IS-IS and OSPF: Network Design

Comparisons and Considerations

Jeff Doyle
Objectives

„ Understand protocol similarities

and differences
„ Understand protocol strengths
and weaknesses
„ Make more informed design decisions
Agenda

„ Overview of link-state protocols

„ A parallel history of IS-IS and OSPF
„ Comparative analysis of IS-IS and OSPF
„ Design considerations
„ Less-tangible considerations
„ Conclusions
Agenda

„ Overview of link-state protocols

„ A parallel history of IS-IS and OSPF
„ Comparative analysis of IS-IS and OSPF
„ Design considerations
„ Less-tangible considerations
„ Conclusions
In the Beginning Was Distance Vector...

„ Also known as Bellman-Ford, Ford-Fulkerson

„ Very simple algorithm
„ Distance Vector Protocols include
z RIP
z BGP (but usually called Path Vector)
z Cisco’s IGRP
z Cisco’s EIGRP
Routing by Rumor

„ Distributed calculation
„ Each router knows only what its neighbor tells it

R1 R2 R3 R4
Subnet
A

A, 0 A, 1 A, 2

Prefix Hops Via Prefix Hops Via Prefix Hops Via Prefix Hops Via
A 0 Local A 1 R1 A 2 R2 A 3 R3
Problems with Distance Vector

„ Slow convergence
z
z A
A direct
direct result
result of
of the
the distributed
distributed calculation
calculation
z
z Triggered
Triggered updates
updates help
help
z
z Kludges
Kludges such
such as
as hold-down
hold-down timers
timers reduce
reduce transient
transient errors,
errors,
but
but increase
increase convergence
convergence time
time
„ Single-hop routing loops
z
z Solution:
Solution: split
split horizon
horizon
„ Counting to infinity
z
z Solution:
Solution: make
make infinity
infinity finite
finite
„ Synchronized periodic updates
z
z Solution:
Solution: update
update jitter
jitter timers
timers
Link-state Protocols

„ Also known as shortest path, Dijkstra

„ Algorithm based on graph theory, providing
better loop avoidance
„ Local computation means faster convergence
„ Link-state protocols include
z OSPF z IBM APPN
z IS-IS z MPLS CSPF
z ATM PNNI
Fundamental Link-state Concepts

„ Adjacency
„ Information flooding
„ Link-state database
„ SPF calculation
SPF Calculation Example
Example Topology Link-state Database
R1-R2, 1
R2
R1-R5, 2
2 1 R2-R1, 2
R1 2 R2-R3, 1
R3 R2-R4, 2
1
3
R3-R2, 3
2
R3-R4, 2
2
R4-R2, 4
R4
3 R5 4 R4-R3, 4
5 4
4 R4-R5, 4
R4-R6, 2
3 2
R5-R1, 3
R5-R4, 5
R6
R5-R6, 3
2 1
R6-R4, 1
R6-R5, 2
SPF Calculation Example
Link-state Database Tentative Path
R1-R2, 1 LS Entry Cost to Root R2-R2, 0
R1-R5, 2 R2-R1, 2 2 R2-R3, 1
R2-R1, 2 R2-R3, 1 1
R2-R3, 1 R2-R4, 2 2
R2-R4, 2
R3-R2, 3
R3-R4, 2
R4-R2, 4
R4-R3, 4
R4-R5, 4
R4-R6, 2
R5-R1, 3
R5-R4, 5
R5-R6, 3
R6-R4, 1
R6-R5, 2
SPF Calculation Example
Link-state Database Tentative Path
R1-R2, 1 LS Entry Cost to Root R2-R2, 0
R1-R5, 2 R2-R1, 2 2 R2-R3, 1
R2-R1, 2 R2-R4, 2 2 R2-R1, 2
R2-R3, 1 R3-R4, 2 3
R2-R4, 2
R3-R2, 3
R3-R4, 2
R4-R2, 4
R4-R3, 4
R4-R5, 4
R4-R6, 2
R5-R1, 3
R5-R4, 5
R5-R6, 3
R6-R4, 1
R6-R5, 2
SPF Calculation Example
Link-state Database Tentative Path
R1-R2, 1 LS Entry Cost to Root R2-R2, 0
R1-R5, 2 R2-R4, 2 2 R2-R3, 1
R2-R1, 2 R3-R4, 2 3 R2-R1, 2
R2-R4, 2
R2-R3, 1 R1-R5, 2 4
R2-R4, 2
R3-R2, 3
R3-R4, 2
R4-R2, 4
R4-R3, 4
R4-R5, 4
R4-R6, 2
R5-R1, 3
R5-R4, 5
R5-R6, 3
R6-R4, 1
R6-R5, 2
SPF Calculation Example
Link-state Database Tentative Path
R1-R2, 1 LS Entry Cost to Root R2-R2, 0
R1-R5, 2 R1-R5, 2 4 R2-R3, 1
R2-R1, 2 R4-R5, 4 6 R2-R1, 2
R2-R4, 2
R2-R3, 1 R4-R6, 2 4
R1-R5, 4
R2-R4, 2
R3-R2, 3
R3-R4, 2
R4-R2, 4
R4-R3, 4
R4-R5, 4
R4-R6, 2
R5-R1, 3
R5-R4, 5
R5-R6, 3
R6-R4, 1
R6-R5, 2
SPF Calculation Example
Link-state Database Tentative Path
R1-R2, 1 LS Entry Cost to Root R2-R2, 0
R1-R5, 2 R4-R6, 2 4 R2-R3, 1
R2-R1, 2 R5-R6, 3 7 R2-R1, 2
R2-R4, 2
R2-R3, 1
R1-R5, 4
R2-R4, 2
R4-R6, 4
R3-R2, 3
R3-R4, 2
R4-R2, 4
R4-R3, 4
R4-R5, 4
„ Tentative is empty
R4-R6, 2
R5-R1, 3 „ All nodes in link-state database
R5-R4, 5 are in path
R5-R6, 3
R6-R4, 1 „ SPF calculation is finished
R6-R5, 2
SPF Calculation Example

„ Loop-free, lowest-cost path to every node

R2
2 1

R1 2
R3 Path
1 3
2 R2-R2, 0
2 R2-R3, 1
R4 R2-R1, 2
3 R5 4 R2-R4, 2
5 4
4 R1-R5, 4
3 2 R4-R6, 4

R6
2 1
Problems with Link State

„ Information flooding load

z Solution: sequence numbers and aging
z Solution: areas
„ Stale LS database entries
z Solution: periodic database refresh

„ .5(n22-n) adjacencies on multi-access

networks
z Solution: designated routers

„ Memory and CPU overload

z Solution: areas
Agenda

„ Overview of link-state protocols

„ A parallel history of IS-IS and OSPF
„ Comparative analysis of IS-IS and OSPF
„ Design considerations
„ Less-tangible considerations
„ Conclusions
In the Beginning Was DECnet...

„ Radia Perlman
„ Adopted by ISO for OSI model
„ IS-IS extended to support IP
z Interim solution until OSI makes it extinct
(don’t hold your breath)
z RFC 1195
z Also known as integrated IS-IS, dual IS-IS
IS-IS = 0

„ Parallel initiative by IETF to develop an IP

routing protocol
„ OSPF based on initial work on and
experience with IS-IS
Protocol History

„ (Disclaimer—biased, foggy memory)

„ 1987
z IS-IS (from DEC) selected by ANSI as OSI
intradomain protocol (CLNP only)

„ 1988
z NSFnet deployed, IGP based on early IS-IS draft
z OSPF work begins, loosely based on IS-IS
mechanisms (LS protocols are hard!)
z IP extensions to IS-IS defined
Protocol History

„ 1989
z OSPF v.1 RFC published
z Proteon ships OSPF
z IS-IS becomes ISO proposed standard
z Public bickering ensues; OSPF and IS-IS are
blessed as equals by IETF, with OSPF
somewhat more equal
z Private cooperation improves both protocols

„ 1990
z Dual-mode IS-IS RFC published
Protocol History

„ 1991
z OSPF v.2 RFC published
z Cisco ships OSPF
z Cisco ships OSI-only IS-IS

„ 1992
z Cisco ships dual IS-IS (part of DEC Brouter)
z Lots of OSPF deployed, but very little IS-IS

„ 1993
z Novell publishes NLSP (IPX IS-IS knockoff)
Protocol History

„ 1994
z Cisco ships NLSP (rewriting IS-IS as side effect)
z Large service providers need an IGP; IS-IS is
recommended due to recent rewrite and OSPF
field experience (and to lesser extent, NSF
CLNP mandate)

„ 1995
z Service providers begin deployment of IS-IS
ƒ Cisco
Cisco implementation
implementation firms
firms up
up
ƒ Protocol
Protocol starts
starts to
to become
become popular
popular in
in niche
niche
Protocol History

„ 1996-1998
z IS-IS niche popularity continues to grow
(some service providers switch to it from OSPF)
z IS-IS becomes barrier to entry for router vendors
targeting large service providers
z Juniper Networks and other vendors ship IS-IS
capable routers

„ 1999-2000
z Extensions continue for both protocols
(for example, traffic engineering)
Agenda

„ Overview of link-state protocols

„ A parallel history of IS-IS and OSPF
„ Comparative analysis of IS-IS and OSPF
„ Design considerations
„ Less-tangible considerations
„ Conclusions
ISOspeak 101

„ Intermediate System (IS)

„ End System (ES)
„ Protocol Data Unit (PDU)
„ Subnetwork Point of Attachment (SNPA)
„ Link State PDU (LSP)
„ Routing Domain
„ Level 2 Area
„ Level 1 Area
Message Types: OSPF

„ Hello packet
„ Database description packet
„ Link-state request packet
„ Link-state acknowledgement packet
„ Link-state update packet
Message Types: IS-IS

„ Hello PDU
„ Link-state PDU
„ Sequence number PDUs
z Complete Sequence Number PDU (CSNP)
z Partial Sequence Number PDU (PSNP)

„ Message types are further divided into Level

1 and Level 2
OSPF LSAs

„ Multiple LSA types

Type LSA
1 Router LSA
2 Network LSA
3 Network Summary LSA
4 ASBR Summary LSA
5 AS External LSA
6 Group Membership LSA
7 Not-so-stubby Area LSA
8 External Attributes LSA
9-11 Opaque LSAs
IS-IS Level 1 LSPs

„ Single LSP, multiple TLVs

Type TLV
1 Area Addresses
2 IS Neighbors
3 ES Neighbors
10 Authentication Information
128 IP Internal Reachability Information
129 Protocols Supported
132 IP Interface Address
IS-IS Level 2 LSPs

„ Single LSP, multiple TLVs

Type TLV
1 Area Addresses
2 IS Neighbors
4 Partition Designated Level 2 IS
5 Prefix Neighbors
10 Authentication Information
128 IP Internal Reachability Information
129 Protocols Supported
130 IP External Reachability Information
131 Inter-domain Routing Protocol Information
132 IP Interface Address
135 Extended IP Reachability (wide metrics)
Message Encoding: OSPF

„ Runs over IP (protocol number 89)

„ 32-bit alignment
„ Only LSAs are extensible
„ All OSPF speakers must recognize
the extensions
Message Encoding: IS-IS

„ Runs directly over data link

„ No alignment
„ All PDUs are extendable
„ Nested TLVs
Media Support

„ OSPF
z Broadcast (LANs)
z Point-to-point
z Point-to-multipoint
z NBMA

„ IS-IS
z Broadcast
z Point-to-point
z No NBMA support
Router and Area IDs: OSPF

„ Router ID and area ID specified separately

„ Each is 32-bit number
„ AID associated with interface
„ RID
z Explicitly specified RID
z Loopback address
z Highest interface IP address
Router and Area IDs: IS-IS

1-13 bytes 6 bytes 1 byte

Area ID System ID SEL

Examples:
01.0000.23a5.7c32.00
49. 0001.0000.23a5.7c32.00
47.0005.80.0000a7.0000.ffdd.0001.0000.23a5.7c32.00

„ Area ID and sysID (router ID) specified in Network Entity Title

(NET)
„ NSAP address format
„ In JUNOS™
™ Internet software, specified on loopback interface
Neighbor Discovery and
Maintenance: OSPF
„ Hello packets
z Establish two-way communication
z Advertise optional capabilities
z DR/BDR election/discovery
z Serve as keepalives
z 10s default hello interval, dead interval 4X
„ Most hello fields must match for adjacency
z Area ID, authentication, network mask, hello
interval, router dead interval, options
z Changing values causes adjacency disruption
Neighbor Discovery
and Maintenance: IS-IS
„ Hello packets
z
z Establish
Establish two-way
two-way communication
communication
z
z L1,
L1, L2,
L2, L1/L2
L1/L2 neighbor
neighbor discovery
discovery
z
z DR
DR election/discovery
election/discovery
z
z Serve
Serve asas keepalives
keepalives
z
z 3s
3s JUNOS
JUNOS default
default hello
hello interval,
interval, dead
dead interval
interval 3X
3X

„ Hellos padded to full MTU size (dubious)

„ Fewer matches necessary for adjacency
z
z Hello
Hello and
and dead
dead intervals
intervals can
can vary
vary
z
z Not
Not even
even IP
IP subnets
subnets must
must match!
match!
Database Synchronization: OSPF

„ Database synchronization driven by

state machine
„ Master/slave election
„ Database synchronization
z Database description packets
z Link-state request packets
z Link-state update packets
z Link-state acknowledgement packets
Database Synchronization: IS-IS

„ Simple synchronization based on flooding of

sequence number PDUs
„ CSNPs
z
z Describe
Describe all
all LSPs
LSPs in
in the
the database
database
z
z Analogous
Analogous to
to OSPF
OSPF DD DD messages
messages
z
z Sent
Sent by
by DR
DR every
every 10
10 seconds
seconds on
on broadcast
broadcast networks
networks
z
z Sent
Sent every
every hour
hour on
on point-to-point
point-to-point networks
networks

„ PSNPs
z
z Request
Request missing
missing or
or newer
newer LSPs
LSPs
z
z Analogous
Analogous to
to OSPF
OSPF LSLS Request
Request messages
messages
Database Refresh: OSPF

„ LSA refresh every 30 minutes

„ MaxAge = 1 hour
„ Up-counting timer
„ Design flaw: cannot change MaxAge
Database Refresh: IS-IS

„ LSP refresh every 15 minutes

z Minus random jitter timer of up to 25 percent

„ LSP Lifetime = 20 minutes (default)

„ Down-counting timer
„ LSP lifetime configurable up to 18.2 hours
„ Major reason IS-IS scales better to large areas
Designated Routers: OSPF

„ Highest priority becomes DR

z 0-255, default 128
z Highest router ID tie-breaker

„ Backup designated router

z Speeds recovery from failed DR

„ DR cannot be pre-empted
z The DR is usually the first active router

„ Adjacencies formed only with DR and BDR

Designated Routers (DIS): IS-IS

„ Highest priority becomes DR

z 0-127, default 64
z Highest MAC address tiebreaker
„ No backup designated router
„ DR can be pre-empted
z Adding a router to a LAN can cause temporary
instability
„ Adjacencies formed with all routers on LAN,
not just DR
z Separate L1 and L2 adjacencies on same LAN
 Area Structure: OSPF

„ Area boundaries fall on routers

„ Router types
z
z Interior (or backbone)
z
z ABR
Area 1
z
z ASBR Area 2
ASBR ABR/
ASBR ABR

Area 0

External
Routes ASBR
Area Structure: IS-IS

„ Area boundaries fall between routers

„ External reachability information in L2 LSPs only
„ Router types
z
z L1
L1 Area 1 Area 2 L1
z
z L2
L2 L1
z
z L1/L2
L1/L2 L1/L2 L1/L2

Area 3

External
Routes L2 L2
IS-IS Optimizations for IP Support

„ Three-way handshaking
„ Dynamic hostname exchange (RFC 2966)
„ >256 pseudonode support
„ Domain-wide prefix distribution (RFC 2966)
„ Wide metrics
Agenda

„ Overview of link-state protocols

„ A parallel history of IS-IS and OSPF
„ Comparative analysis of IS-IS and OSPF
„ Design considerations
„ Less-tangible considerations
„ Conclusions
Metrics: OSPF

„ Dimensionless metric
„ Large metric field
z
z Type
Type 11 LSA
LSA == 16
16 bits
bits
z
z Type
Type 3,
3, 4,
4, 5,
5, and
and 77 LSA
LSA == 24
24 bits
bits
„ Cost
z
z Cost
Cost == Reference
Reference BW BW // Interface
Interface BW
BW
z
z Default
Default Reference
Reference BWBW == 100
100 Mbps
Mbps
z
z IfIf (Ref
(Ref BW
BW // Interface
Interface BW)
BW) >> 1,
1, Cost
Cost == 11
z
z Cost
Cost cancan also
also bebe set
set arbitrarily
arbitrarily
„ External Metrics
z
z Type
Type 11 (E1)
(E1) == Assigned
Assigned cost
cost ++ cost
cost to
to ASBR
ASBR
z
z Type
Type 22 (E2)
(E2) == Assigned
Assigned cost
cost only
only
Metrics: IS-IS

„ Dimensionless metric
„ ISO 10589 defines 4 metric fields
z
z Only
Only default
default used
used in
in practice
practice
„ Small 6-bit metric field
z
z Default
Default == 10
10 for
for all
all interfaces
interfaces
z
z Maximum
Maximum interface
interface value
value == 64
64
z
z Maximum
Maximum routeroute metric
metric == 1,023
1,023
z
z Possible
Possible limited
limited metric
metric granularity
granularity in
in large
large networks
networks
z
z Originally
Originally intended
intended to to simplify
simplify SPF
SPF calculation
calculation (irrelevant
(irrelevant
with
with modern
modern CPUs)
CPUs)
„ Wide Metrics
z
z Extends
Extends metric
metric field
field to
to 32
32 bits
bits
„ Metrics tagged as internal or external (I/E bit)
LSA Scalability: OSPF

„ Famous “rules of thumb” carry little real meaning

„ 64 KB maximum LSA size
„ Only router (type 1) LSAs likely to grow large
z
z 24
24 bytes
bytes of
of fixed
fixed fields
fields
z
z 12
12 bytes
bytes to
to represent
represent each
each link
link
z
z 5,331
5,331 links,
links, maximum
maximum (but(but isn’t
isn’t this
this enough?)
enough?)

„ Types 3, 4, 5, 7 LSAs
z
z One
One destination
destination prefix
prefix per
per LSA
LSA
z
z Be
Be careful
careful what
what you
you redistribute!
redistribute!
LSP Scalability: IS-IS

„ Single LSP per router, per level

„ Fragmentation supported, but...
z Maximum fragment size = 1,470 bytes
z Maximum number of fragments = 256
z …Isn’t this enough?

„ Be careful what you redistribute!

Stub Areas

„ Trade routing precision for improved scalability

„ OSPF
z Stub areas eliminate type 5 LSA load
z Totally stubby areas extend the concept
z All area routers must understand stubbiness

„ IS-IS
z L1 routers are “totally stubby” by default
z Attached (ATT) set by L1/L2 router
IS-IS Inter-Area Route Leaking

„ Why leak routes?

z
z Improved
Improved routing
routing precision
precision
z
z More
More accurate
accurate BGP
BGP next-hop
next-hop resolution
resolution
z
z Using
Using IS-IS
IS-IS metric
metric as
as BGP
BGP MED
MED

„ L1-->L2 route leaking happens by default

z
z Internal
Internal routes
routes only
only
z
z External
External routes
routes require
require policy
policy

„ L2-->L1 route leaking requires policy

z
z Internal
Internal or
or external
external
z
z Up/down
Up/down bit
bit prevents
prevents looping
looping
Not-So-Stubby Areas

„ OSPF feature
z “Trick” to allow advertisement of external routes
through stub areas (type 5 LSAs illegal)
z All routers in area must understand type 7 LSAs

„ Similar function with IS-IS

z Using simple L1-->L2 policy
NBMA Networks

„ OSPF
z Point-to-point
z Point-to-multipoint mode
z NBMA mode (but why?)
z P-T-MP and NBMA require manual specification
of neighbor addresses

„ IS-IS
z No multipoint support
z Must configure interfaces as logical P-T-Ps
Virtual Links

„ Useful for
z Patching partitioned backbone areas
z Area migrations

„ Should be a temporary solution!

„ Full OSPF support
„ No IS-IS support
z Specified in ISO 10589, but not implemented by
major router vendors
Overload Bit

„ IS-IS feature
z Enables router to signal memory overload
z No transit traffic sent to overloaded router
z Set separately for Level 1 and Level 2
z Can be manually set, useful for graceful router
turn-up

„ No comparable OSPF feature

Mesh Groups

„ IS-IS feature (RFC 2973)

z Can sharply curtail LSP flooding in full-mesh
topologies
z Each router in mesh group receives only one
copy of each LSP (one-hop flooding)
z Risk of lost LSPs—ensure design is robust
enough!
z Interfaces can be manually configured to block
LSPs (increased scalability, but increased risk)

„ OSPF has no comparable feature

Security

„ Both protocols support authentication

z Plain-text passwords (sniffable!)
z MD5 cryptographic hash
„ Authentication especially important with
OSPF
z Runs over IP, so subject to spoofing and
other attacks
„ Non-IP nature makes IS-IS inherently
more secure
z But authentication still a good idea
Traffic Engineering Support

„ Both protocols extended to disseminate

traffic engineering parameters
„ OSPF
z
z Type 10 Opaque (area scope) LSAs

„ IS-IS
z
z Extended IS reachability (type 22) TLV
ƒ Traffic
Traffic engineering
engineering parameters
parameters in
in sub-TLVs
sub-TLVs

z
z Extended IP reachability (type 135) TLV
ƒ Wide
Wide metrics
metrics and
and up/down
up/down bit
bit

z
z Limited to area by L1 or L2 PDU type
Agenda

„ Overview of link-state protocols

„ A parallel history of IS-IS and OSPF
„ Comparative analysis of IS-IS and OSPF
„ Design considerations
„ Less-tangible considerations
„ Conclusions
Extensibility

„ OSPF
z New extensions require new LSAs (usually)
z All routers must understand new LSAs (usually)
z IPv6 support will require new OSPF version
„ IS-IS
z New extensions require new TLVs (usually)
z Fewer compatibility issues than OSPF
z Small community of interest (big ISPs) with big
vendor clout means faster rollout of extensions
z Extendable for IPv6
Optimality

„ Optimality
z OSPF was optimized for things that don’t matter
any more (link bandwidth, CPU alignment)
z IS-IS was optimized for things that don’t matter
any more (large LANs, SPF cost)
z Optimizations turn out to add complexity, but not
much value
z A lot has changed in 10 years
Guru Availability

„ OSPF
z Broad experience base
z Many books, RFCs, training classes available

„ IS-IS
z Significantly smaller experience base
z Scarcity of documentation makes it mysterious
z Simpler than OSPF, easy to learn (with a few
shifts in thinking)
 Writing OSPF Code

„ OSPF spec is an excellent implementation guide

z If followed to the letter, a working, if naïve,
implementation will likely result
z Spec is complex, but has almost no “why”
information; hence, other (potentially more scalable)
implementation approaches are at the implementer’s
own risk
z Barrier to entry in high-end router market
(you need to know the protocol intuitively)
Writing IS-IS Code

„ IS-IS spec uses arcane ISOspeak and has

very few implementation hints
z Spec is inherently simple (once you get the
lingo), with fewer implementation issues
z Boilerplate at front and back of spec means you
can lose pages without affecting content
z Barrier to entry in high-end router market (you
need to know the protocol intuitively)
Agenda

„ Overview of link-state protocols

„ A parallel history of IS-IS and OSPF
„ Comparative analysis of IS-IS and OSPF
„ Design considerations
„ Less-tangible considerations
„ Conclusions
Conclusion

„ Both protocols are mature and stable

(with the right vendor)
„ Both protocols continue to be extended
„ Enterprise networks
z
z IGP requirements can be complex
z
z OSPF is a nobrainer

„ Service provider networks

z
z IGP requirements usually simpler
z
z Scalability and stability are paramount
z
z Consider your requirements carefully; pick the
protocol that fits
 Thank You
http://www.juniper.net