RT3 is DR) DR RT3 RT3 RT4 RT4 RT4 RT6 N RT5 N RT3's network LSA LSA hdr Network LSA hdr RT6 RT3, RT4, RT5, RT6

40 External routes An external route is a prefix that OSPF has learnt from another protocol (or static route) Has been redistributed into OSPF External routes come in two flavors based on the metrics: External Type (E): use same metrics as internal External Type (E): external metric takes precedence If RIP routes are imported as E, and OSPF uses hop count metric, then OSPF and RIP can work seamlessly BGP routes are imported as E, where metric is AS path length

41 AS External LSA (Type 5) But how are the external routes communicated to the network? Router and network LSAs are not applicable AS External LSAs Originated by AS boundary routers Announces an external particular prefix Redistributed route from another protocol A forwarding address (may be different than AS boundary router) External route tag Eg an BGP AS path would enable the use of OSPF instead of IBGP AS External LSAs are flooded throughout the AS RT3 N RT3 N ASBR ASBR RT3's AS External LSA LSA hdr AS External LSA hdr N

42 From network to FIB: Example Network > Database > OSPF Routing table > RIB > FIB

43 Network example N 3 RT 3 RT4 N3 N N4 N 6 6 RT RT5 7 N3 6 RT3 RT6 Ia 7 N4 N5 Ib 6 RT0 3 N 3 N9 H RT9 0 6 RT N6 RT N RT 4 N0 N7 RFC 3 fig 9 RT7

44 Building a database With LSA type and, we can build databases by combining the LSA views * * T O * * **FROM** RTRTRTRTRTRTRTRTRTRTRTRT N3N6NN RT 0 RT 0 RT3 6 0 RT4 0 RT5 6 6 RT6 7 5 RT7 6 0 RT 0 RT9 0 RT RT 0 0 RT 0 N3 N 3 N3 N4 N6 N7 4 N 3 N9 N0 N 3 N N3 N4 N5 9 H 0 RFC 3 fig 3

45 N4 Database, graphical form N 3 RT 0 N3 0 N 0 RT4 7 RT3 RT5 0 3 RT N3 N LSA Type 5: AS External RT6 7 7 Ib N4 7 N5 5 Ia 6 5 RT7 RT0 3 N 3 RT9 0 H 0 RT N N RT N6 0 RT 4 N0 N7 9

46 N4 Shortest path tree for RT6 N 3 RT 0 N N3 0 N RT5 RT4 0 N3 3 RT 6 RT3 6 RT6 7 7 N4 Ib N5 Ia 5 RT7 RT0 3 N 3 RT9 H N9 0 RT N N6 0 0 RT RT 4 N0 N7 RFC 3 fig 5 9

47 Building a routing table Local routing table (RIB) computed from Dijkstra shortest path calculation Next hop routing: only nexthop router even if complete path is known Example: RT6 Local destinations RFC 3, table Destination Next Hop Distance N RT3 0 N RT3 0 N3 RT3 7 N4 RT3 Ib * 7 Ia RT0 N6 RT0 N7 RT0 N RT0 0 N9 RT0 N0 RT0 3 N RT0 4 H RT0 RT5 RT5 6 RT7 RT0 Remote destinations(type ) RFC 3, table 3 Destination Next Hop Distance N RT0 0 N3 RT5 4 N4 RT5 4 N5 RT0 7

48 OSPF Network Topology Area 0 is the backbone area. All (inter area) traffic goes via the backbone. All other areas are connected to the backbone ( level hierarchy) A Border area router (ABR) has one interface in each area. An AS Boundary Router (ASBR) attaches to other AS:s Backbone router at least one interface in backbone area ASY ASZ AS boundary router: External routing AS Area 0 ASX Internal router Area Border Router: Interfaces in different areas All areas connected to backbone area Area Area Area 3 Internal router + ASBR

49 OSPF Areas Divides the OSPF domain into smaller zones Smaller link state database in each zone Also decreases signaling traffic Routers have limits on processing power and memory Router CPUs are typically much slower than PCs CISCO used to recommend ~0 routers as a limit in a single area You need a large network to benefit from areas Typical large companies Example: KTHLAN using OSPF with 5 0 routers used to have areas but now only uses area 0. However, areas are less used today. More often divide your internal network into BGP confederations, for example

50 Smaller database Using areas makes the database smaller That is, fewer and more compact LSAs The destinations inside the area is still fully described by type router and type network LSAs Full Dijkstra algorithm But destinations outside the area are summarized Only the (cumulative) metric and prefix necessary Not full link state This leads to a smaller database and less processing to compute shortest path

51 Route summarization When the details of an area has been hidden it makes sense to aggregate the prefixes Typically, all networks within an area, can be summarized into one LSA Routes can also be summarized at redistribution to/from another protocol The metric uses the max of all summarized metrics In the example, area 's routes are summarized: N9 N, H With max cost (to H)

52 Summary LSAs: types 3 and 4 To distribute the more summary information, we need two new LSAs Type 3 Network summary LSA Destination, network mask and cumulative metric N Type 4 ASBR Summary LSA Same as type 3, but destination is an AS boundary router Next hop for external routes ABSR Why is LSA type 4 needed? Because type 5 AS external LSAs are flooded throughout the AS, but the ASBR might not be visible from inside an area

53 N4 N 3 RT 3 RT4 N3 N N 6 6 RT RT5 7 N3 6 RT3 RT6 Ia 7 N4 Area N5 Ib RT0 3 N 3 N9 H RT Area 3 RT9 0 N0 N6 RT N RT 4 Area RFC 3 fig 6 N7 RT7

54 Virtual links The backbone must be logically connected But it does not have to be physically connected You can use virtual links (tunnels) to make the backbone virtually connected Traffic passing in the backbone may then physically use a non backbone area: this is called a transit area. Example: A virtual link is (manually) configured between RT0 and RT For robustness, RT7 and RT may also have a virtual link

55 Virtual link example A virtual link is (manually) configured between RT0 and RT For robustness, RT7 and RT may also have a virtual link Why? Area is now a transit area RT0 3 N6 RT N RT 4 Area N7 RT7

56 Example RFC 3, section 3 Using Area 0 and as examples shows Note : The Area Border Routers (RT3 and RT4) injects summaries both Into Area from the backbone and other areas Into area 0 (backbone) from area Note : The external routes are flooded through all areas Note 3: Area has two points of exits Internal routers can make intelligent decisions, and load balance between exit points Example: RT uses RT4 to N6, RT3 to N0, and load balance to N!

57 N4 Backbone Database: RFC 3 fig N RT4 N N3 RT3 N3 RT5 7 6 N 6 6 RT6 7 N4 Ib Area N5 5 Ia 6 5 Max of all individual metrics N9-N, H Area 3 3 RT7 RT0 RT N N6 Area N7 9

58 Area 's database: RFC 3 fig 7 N4 LSA Type 5: AS External N3 RT5 LSA Type : Router RT 0 0 N 3 RT RT7 9 N5 N LSA Type 4: ASBR Summary LSA Type : Network N N3 0 Ia Ib RT4 0 0 N6 RT3 N7 N N4 Area N9-N, H LSA Type 3: Network Summary

59 Stub areas A problem with flooding external LSAs: Suppose many external routes are injected into OSPF Maybe the core carries transit traffic (between other AS:s) But large parts of the areas do not High performance routers Simpler routers This will give a high burden on the smaller routers In a stub area, the ABR does not flood external LSAs into the area Instead, one (or many) default route is injected Then all external traffic must use the default route announced by the ABRs But the inter area traffic are still announced by summaries from the ABRs

60 Stub area example N4 N 3 RT N RT5 7 default RT Area RT4 N3 3 N3 N RT6 RT3 N4 N5 6 9 RT7

61 Motivation for NSSA Sometimes, the restrictions on stub areas are too strict: You would like to import a limited number of external routes Example: You want to block large routing tables from transit traffic, but want to import a small number of routes But in stub areas, you cannot import any external routes.

62 Example: NSSA motivation A Peering with other AS C Large amount of transit routes Area Area 0 Want to import routes from C but not from A and B Peering with other AS B

63 Not So Stubby Area (NSSA) RFC 30 NSSA allows to inject external routes into a stub area LSA type 7 are spread through the NSSA At ABRs, the Type 7 LSA are translated to Type 5 (External AS LSA) and spread through the AS But other External AS LSAs are still not inserted in the NSSA

64 Example: NSSA solution A Peering with other AS C Routes from C spread as Type-7 LSA Area Area 0 Routes from C translated to Type-5 LSAs Peering with other AS B

65 Totally stub areas Totally stub area Do not distribute inter area routes into an area Just use default route CISCO specific NSSA totally stub area Combination of NSSA and totally stub area

66 Summary of LSAs: regular areas Regular area Area 0 Regular area ABR > -----> 3* -----> -----> >x ----->x -----> -----> > 5 Note: Area 0 ABR (stub links) 3* <----3 <----3 <----4 <----5 <----4 <----- <---- <---- < < < (*) Only stub links of type translated to type 3 Type 5 passes through but generates a new type 4 (ASBR) Type 3 and 4 from other areas passes through but changes origin (to ABR) Summaries (3 and 4) are not forwarded into backbone

67 Summary of LSAs: stub areas Stub area Area 0 Stub area ABR > -----> 3* -----> -----> >x ----->x ----->x Note: Area 0 ABR (stub links) 3* <----- < <----- < <----- < x< x< /0 <----- All type 5 blocked. Replaced with default route (in a type 3 LSA)

68 Summary of LSAs: not so stubby areas Stub area Area > -----> 3* -----> -----> >x ----->x ----->x -----> -----> > 5 Stub area Area 0 3* <----- < <----- < <----- < x< x< /0 <----x<---- 7

69 Opaque LSA Option RFC 370 For Future extensibility Standard LSA header Followed by application specific information Three new LSA, difference in scope: Type 9 LSA: Link local scope Type 0 LSA: Area local scope Type LSA: AS local scope

70 Summary This was OSPF essentials But there are many more issues, for more reading consult: RFC 3 J Moy, OSPF Anatomy of an Internet Routing Protocol Lots of vendor documentation

71 Traffic Engineering extensions RFC 370 It adds bandwidth and administrative constraints So that a (network) manager can control traffic in more detail Distribute it in an area Uses Type 0 opaque LSA, area scope Call it Traffic Engineering LSA The LSA payload contains nested TLVs, for example: Traffic engineering metric Maximum bandwidth Maximum reservable bandwidth Unreserved bandwidth Administrative group

72 OSPFv3 OSPF for IPv6 is OSPFv3 Unchanged: Flooding, DR election, area support, SPF calculations, etc Authentication removed (use IPSEC) New LSAs for IPv6 addresses Addressing semantics removed from basic LSAs and msgs Avoid IPv4/IPv6 addresses prefer RouterID Network protocol independence Renaming: Type 3 summary LSA > Inter Area prefix LSA Type 4 summary LAS > Inter Area router LSA