Routing &
Protocols
1
Paul Traina
cisco Engineering
2
Today's Talk
• erminology
T
•Routing
•Static Routes
•Interior Gateway Protocols
•Exterior Gateway Protocols
•Building an ISP network
Paul Traina / INET '95 3
Developing Countries Workshop
Terminology
• etwork number
n
•prefix
•mask (or length)
Paul Traina / INET '95 4
Developing Countries Workshop
Static routes
hand configured routing
t•ell the router which way to send
packets
•based upon final packet destination
Paul Traina / INET '95 5
Developing Countries Workshop
Static routes
S3
171.65.3.4
• p route 10.0.0.0
i
255.0.0.0 serial 3
•ip route 131.108.0.0
255.255.0.0 171.65.3.4
Paul Traina / INET '95 6
Developing Countries Workshop
Terminology
Interior Gateway Protocol (IGP)
• IP, IGRP, HELLO, OSPF
R
•Primary goal is optimal connectivity
•Strong distance metrics
•May not have good administrative
controls
Paul Traina / INET '95 7
Developing Countries Workshop
Terminology
Distance vector protocols
l•isten to neighboring routers
•install routes in table, lowest distance
wins
•advertise all routes in table
•very simple
•very stupid
Paul Traina / INET '95 8
Developing Countries Workshop
Terminology
Distance vector protocols
D
A
G
E
B
H
C
A
B
C
G
H
1
1
1
sup
1
I
A 2
B 2
C 2
D 1
E sup
F 1
G 1
H 1
I1
F
Paul Traina / INET '95 9
Developing Countries Workshop
Terminology
Link state protocols
•information about adjacencies sent to
all routers
•each router builds a topology database
•a "shortest path" algorithm is used to
find best route
•converge as quickly as databases can
be updated
Paul Traina / INET '95 10
Developing Countries Workshop
Terminology
Link state protocols
A
D
2
1
G
E
B
3
H
I
C
router 1
A, B, C, G, H
router 3
H, I
F
router 2
D, E, F, G, I
A - 1 - G - 2 - D
Paul Traina / INET '95 11
Developing Countries Workshop
Interior Gateway Protocols
Routing Information Protocol (RIP)
I•P only
•distance vector protocol
•slow convergence
•does not carry mask information
•reasonably simple design &
configuration
•does not scale (maximum 15 hops)
•poor metrics (hop-count)
Paul Traina / INET '95 12
Developing Countries Workshop
Interior Gateway Protocols
Interior Gateway Routing Protocol
(IGRP)
I•P only
•distance vector protocol
•slow convergence (like RIP)
•does not carry mask information
(like RIP)
•very simple design & configuration
powerful proprietary metric
–
load sharing across diverse links
–
Paul Traina / INET '95 13
Developing Countries Workshop
Interior Gateway Protocols
The IGRP metric
»always get optimal routing
•metric vector, not single value
bandwidth
–
delay
–
hops
–
reliability
–
loading
–
Paul Traina / INET '95 14
Developing Countries Workshop
Interior Gateway Protocols
Enhanced IGRP
• ulti-protocol (IP, IPX, Appletalk)
m
•fast convergence (like OSPF)
•very simple design & configuration
(like IGRP)
IGRP metric
–
allows load sharing across diverse
–
links
Paul Traina / INET '95 15
Developing Countries Workshop
Interior Gateway Protocols
Enhanced IGRP
• istance vector based protocol
d
•NOT a Bellman-Ford protocol
Uses "dual" algorithm
•alternative to OSPF & I-ISIS
•can be bandwidth intensive on slow
links
Paul Traina / INET '95 16
Developing Countries Workshop
Interior Gateway Protocols
Integrated IS-IS (I-ISIS)
• ulti-protocol (CLNP, IP, IPX, ...)
m
•link state protocol
•fast convergence
•design and architecture moderately
complex
•configuration may be simple
Paul Traina / INET '95 17
Developing Countries Workshop
Interior Gateway Protocols
Open Shortest Path First (OSPF)
•IS - IS = 0
Paul Traina / INET '95 18
Developing Countries Workshop
Interior Gateway Protocols
Open Shortest Path First (OSPF)
I•P only
•link state protocol
•fast convergence
•design and architecture very complex
•configuration can be simple
Paul Traina / INET '95 19
Developing Countries Workshop
Interior Gateway Protocols
Which to use?
• our interior network is actually VERY
Y
simple.
•Your IGP should only carry your routes
and your direct customers'
Paul Traina / INET '95 20
Developing Countries Workshop
Interior Gateway Protcols
Problems with "classic" protocols
• low convergence
s
•count to infinity
•no mask information
no CIDR
–
no VLSM
–
no subnet 0
–
Paul Traina / INET '95 21
Developing Countries Workshop
Interior Gateway Protocols
Slow convergence
•advertisement period
entire routing table dumped every n
–
seconds
•timeout period
usually 3 times advertisement period
–
•RIP values are normally 30 and 90
seconds!
Paul Traina / INET '95 22
Developing Countries Workshop
Interior Gateway Protocols
Count to infinty problem
1
4
2
A
3
B
1 (1 hop)
C
1 (2 hops)
1 (3 hops)
1
4
2
A
3
B
C
route cleared
1 (3 hops)
1
4
2
A
3
B
1 (4 hops)
C
1 (3 hops)
Paul Traina / INET '95 23
Developing Countries Workshop
Interior Gateway Protocols
Count to infinity: split-horizon
• on't feed selected route back to
D
source
no feedback on source interface
–
no feedback to source neighbor
–
Paul Traina / INET '95 24
Developing Countries Workshop
Interior Gateway Protocols
Count to infinity: split-horizon
1
4
2
A
1
3
B
1 (1 hop)
C
1 (2 hops)
1 (3 hops)
4
2
A
3
B
1
C
route cleared
2
A
1 (3 hops)4
3
B
route cleared
C
route cleared
Paul Traina / INET '95 25
Developing Countries Workshop
Interior Gateway Protocols
Count to infinity: hold-down
• plit horizon not sufficient!
S
•Holddown period
interval during which "less
–
attractive" updates are ignored
Paul Traina / INET '95 26
Developing Countries Workshop
Interior Gateway Protocols
Count to infinity: hold-down
Paul Traina / INET '95 27
Developing Countries Workshop
Interior Gateway Protocols
The universal rule
• ou will always trade bandwidth for
Y
speed of convergence
Paul Traina / INET '95 28
Developing Countries Workshop
Interior Gateway Protocols
OSPF configuration
•myth
OSPF is hard to use
–
•reality:
– outer ospf 1
r
network 192.111.107.0 0.0.0.255 area 0
Paul Traina / INET '95 29
Developing Countries Workshop
Interior Gateway Protocols
OSPF operation
•every OSPF router sends out 'hello'
packets
•hello packets used to determine if
neighbor is up
•hello packets are small easy to process
packets
•hello packets are sent periodically
(usually short interval)
Paul Traina / INET '95 30
Developing Countries Workshop
Interior Gateway Protocols
OSPF operation
• nce an adjacency is established, trade
o
information with your neighbor
•topology information is packaged in a
"link state announcement"
•announcements are sent ONCE, and
only updated if there's a change
(or every 45mins...)
–
Paul Traina / INET '95 31
Developing Countries Workshop
Interior Gateway Protocols
OSPF operation
• hange occurs
c
•broadcast change
•run SPF algorithm
•install output into forwarding table
Paul Traina / INET '95 32
Developing Countries Workshop
Interior Gateway Protocols
making OSPF scale
• ach link transition causes a broadcast
e
and SPF run
•OSPF can group routers to appear as
one single router
•OSPF areas
Paul Traina / INET '95 33
Developing Countries Workshop
Interior Gateway Protocols
OSPF areas (before)
Paul Traina / INET '95 34
Developing Countries Workshop
Interior Gateway Protocols
OSPF areas (after)
Paul Traina / INET '95 35
Developing Countries Workshop
Interior Gateway Protocols
OSPF areas - partitioning
Paul Traina / INET '95 36
Developing Countries Workshop
Interior Gateway Protocols
OSPF areas - partition repair
Paul Traina / INET '95 37
Developing Countries Workshop
Interior Gateway Protocols
OSPF areas
•rule of thumb:
no more than 150 routers/area
•reality:
no more than 500 routers/area
•backbone "area" is an area
•proper use of areas reduce bandwidth
& CPU utilization
Paul Traina / INET '95 38
Developing Countries Workshop
Interior Gateway Protocols
EIGRP operation
•design goals were
make it as fast as OSPF & IS-IS
–
make it trivial to configure
–
easy migration from IGRP
–
Paul Traina / INET '95 39
Developing Countries Workshop
Interior Gateway Protocols
EIGRP operation
• outer eigrp 1
r
network 192.108.0.0 mask 255.255.0.0
Paul Traina / INET '95 40
Developing Countries Workshop
Interior Gateway Protocols
EIGRP operation - caveats
• othing is for free
n
•EIGRP works best on high speed links
•EIGRP doesn't scale well in
high-meshed frame-relay networks
star networks OK
–
Paul Traina / INET '95 41
Developing Countries Workshop
Interior Gateway Protocols
summarization
150.196.4.0
.1.0
150.196
.3.0
.2.0
131.108.3.0
131.108
• lassful routing protocols naturally
c
summarize to network numbers at
boundaries
Paul Traina / INET '95 42
Developing Countries Workshop
Interior Gateway Protocols
summarization
131.108.0.0/16
131.108.3.64/27
131.108.4.32/29
131.108.4.0/24
• lassless routing protocols summarize
c
at arbitrary bit boundaries
Paul Traina / INET '95 43
Developing Countries Workshop
Interior Gateway Protocols
route filtering
• seudo-security (bad idea!)
p
•low bandwidth links
•eliminate unnecessary information
Paul Traina / INET '95 44
Developing Countries Workshop
Interior Gateway Protocols
route filtering
B
C
A
131.108/16
150.136/16
44/8
0/0
Paul Traina / INET '95 45
Developing Countries Workshop
Interior Gateway Protocols
redistribution
OSPF
RIP
• ou run OSPF
y
•your neighbor runs RIP
Paul Traina / INET '95 46
Developing Countries Workshop
Interior Gateway Protocols
redistribution
•run RIP on their interface
•router rip
network 192.111.107.0
•configure OSPF to redistribute RIP
•router ospf 1
network 135.111.104.0
0.0.0.255 area 0
redistribute rip metric 10
Paul Traina / INET '95 47
Developing Countries Workshop
Interior Gateway Protocols
redistribution
• i-directional redistribution MUST be
b
filtered!
Paul Traina / INET '95 48
Developing Countries Workshop
Interior Gateway Protocols
redistribution
• outer rip
r
network 192.111.107.0
•router ospf 1
network 135.111.104.0 0.0.0.255
area 0
redistribute rip metric 10
distribute-list 1 out rip
•access-list 1 permit 192.111.107.0
0.0.0.255
Paul Traina / INET '95 49
Developing Countries Workshop
Exterior routing
xx
Job Number Goes Here 50
Exterior routing
• erminology
T
•What is exterior routing?
•Routing protocols
•Overview of BGP
•Putting it all together
•Further information
Paul Traina / INET '95 51
Developing Countries Workshop
Terminology
Autonomous System
•A set of networks sharing the same
routing policy.
•Internal connectivity
•One contiguious unit
•Identified by "AS number"
•Examples
service provider
–
multi-homed customer
–
anyone needing policy discrimination
–
Paul Traina / INET '95 52
Developing Countries Workshop
Terminology
Exterior routes
• outes learned from other autonomous
R
systems
Paul Traina / INET '95 53
Developing Countries Workshop
Terminology
Exterior Gateway Protocol
•egp vs EGP
•EGP, BGP, IDRP
•Primary goal is to provide reachability
information outside administrative
domain
•Secondary goal is administrative control
•Metrics may be arbitrary or weak
Paul Traina / INET '95 54
Developing Countries Workshop
Terminology
Natural network mask
•Classful mask
Class A = 8 bits
–
networks 1...127
–
Class B = 16 bits
–
networks 128.0...191.255
–
Class C = 24 bits
–
networks 192.0.0...223.255.255
–
Paul Traina / INET '95 55
Developing Countries Workshop
Terminology
DMZ network
• e-militarised zone
d
•area between North and South Korea
•shared network between ASs
before, neither AS carried it in IGP
–
now, both carry it in IGP
–
Paul Traina / INET '95 56
Developing Countries Workshop
Terminology
DMZ network
DMZ networks
Paul Traina / INET '95 57
Developing Countries Workshop
Why do we need exterior routing?
Why not make entire internet a single
cloud?
• eparate policy control
s
•filtering on networks doesn't scale well
•service provider selection given
multiple choices
•everything must scale to hundreds of
thousands of routes
Paul Traina / INET '95 58
Developing Countries Workshop
Exterior Routing
• tatic routes
s
•multiple IGP instances
•OSPF inter-domain routing
•EGP
•IDRP
•BGP version 4
Paul Traina / INET '95 59
Developing Countries Workshop
Exterior Routing
Static routes
• o path information
n
•very versatile
•low protocol overhead
•high maintenance overhead
•very very very bad convergence time
requires manual configuration
–
Paul Traina / INET '95 60
Developing Countries Workshop
Exterior Routing
Multiple IGPs with route leaking
•Run an instance of an IGP at each site
for local routing
•Run a backbone IGP at each border
router
•redistribute local IGP into backbone IGP
•redistribute backbone IGP into local IGP
(or default)
•backbone routers share common
administration
Paul Traina / INET '95 61
Developing Countries Workshop
Exterior Routing
Multiple IGPs with route leaking
rip default
redistributed into
customer's IGP
RIP routes learned from
customer redistributed
into service provider's IGP
after filtering
RIP run over wire
IGRP 109
customer
OSPF 690
service provider
Paul Traina / INET '95 62
Developing Countries Workshop
Exterior Routing
Multiple IGPs with route leaking
•backbone IGP
router ospf 690
–
network 129.119.0.0 0.0.255.255
area 0
redistribute rip metric 5
distribute-list 1 rip out
•local IGP
– outer igrp 109
r
network 131.108.0.0
ip default-network 140.222.0.0
Paul Traina / INET '95 63
Developing Countries Workshop
Exterior Routing
OSPF inter-domain routing
• oute leaking formalised for one
R
protocol
•OSPF tag carries originating AS
limited policy control
–
only have 32 bit OSPF tag
–
OSPF tag contains originating AS
–
Paul Traina / INET '95 64
Developing Countries Workshop
Exterior Routing
Exterior Gateway Protocol
• istorical protocol
h
•obsolete
•assumes a central core
•no transit service except via core
Paul Traina / INET '95 65
Developing Countries Workshop
Exterior Routing
Exterior Gateway Protocol (historical)
•RIP by any other name
•fancy "hello dance"
•periodic update protocol
•entire routing table sent with each
update
•no metric
everything is one hop from core
–
Paul Traina / INET '95 66
Developing Countries Workshop
Exterior Routing
Exterior Gateway Protocol
109
110
core
• S 110 may not advertise AS 109 to
A
core
Paul Traina / INET '95 67
Developing Countries Workshop
IDRP (future expansion path)
Inter-domain routing protocol
•IDRP is an almost identical clone of
BGP-4
•IDRP is multi-protocol
IP
–
CLNP
–
IPX
–
•For purposes of this talk:
g/BGP-4/s//IDRP/g
Paul Traina / INET '95 68
Developing Countries Workshop
BGP-4
Border Gateway Protocol version 4
•carries external routes only
•uses reliable transport mechanism
(TCP)
•not a periodic routing protocol
•allows limited policy selection
•AS path insures loop free routing
•"best path" determined at AS granularity
Paul Traina / INET '95 69
Developing Countries Workshop
BGP peer relationships
External BGP
AS 110
AS 109
• eighbor is in a different AS
n
•neighbors share a common network
Paul Traina / INET '95 70
Developing Countries Workshop
BGP peer relationships
Internal BGP
• eighbor in same AS
n
•may be several hops away
•full neighbor mesh required
Paul Traina / INET '95 71
Developing Countries Workshop
Common BGP networks
Stub customer
• GP only at border
B
•default to border
Paul Traina / INET '95 72
Developing Countries Workshop
Common BGP networks
Multi-homed customer
I•nternal BGP used with IGP
•IBGP only between border gateways
•Only border gateways speak BGP
•Synchronization with IGP required
•May use one IGP for exterior routes,
and another for internal nodes
exterior routes must be redistributed
–
into IGP
Paul Traina / INET '95 73
Developing Countries Workshop
Common BGP networks
Multi-homed customer
Paul Traina / INET '95 74
Developing Countries Workshop
Common BGP networks
Service provider
I•nternal BGP used to carry exterior
routes
IGP carries local information only
–
Full mesh required if no IGP
–
synchronization
Paul Traina / INET '95 75
Developing Countries Workshop
Common BGP networks
Service provider
Paul Traina / INET '95 76
Developing Countries Workshop
Common BGP networks
Service provider confederation
• group of service providers
A
•Multiple connectivity points
multi-exit discriminator useful
–
•Not a special case
Paul Traina / INET '95 77
Developing Countries Workshop
The BGP protocol
Update messages
• ithdrawn routes
w
•attributes
•advertised routes
Paul Traina / INET '95 78
Developing Countries Workshop
Update messages
Network reachability information
•prefix length
number of significant bits
–
•network prefix
0 to 4 bytes
–
•Example:
131.108/16
–
131.108.0.0 255.255.0.0
–
193/8
–
193.0.0.0 255.0.0.0
–
Paul Traina / INET '95 79
Developing Countries Workshop
Update messages
Attributes
•AS path
•next hop
•origin
•local preference
•multi-exit discriminator
•atomic aggregate
•aggregator
Paul Traina / INET '95 80
Developing Countries Workshop
AS path
AS sequence
•a list of AS's that a route has traversed
109 200 690 1755 1883
–
Paul Traina / INET '95 81
Developing Countries Workshop
AS path
AS sequence
193.0.34/24
1881
193.0.33/24
1880
193.0.33/24 1880 1881
193.0.34/24 1880
193.0.35/24 1880 1882
193.0.35/24
1882
Paul Traina / INET '95 82
Developing Countries Workshop
AS path
AS set
• ath traversed one or more members of
p
a set
{1880,1881,1882}
–
Paul Traina / INET '95 83
Developing Countries Workshop
AS path
AS set
1881
193.0.33/24
193.0.31/24
1880
193.0.32/22
{1880,1881,1882}
193.0.32/24
1882
Paul Traina / INET '95 84
Developing Countries Workshop
AS path
Sets and sequences combined
•local aggregation
109 200 690 1755
–
{1881,1882,1883}
•regional aggregation
109 200 690
–
{1755,1881,1882,1883,...}
Paul Traina / INET '95 85
Developing Countries Workshop
BGP path selection
BGP maintains multiple "feasable"
paths to a destination
f•ast convergence
•routing based upon preferences
•Example:
131.108/16 may be reached via
–
AS path 690 200 109 or via AS
path 690 1340 109
Paul Traina / INET '95 86
Developing Countries Workshop
BGP path selection algorithm
Initial route determination
• o not consider path if no next hop route
d
•largest weight
local to router
–
•highest local preference
global within AS
–
•shortest AS path
Paul Traina / INET '95 87
Developing Countries Workshop
BGP path selection
Tie breaking
•multi-exit discriminator
only considered if AS paths identical
–
•external routes
•best IGP metric to next hop
•highest IP address
Paul Traina / INET '95 88
Developing Countries Workshop
Policy Control
•distribute list
filter individual networks
–
•filter list
filter by AS path
–
•route maps
general policy control and tuning
–
Paul Traina / INET '95 89
Developing Countries Workshop
More information
Technical information on BGP
•RFC-1772
application of the Border Gateway
–
Protocol
•RFC-1771
BGP-4 protocol reference
–
document
•RFC-1745
BGP <-> OSPF interaction
–
Paul Traina / INET '95 90
Developing Countries Workshop
Building an
Internet
xx
Job Number Goes Here
91
Putting it all together
General philosophy
•Your network is going to grow at an
exponential rate!
•Design to scale...but be prepared to
reorganize from scratch
•Don't be afraid of change!
Most network redesigns are only
–
configuration changes
Paul Traina / INET '95 92
Developing Countries Workshop
Putting it all together
• equirements for IGPs for backbones
R
•IGP connects your backbone together,
not your client's routes
•Must
converge quickly
–
•Should
carry netmask information
–
Paul Traina / INET '95 81
Developing Countries Workshop
Putting it all together
connecting to a customer
•static routes
you control directly
–
no route flaps
–
no packets to be charged
–
•shared routing protocol or leaking...
you MUST filter your customers info
–
route flaps
–
•BGP for multi homed customers
Paul Traina / INET '95 82
Developing Countries Workshop
Putting it all together
building your backbone
• eep it simple
k
•redundancy is good, but expensive
•use an IGP that carrys mask
information
•use an IGP that converges quickly
•use OSPF, ISIS, or EIGRP
Paul Traina / INET '95 83
Developing Countries Workshop
Putting it all together
connecting to other ISPs
• se BGP-4
U
•advertise only what you serve
•take back as little as you can
Paul Traina / INET '95 84
Developing Countries Workshop
Putting it all together
the internet exchange
l•ong distance connectivity is expensive
•connect to several providers at a single
point
Paul Traina / INET '95 85
Developing Countries Workshop
Internet exchanges - FIX
Federal internet exchange (historical)
• umb ethernet connecting a group of
d
service providers
Paul Traina / INET '95 86
Developing Countries Workshop
Internet exchanges - FIX
Federal internet exchange
• ingle primary media all systems share
s
•secondary media may be shared by a
subset of systems to reduce load on
primary media
Paul Traina / INET '95 87
Developing Countries Workshop
Non-Internet exchange - CIX
Commercial internet exchange
(historical)
CIX
• ctually a one-router transit AS
a
•CIX clients only receive best path as
determined by CIX router
Paul Traina / INET '95 88
Developing Countries Workshop
Internet exchanges - d-GIX
Distributed global internet exchange
•emulates a single ethernet
Paul Traina / INET '95 89
Developing Countries Workshop
Internet exchanges - d-GIX
Distributed global internet exchange
• hare the cost of high speed lines
s
•single virtual level-2 media
bridges, not routers, connect the link
–
access points
bridge table entries are static
–
don't need spanning tree
–
mac address filtering used
–
Paul Traina / INET '95 90
Developing Countries Workshop
Internet exchanges - d-GIX
Distributed global internet exchange
t•he GIX itself still has no routing policy
•in that case, how do you pay for it?
•the GIX does have connectivity policy
charge for MAC address filters
–
(source/destination filtering)
Paul Traina / INET '95 91
Developing Countries Workshop
Internet exchanges - multi-NAP
Multiple-media network access point
Frame Relay
ATM
Network
Access
Point
local ethernet
Paul Traina / INET '95 92
Developing Countries Workshop
Internet exchanges - multi-NAP
Multiple-media network access point
•Problem:
How do you allow one NAP client to
–
connect via Frame Relay and another
customer connect via ATM?
•Answer:
Don't do this! Extend the NAP and
–
keep it policy free.
Paul Traina / INET '95 93
Developing Countries Workshop
Interenet exchanges - multi-NAP
Multiple-media network access point
•NAPs and IXs need to be policy free
•Routers implicity have an 'advertise
only what you use' policy.
•If routers are used, NAP becomes a
transit AS, not an "IX," and clients of the
NAP are limited by the NAP's route
selection policy.
Paul Traina / INET '95 94
Developing Countries Workshop
More information
•Original GIX proposal
ftp://ftp.ripe.net/ripe/docs/ripe-082.ps
–
– tp://ftp.ripe.net/ripe/drafts/
f
gix15jun.txt
• -GIX - distributed global internet
d
exchange
ftp://ftp.ripe.net/ripe/drafts/
–
d-gix-proposal.ps
Paul Traina / INET '95 95
Developing Countries Workshop
Routing registries
What are they?
•database containing
route prefix/
–
origin autonomous system
autonomous system/
–
connectivity policy
•RIPE-181 aka RC-1786
Paul Traina / INET '95 96
Developing Countries Workshop
Classless routing
xx
Job Number Goes Here 97
Why CIDR?
I•P route advertisements have been
growing exponentially.
•Class A networks are too big
•Class C networks are too small
•Only 65534 class B networks available
Paul Traina / INET '95 98
Developing Countries Workshop
Routing Table Growth
60000
50000
40000
30000
20000
10000
0
87 88 89 90 91 92 93 94 95 96
Paul Traina / INET '95 99
Developing Countries Workshop
Why CIDR?
Classful networks mis-sized
•Class A networks are too big
not desirable because of connectivity
–
constraints
•Class B address space is depleted
•Class C networks are useful only for
small customers
large gap between "C" customer and
–
"B" customer
Paul Traina / INET '95 100
Developing Countries Workshop
Classless routing
CIDR at the service provider level
•Service provider given CIDR blocks by
numbering authority
•Example:
198.24/15 == 512 class "C" nets
–
•Service provider advertises only a
summary route for CIDR block to
neighboring providers, not 512
separate class "C" routes.
Paul Traina / INET '95 101
Developing Countries Workshop
Classless routing
The client interface
• artition local CIDR block and assign to
P
customers
•Example:
198.24.62/23 == 2 "C" nets
–
198.24.192/18 == 64 "C" nets
–
198.24.61/24 == 1 "C" net
–
Paul Traina / INET '95 102
Developing Countries Workshop
Classless routing
Do's and don'ts
•Don't assign blocks smaller than class
"C" sized networks without prior
agreement from customers
most hosts & routing protocols are
–
not classless
•Do help customers use their address
space wisely!
Paul Traina / INET '95 103
Developing Countries Workshop
Classless routing
Do's and don'ts
• o give customers enough address
D
space for what they need
•Do parition your CIDR block to provide
for customer growth
get the tree program
–
understand RFCs 1519 and 1219
–
Paul Traina / INET '95 104
Developing Countries Workshop
Classless routing
Do's and don'ts
• on't be afraid of "holes" when
D
aggregating
•Longest match routing means "he who
has the longest prefix wins"
Paul Traina / INET '95 105
Developing Countries Workshop
Classless routing
Getting the most out of your allocation
I•t's natural, but inefficent to subnet on
8 bit boundaries
131.108.1 = subnet 1
–
131.108.2 = subnet 2
–
131.108.3 = subnet 3
–
•254 subnets with up to 254 hosts per
subnet out of a 16 bit address
allocation
Paul Traina / INET '95 106
Developing Countries Workshop
Classless routing
There are NO NETWORK NUMBERS!!!
•...just address space prefixes
131/8
–
131.0/12
–
131.108/16
–
131.108.5/24
–
131.108.5.32/29
–
131.108.5.33/32
–
Paul Traina / INET '95 107
Developing Countries Workshop
Classless routing
There are NO SUBNET MASKS!!!
•It's no longer a mask, just a prefix
length
•There can be no '0' holes in the mask
•/16 = 255.255.0.0
•/32 = 255.255.255.255
•/14 = 255.252.0.0
•/0 = default = 0.0.0.0
Paul Traina / INET '95 108
Developing Countries Workshop
Classless routing
Getting the most out of your allocation
•Unnumbered serial links
•Variable length subnet masks
•Small ethernet
28 bit mask = 14 hosts
–
•Larger ethernet
26 bit mask = 62 hosts
–
•VLSM allocation rules are the same as
CIDR allocation
Paul Traina / INET '95 109
Developing Countries Workshop
Classless routing
restrictions removed
•no such thing as a "subnet" anymore
subnet 0 is no longer special
–
all 1's subnet is no longer special
–
no such thing as a disconnected
–
subnet
Paul Traina / INET '95 110
Developing Countries Workshop
Classless routing
Mickey Mouse topology is OK
131.108.128/17
131.108.0/17
192.111.107/24
Paul Traina / INET '95 111
Developing Countries Workshop
Classless routing
Plan for entropy
• hat is your policy when customers
W
move to a different service provider?
do you own the numbers in the CIDR
–
block?
will new service provider supply
–
more specific routing information?
Paul Traina / INET '95 112
Developing Countries Workshop
Classless routing
Allocate addresses efficiently!
• ou don't get very many
y
•what happens as organizations grow?
•what happens when your customers lie
to you?
Paul Traina / INET '95 113
Developing Countries Workshop
More information
Technical information on classless
routing
•RFCs 1517, 1518, and 1519
address assignment and aggregation
–
strategy
•RFC1219
assignment of subnet numbers
–
•ftp://ftp.sesqui.net/pub/tools/tree.tar
– rogram to help calculate address
p
assignment
Paul Traina / INET '95 114
Developing Countries Workshop
More information
Technical information on address
allocation
•RIPE NCC address allocation guidelines
Paul Traina / INET '95 115
Developing Countries Workshop