CEENet Workshop

1
LINK STATE PROTOCOLS(contents)

• Disadvantages of the distance vector protocols
• Link state protocols
• Why is a link state protocol better?
• The design of OSPF
• OSPF versus RIP
• Other link-state protocols

2
Disadvantages of the Distance Vector Protocols

• Slow convergence
• Counting to infinity problem
• Lack of variety of metrics
• No possibility of hierarchical routing
• Bad performance in large networks

3
Link State Protocols

• developed to overcome the disadvantages of the
  distance vector protocols
• centralized database describes the topology of
  the whole network
• calculation and routing are still distributed

4
Characteristics

• information about adjacencies sent to all routers
  only when there is a change
• each router maintains an identical database
• a shortest path algorithm is used to find the
  best route
• converge as quickly as databases can be updated

5
Link State Database
LS seq. num.
From To Link Cost A B 1
1 2 A D 3 1
2 B A 1 1 2 B
C 2 1 1 B E
4 1 2 C B 2
1 1 C E 5 1
1 D A 3 1 2 D
E 6 1 1 E B
4 1 2 E C 5
1 1 E D 6 1
1
Link State Announcement (LSA)
From A to B, link 1, distance 1 number 2
6
The Sequence Numbers

• modulo N numbering is used to avoid very large
  numbers
• The sequence numbers vary between 1-N and N-2,
  where N 231
• numbers -N and N-1 are not used
• at start, router uses the negative number 1-N
• when N-2 is reached, the next value will be 0
• the numbers will continue to cycle in the
  positive segment of the sequence space

7
The Sequence Numbers 2

• the lollipop analogy is implemented
• comparison of numbers
• if at least one number is negative - direct
• if both numbers are positive - cyclic
• if the difference is smaller than half a cycle,
  than the bigger number is newer, otherwise the
  smaller number is newer

8
The Flooding Protocol

• every node sends the message on every link except
  the one from where it received the message
• very fast and very reliable, but wastes bandwidth
• used for ordinary traffic in military networks
• messages are sent only when there is a change or
  every 45 minutes

9
The Flooding Protocol 2

• When a link breaks, A and B send the information
  to all other nodes about state of link 1.
• Each node compares this entry in the data base,
  if it is newer than the received message it
  ignores the message, otherwise it updates the
  entry.

10
Bringing up Adjacency
link 1 and link 6 are down and during that time
link 2 breaks
3
When link 6 comes up it has no information about
link 2

• synchronizing databases via comparison of
  sequence numbers
• interesting records - the sequence numbers are
  different or not present in database

11
Securing the Map Updates

• the flooding procedure includes hop-by-hop
  acknowledgments
• the database description packets are transmitted
  in a secure fashion
• each link state record is protected by a timer
  and is removed from the database if a refreshing
  packet does not arrive in due time
• all records are protected by checksum
• the messages can be authenticated, e. g. by
  passwords

12
Shortest Path Routing Algorithms

• Determine the shortest path tree at each node
• Dijkstras algorithm
• complexity (MlogM)
• Bellman-Ford algorithm
• complexity (MN)

M - number of links N - number of nodes
13
Dijkstras Algorithm
Average time needed to compute the routing table
is about 200ms for a 200 node network on a
typical router.
Step N d(B) d(C) d(D) d(E) Initial A 1
1 1 A, B 1 2 1
2 2 A, B, D 1 2 1 2
3 A, B, C, D 1 2 1 2 4
A, B, C, D, E 1 2 1 2
14
Why Is a Link State protocol Better?

• fast loopless convergence
• support of precise metrics and, if needed
  multiple metrics
• support of a multiple paths to a destination
• splitting very large networks in areas

15
Link State ProtocolsDisadvantages

• more memory required
• the link state database is needed in addition to
  the routing tables
• much more complex procedure
• higher probability for a bug in the program

16
OSPF

• link state or SPF technology
• developed by OSPF Working Group of IETF (not
  proprietary)
• designed for TCP/IP Internet environment
• documented in rfc 1247

17
The Design of OSPF

• Strict separation of hosts and routers
• Broadcast networks such as Ethernet or FDDI
• Non broadcast networks as X.25 or ATM
• Splitting very large networks in areas

18
OSPF - Advantages

• fast convergence
• load balancing
• low bandwidth utilization
• optimal path utilization
• authenticated routing updates
• external routes

19
Fast convergence
detection of failure
LSA/SPF
R2
alternate path
N1
N2
XXX
R1
R3
primary path
20
Load Balancing by Multiple Path
equal or proportional cost multiple paths
21
Low Bandwidth Utilization
FDDI Dual Ring

• only changes propagated
• multicast on multiaccess
• broadcast network
• database synchronization

LSA
LSA
XXX
22
Support of Multiple Metrics
Type of metrics

• The algorithm requires

• throughput
• delay
• cost
• reliability

• several metrics for each link
• different routing tables for each link
• presenting selected metric in the packet

23
Optimal Path Utilization
Cost 1
N2
N3
Cost 1
R2
R3
N5
R1
N1
Cost 10
Cost 10
Optimal path is determined by the sum of the
interface cost
R4
Cost 10
N4
24
Optimality Depends on Metric
R2
64Kbs/20ms
64 Kbs/20ms
min. delay
max. throughput
R1
R3
1.5 Mbs / 295ms
1.5 Mbs / 295ms
1.5 Mbs / 295ms
R4
R5
25
The Cost and The Bandwidth

• formula cost 108 /bandwidth in bps

56 Kbps serial link 1758 64 Kbps serial
link 1562 T1 (1.544 Mbps seral link) 65 E1
(2.048 Mbps serial link) 48 4 Mbps token
ring 25
Ethernet 10 16 Mbps token ring 6 FDDI
1
26
Authenticated Routing Updates

• Two possibilities are defined
• no authentication
• simple authentication using passwords
• network administrator can configure a different
  password for each network, e. g. for each
  point-to-point connection or each Ethernet

27
IP Subnetting Support

• Network number
• Variable length subnet mask (VLSM)
• Discontiguous subnets
• Supernets/subnet prefixes

28
Route Summarization
R2
Network Next hop 1
R1
With summarization
Network Next hop 1, A
R1 1, B R1 1, C
R1
Without summarization
R1
1,A 1,B 1,C are stub networks
1,B
1,A
1,C
29
Route Summarization Benefit
30
Route Tagging

• Autonomous System B wants to
• propagate routes from A --gt D, but not propagate
  from C --gt D
• OSPF tags routes with AS input
• the information can be used when redistributing
  routs

31
TOS - Based Routing

• IP header supports 3 bit priority field
• IP header supports 4 special type of services
• bandwidth
• delay
• reliability
• cost
• currently only TOS 0 is supported

32
External Routes

• one gateway (router) to external word
• only advertise default route
• several gateways
• pick one that is closest
• pick one that carry data more efficiently
• external routes are added to the database as
  gateway link state records

33
Broadcast Media Problems

• N neighbors - N(N-1)/2 adjacencies
• Not optimal
• Wasted bandwidth
• does not scale

34
Designated Router

• One per multi-access network
• generates network links advertisements
• assists in database synchronization

Backup designated router
Designated router
35
Broadcast Media

• select a designated router (DR)
• all routers become adjacent to DR
• exchange routing information with DR via
  multicast
• DR updates all the neighbors
• less routing traffic generated

36
Non Broadcast Networks

• for N routers N(N-1)/2 virtual circuits are
  needed to have full connectivity
• may be costly (does not scale)
• designated router plays the same roll as in
  broadcast media
• instead of multicast the LSAs are sent
  point-to-point between the designated router and
  all the others

37
Multiple Areas

• network increases gt increase in
• link state database
• route computation
• volume of messages
• the solution to these problems
• split network into areas and the backbone
• the size of the routing proportional to the size
  of the area, not the whole network

38
Hierarchical Structure

• backbone area needed to connect all the other
  areas

Area 2
Area 1
Area 3
39
OSPF Areas

• group of contiguous hosts and networks
• one database per area
• backbone area (contiguous)
• virtual links
• inter-area routing

Area 3
Area 2
area 0
Area 4
Area 1
40
OSPF Areas 2

• a router has separate LS database for each area
  that it belongs
• all routers belonging to the same area have
  identical database
• SPF calculations are performed separately for
  each area
• LSA flooding is bounded by area

41
Area Link State Database

• area database is composed of
• router links advertisements
• network links advertisements
• summary links advertisements
• AS external advertisements

42
Classification of Routers

• IR - internal router
• ABR - area border router
• BR - backbone router
• ASBR -autonomous system border router

IR
ABR/BR
IR/BR
to other AS
ASBR
43
OSPF Area Mapping

• area can be one or more networks
• area can be one or more subnets
• any combination of networks and subnets possible
  (but bad in practice)
• for summarization subnets must be grouped

44
The Link State Header

• LS age
• time since the LS record was first advertised
• options
• E - external links
• T - TOS (type 0 doesnt support any TOS
• LS type(router link, network link, summary link
  (IP network, summary link, to a border router,
  external link)

31
0
LS age options LS type
Link State ID
Advertising Router
LS sequence number
LS checksum length
45
The Router Links

• summarizes all links that start from the router
• bits 6 and 7 of the first word indicate the type
  of the router

46
The Network Links

• advertised by designated routers

47
The Summary Links

• advertised by area-border routers
• the network mask is followed by a set of metrics

48
The External Links

• advertised by border routers
• required by EGPs
• E indicates that TOS is not comparable with that
  of internal routes

network mask
E, TOS 0 TOS 0 metric
external route tag 0
E,TOS x 0 TOS x metric
external route tag x
- - -
- - -
- - -
E,TOS z 0 TOS z metric
external route tag z
49
Protocols within OSPF

• common header
• hello protocol
• exchange protocol
• flooding protocol
• aging link state record

50
The Common Header
0
31
version (1) type (1)
packet length (2)
Router ID (4)
Area ID (4)
Checksum (2)
autype (2)
Authentication (4)
51
The Hello Protocol
0
31
OSPF packet header, type 1 (hello)
Network mask
Hello interval
Options
Priority
Dead interval
Designated router
Backup designated router
Neighbour
- - - -
Neighbour
52
The Exchange Protocol

• uses database description packets
• asymmetric protocol (master-slave)
• master sends database description packets
• slave sends the acknowledgments

53
The Exchange Protocol 2

• request records
• send in case when sequence number of the LS is
  smaller
• the other router will answer with a LS update

54
The Flooding Protocol

• when a link changes state
• a router responsible for that link issues a new
  version of the link state
• the update is retransmitted in regular interval
  until an acknowledgment is received

0
31
OSPF packet header, type 5 (ack)
Link State advertisements headers
- - - -
55
Aging Link State Records

• old or stale records need to be removed from the
  link state database
• the procedure needs to be synchronized
• the age is set to 0 when the record is first
  issued
• it is incremented on each hop and by 1 every sec.
• when it reaches maxAge, the router needs to
  remove it
• the neighbors have to be informed about this

56
Scaling OSPF

• Rule of thumb
• no more than 150 routers /area
• Reality
• no more than 500 routers/area
• Backbone area is an area
• always marked as area 0
• proper use of areas reduces bandwidth
• summarized routes
• instability is limited within the area

57
Route Redistribution
RIP Domain
OSPF Domain

• UNIX host ruining routed

• the router redistributes RIP into OSPF and vice
  versa

58
Conclusions

• more complex than RIP
• the documentation is five times thicker
• the management needs more information
• the implementation needs more code
• why design such complex procedure?
• routing is important
• requires less signalization messages
• compute better routes

59
Conclusions 2

• OSPF is not a perfect protocol
• IETF keeps making it better
• O in OSPF stands for Open
• OSPF is not the only link state protocol
• IS-IS protocol is part of OSI routing framework
  for CLNP
• similar in design to OSPF
• uses different terminology