Download presentation
Presentation is loading. Please wait.
1
IP Design and Migration
1054_03F8_c2
2
sraza@cisco.com Cisco Systems GSE
Khalid Raza Cisco Systems GSE
3
Network Design Design a hierarchical network
Break your network into regions Exchange regional routes via BGP
4
Network Design Advantages of regionalizing your network
You can have multiple instances of routing protocols Your IGP does not need to carry unnecessary routes from other regions Regions can grow independently Merging organizations is less painful
5
Network Design Case study Organization X.com has merged with Y.com
Both have registered networks X.com is a large organization and owns a class B network Y.com has 16 of class C networks, they own a CIDR block
6
Network Design Case Study Both organizations have different IGPs
Both of their individual Internet connections This new merged organization wants to use their existing Internet connections They want to avoid Asymmetric routing
7
Internet Connection to ISP A
Network Design Internet Connection to ISP C Internet Connection to ISP A SJ1 LA1 X.com Y.com /20 NY1 DC1 Internet Connection to ISP B
8
Network Design Case Study More details
CIDR block for network Y.com is out of ISP A’s address space Y.com want to keep ISP A as their primary service provider X.com wants to have ISP C as their primary service provider
9
Network Design First set Run BGP between the two organizations
Make sure both the organizations advertise all the networks via BGP to all the ISPs for redundancy
10
Network Design Redistribute the BGP routes form Y.com only into the IGP of X.com if optimal routing is desired Redistribute the BGP routes form X.com only, into the IGP of Y.com if optimal routing is desired If optimal routing is desired each of the router connecting the two organizations should send MED
11
Network Design Configurations LA1# router bgp 321
network mask network mask aggregate-address summary-only neighbor remote-as 132 neighbor remote-as 331 neighbor remote-as 132 unsupress-map foo route-map foo permit 10 match ip address 1 set metric 2 set community no-export route-map foo 20 match ip address 2 access-list 1 permit access-list 2 permit
12
Network Design X.com is running OSPF is there IGP Y.com 205.10.16.0/20
SJ1 /21 LA1 X.com NY1 /21 DC1
13
Network Design X.com is running OSPF
MED form X.com should be reflected as OSPF metric Redistribute BGP into OSPF as external type 2 metric
14
Network Design Configuration SJ1 router ospf 1
network area 0 redistribute bgp 132 route-map MED route-map MED permit 10 match ip address 2 set metric 5 match ip address 1 access-list 1 permit access-list 2 permit
15
Network Design X.com is advertising different ranges of the same class B to the Y.com X.com want Y.com to do optimal routing X.com will advertise MEDs to Y.com for optimal routing
16
Network Design Configuration SJ1 router bgp 132
network mask network mask aggregate-address summary-only neighbor remote-as 321 neighbor remote-as 441 neighbor remote-as 321 unsupress-map foo1 route-map foo1 permit 10 match ip address 1 set metric 2 set community no-export route-map foo permit 20 match ip address 2 access-list 1 permit access-list 2 permit
17
Network Design Y.com is running EIGRP as their IGP
MED should be redistributed into EIGRP metric properly EIGRP metric should be such that every router selects the correct egress routers
18
Network Design Y.com X.com 131.108.0.0/16 205.10.0.0/16 131.108.0.0/17
B=56K D= 2000 usec LA1 B=155MB D= 100 usec /17 SJ1 B=10MB D= 1000 usec R2 Y.com X.com /16 Network A R1 /16 DC1 B=100 MB D= 100 usec B=128K D= 2100 usec NY1 /17 NY1 B=56K D= 1000 usec Network B
19
[ ( ) ] Network Design Lets talk about EIGRP metric: EIGRP metric 10 *
7 10 min BW [ ( ) ] + Sum of delays * 256
20
Network Design Metric for router R1 to reach LA1 is Minimum BW = 10 MB
Total delay = = 1100 R1 metric = [( /10000)+1100)] *256 R1 metric =
21
Network Design Metric for router R1 to reach DC1 Minimum BW = 128K
Total delay = = 2200 R1 metric = [( /128) ] * 256 R1 metric =
22
Network Design Lets look at the metric for the redistributed route form BGP R1 should always prefer LA1 for /17 R1 should always prefer DC1 for /17 For set metric in route-map the values are set metric BW Delay Reliability Load MTU
23
Network Design Configuration LA1# router eigrp 321
redistribute bgp 321 route-map MED route-map MED permit 10 match ip address 1 set metric route-map MED permit 20 match ip address 2 set metric access-list 1 permit access-list 2 permit
24
Network Design Configuration DC1# router eigrp 321
redistribute bgp 321 route-map MED route-map MED permit 10 match ip address 1 set metric route-map MED permit 20 match ip address 2 access-list 1 permit access-list 2 permit
25
Network Design How did we come up with metric values:
R1 metric to LA1 = R1 metric to DC1 = So R2 metric for destination should be larger than for network /17 [ ( /155000) (current delay) + x ] * 256 = [ ( x) ] = /256 x = 80162 We rounded up to 80200to be on the safe side
26
Network Design Metric verification For Subnets 131.108.0.0/17
We want R2 to prefer LA1 Total delay = =101 R1 metric = [ ( /155000) + 101] * 256 R1 metric = 42496
27
Network Design R2 metric to reach network 131.108.128.0/17 via LA
Total delay = = R2 metric = [( /155000) ] *256 R2 metric =
28
Network Design R1 metric to reach 131.108.0.0/17 via DC1
Total delay = = 2201 R1 metric = [ ( /128) ] * 256 R1 metric =
29
Network Design R2 metric to reach 131.108.128.0/17 via DC1
Total delay = = 3201 R2 metric = [ ( /128) ] * 256 R2 metric =
30
Network Design Comparing the metrics for network /17 for R2 Via LA1 = Via DC1 = You can do same thing with BW as well
31
Network Design Y.com wants to keep ISP A as their primary service provider What will happen ISP A ISP C C Will Have Two Choices Shorter AS Length Will Take Care of Symmetrical Routing Packet Flow X.com Y.com Direct Connect and Longest Prefix Match
32
Network Design X.com is multihomed
X.com want ISP C to be primary and ISP B to be secondary ISP C ISP B AS 1 X.com Packet Flow
33
Network Design X.Com needs has two options
1. Community based local preference at the ISP 2. Set as-path prepend
34
Network Design Communities
Grouping of destinations into a community for applying common policies Each destination can belong to multiple communities Second global attribute in BGP after AS-path
35
Network Design X.com wants symmetrical routing
ISP will tell you the community number ISP will do the community-based local-pref matching
36
Network Design Config on X.com side on router that connects to ISP B
NY1# router bgp 132 aggregate-address summary-only network mask network mask neighbor remote-as 703 neighbor send-community neighbor route-map set community out route-map set community permit 10 match ip address 2 set community 132 :70 access-list 2 permit any
37
Network Design Config on the ISP B router ISPB# router bgp 703
neighbor remote-as 132 neighbor route-map community in route-map community permit 10 match community 1 set local-preference 70 route-map community permit 20 match community 2 ip community-list 1 permit 132:70 ip community-list 2 permit Internet
38
Network Design Use of as-path prepend config of NY1 NY1# router bgp132
neighbor remote-as 703 neighbor route-map path route-map path match ip address 1 set as-path prepend access-list 1 permit any
39
Network Design Case Study # 2
Organization Z.com has grown significantly They have four regions Their IGP is having scaling problems They want to isolate regional flaps Each region has a preferred service provider
40
Network Design One region has a very big hub and spoke setup
Their IGP is OSPF and they want a backup solution for each area Their addressing is not contiguous
41
Network Design They have one customer who has a router that is attached to two areas They have a situation where they have to bring one interface in two areas They want to do good summarization
42
Network Design Network setup for Z.com ISP Connection ISP Connection
Region 1 Region 2 Region 4 Region 3 ISP Connection ISP Connection
43
Network Design IGP is having scaling problems
Introduce BGP between the regions Flaps will get confined to regions only Send default form BGP routers
44
Network Design Each region with preferred service provider
If connected to same ISP just send MEDs If connected to different service providers, ask the ISP to do communities based local preference to avoid asymmetric routing If ISP does not support communities, use Cisco feature (as-path prepend)
45
Network Design When connected to same provider
For rest of the world they are connected to one ISP Easier solution to asymmetrical routing MED send to the ISP will take care of return path Each regional BGP router will send out default within its region
46
Network Design Multihomed to different ISP
Use community-based, local-preference Ask ISP to set their policies according to your requirement
47
Network Design One region has a very large hub and spoke Problem
Protocols with neighbor relationship will not scale Lot of spoke sites You have multiple hub routers Static config becomes a problem
48
Network Design Hub and spoke Two options Multipoint interface
Point-to-point interface
49
Network Design 300 Stub Sites
50
Network Design IF you want to run OSPF
Cisco 7500 hundred can have up to 250 neighbors on a RSP4 without a problem Remote routers will have problems with database sync, flooding and SPF If we divide 30 remote routers in a single area then the Cisco 7500 is connected to 10 area it will have to run 10 SPFs, plus processing of summary LSA is more CPU intensive then router and network LSAs
51
Network Design If you run EIGRP on the hub router
Just send default route to the remote routers Problem still exist of queries, router will query all the neighbors for a lost route
52
Network Design Three choices OSPF with totally stubby areas
RIP2 if the interface is multipoint ODR if the interface is multipoint or point to point
53
Network Design OSPF with totally stubby area
Central router will have connection to 10 areas Flap within an area will just cause one summary LSA for area 0 and two SPF’s one for the local area where flap happened and one for area 0 Remote areas don’t have to maintain large database
54
Network Design RIPv2 is the second option
Just send default route via RIP Receive updates form remote routers via RIP RIP works well with multipoint interface
55
Network Design ODR is third option
Install IP stub route on the hub router Does not require to configure IP routing protocol on the spoke router Spoke router advertises IP prefixes of all the connected interfaces
56
Network Design Routing information is propagated via CDP
Use IP filtering to limit the number of network prefixes that a hub router will permit Configuration Router odr 1 Distribute-list 1
57
Network Design Stub routing On stub router do not configure IP routing
Router is automatically considered to be a stub if no dynamic routing protocol is configured Configure default route on the stub router Configure IP classless on stub router
58
Network Design How should we propagate all the stub routes in the core
On the hub router create a static null route for all the remote sites redistribute the null route into the ospf
59
Network Design Each area must have back up if the primary link to area 0 is down Problem Each backup router must have a backup connection within its area] If we have large number of areas then it does not scale
60
Network Design Problem Backup to a site that belongs to same area
ISDN Primary ATM Area 1 R1 R2
61
Network Design Solution Use virtual profiles Put a central NAS
Depending on the dialing router, virtual interface gets assigned to that area Area config and IP address for virtual interface are dynamic Do not configure the dialing routers interface as demand circuit
62
Network Design Virtual profiles Area 3 Area 4 Area 0 AAA Server
Virtual Interface Area 3 Area 4 NAS Area 0 Dial Backup Primary Link
63
Network Design Discontiguous subnets Area 2 Area 1 D Region 2 A B C
Backbone Area 0 BGP Router F E Subnet 1–63 FDDI Dual Ring Area 2 FDDI Dual Ring Area 1 Subnet 18, 19, 21
64
Network Design Problem Solution
I have my subnets scattered all around the regions, can I summarize? Solution Summarize at the site that has the largest blocks and then leak specifics Longest prefix matching will take care of routing Migrate the address to the correct region
65
Network Design Solution (Cont.)
Router F should advertise the least specific summary No configuration needed for router E let it leak the specific subnets Config for router F router ospf 1 network area 2 range ip route null0
66
Network Design Solution (Cont.)
For BGP regional routers leak the subnet routes through BGP from router D Config for router D router bgp 2 network mask network mask network mask no auto-summary
67
Network Design Solution (Cont.)
From router A generate a longer summary then router F Config for router A router bgp 1 network mask no auto summary ip route null0
68
Network Design Router B and C will not advertise the specific subnets via BGP, because they are covered in the range Both routers do have the specific routes via OSPF Routing table of router B and C will have
69
Network Design Core routers will have all the routes
Routing table of A and D will have following route
70
Network Design Where do I make my area? BGP Router Area 2 Area 1
FDDI Dual Ring Area 2 FDDI Dual Ring Area 1 Customer Router
71
Network Design Problem Solution Where do I define my area
Critical site that is dual attached Solution Don’t run OSPF on the Ethernet Use static routes at the customer router Redistribute connected for the Ethernet on both routers
72
Network Design Problem Which area do I put my interface in? Area 0
Should this Interface Be in Area 1 or Area 0? Area 1
73
Network Design Sub optimal routing if in area 0
Sub optimal if in area 1 Bring the interface in both areas How: Create a virtual link between the two ABRs
74
Network Design Use maximum route summarization
Configure a null route to the summary on the ABR to avoid routing loops Use loopback on the ABR to generate the metric for the summary LSA Currently OSPF uses lowest cost for the summary route, rfc2178 has changed this behavior to the highest metric
75
Network Design Area 2 Summary route to area 0 ABR
FR FDDI Dual Ring Subnets to
76
Network Design Config of ABR router ospf 1
network area 2 area 2 range int loop 0 ip address ip route null0
77
Network Design NBMA networks (ATM, frame, etc.) Four choices Broadcast
Non-broadcast Point to point Point to multipoint
78
Network Design Broadcast model Optimal flooding via DR
No per VC costing Requires full mesh all the time Good if your NBMA cloud has rerouting capabilities
79
Network Design NBMA Very similar to broadcast model
Flooding same as broadcast model Requires more config because neighbors need to be defined
80
Network Design Point to point
Each VC is like a regular point-to-point interface Per VC costing Faster convergence Partial mesh allowed Flooding is a problem
81
Network Design Point to multipoint One IP address for the cloud
No per VC costing available for now Sub optimal flooding Maintains host routes for every router in the cloud
82
Network Design Which model to choose
If rerouting at Layer 2 is not an issue or your cloud is capable of rerouting use broadcast model for large mesh Otherwise use point to point, it is more robust
83
Network Design If you are running IS-IS use point to point
Use mesh groups for IS-IS OSPF does not have mesh groups for now but will have it soon
84
Network Design Mesh groups Normal interface (do normal flooding)
Blocked (never send LSPs over this interface) Part of a mesh group When LSP is received on non-meshgroup interface, flood it out over all interfaces (meshgroup or not) If LSP is received on meshgroup interface, flood on non-meshgroup interfaces, but don’t flood on other interfaces in this mesh group
85
Network Design Mesh groups Normal interface (do normal flooding)
Blocked (never send LSPs over this interface) Part of a mesh group When LSP is received on non-meshgroup interface, flood it out over all interfaces (meshgroup or not) If LSP is received on meshgroup interface, flood on non-meshgroup interfaces, but don’t flood on other interfaces in this meshgroup
86
Network Migration Reasons Current routing protocol is classfull
Does not support VLSM Slower convergence Having scaling problems
87
Network Migration Classless protocols
No distinction between classes of network Groups class C or B networks What does IP classless command do on Cisco router
88
Network Migration First step
If large network, divide the network using BGP Configure BGP at the core routers Passive existing protocol on the core links Generate default from the core routers
89
Network Migration Configuration of router for region 1
Similar configuration for each core router router bgp 1 network network mask neighbor remote-as 2 neighbor remote-as 3 neighbor remote-as 4
90
Network Migration Migrating from distance vector to link state
Configure link state with higher admin distance Leave your current distance vector untouched Start removing your distance vector Link state should install the route once the distance vector route is gone Make sure you take care of offset-list (if you have defined it) for RIP
91
Network Migration Configuration router rip network 140.10.0.0
router ospf 1 network area 0 distance 130
92
Network Migration Migration from IGRP to EIGRP
Automatic redistribution IGRP routes are advertised into EIGRP as external Metric is compared before installing the route in the routing table
93
Network Migration Discontiguous networks Area 2 Area 1 D Region 2 A B
Subnet 27, 28, 33 Subnet 48–55 Backbone Area 0 BGP Router F E Subnet 1–63 FDDI Dual Ring Area 2 FDDI Dual Ring Area 1 Subnet 18, 19, 21
94
Network Migration Migration for discontiguous subnets
Move the addresses from area 1 of region 1, no config is required Move the addresses from region 2, BGP will not find the matching mask and will stop advertising the routes Later take the network and mask commands out
95
Network Migration Address moved from region 3 needs changes
Move this range last so the routing table has already shrunk Stop aggregation from region 3 Start leaking specific routes so you can freely move addresses Again once addresses have moved remove the network and mask command from the BGP config
96
1054_03F8_c2 96
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.