Introduction to routing policy

1
Introduction to routing policy

• ECE697A Advanced Computer Networking
• Lei Liang
• October 2002

2
Outline

• Introduction to routing policy
• Implementation of routing policy
• Redundancy
• Symmetry
• Load balancing
• Multihoming

3
Introduction to routing policy

• What is routing policy?
• What is routing policy source and purpose
• Why we define routing policy?
• Example
• Elements of a routing policy

4
What is routing policy?

• Public description of the relationship between
  external BGP peers, and how network prefixes are
  exchanged between that AS and other ASes.
• Can also describe internal BGP peer relationship
• Allows you to control the routing information
  between the routing protocols and the routing
  tables

5
Routing policy source and purpose

• Source Routing information is generated by
  internal networking peers.
• Purpose To control the size and content of the
  routing tables, which routes are advertised, and
  which routes are considered the best to reach
  various destinations.
• Control points
• Import
• export

6
Why define a routing policy?

• Documentation
• Allows automatic generation of router
  configurations
• Scalability easy to scale large number of peers
  or transit customers
• Troubleshooting identify where routes are
  entering the network
• What routes are preferred when multiple routes
  exist

7
Routing policy example

• AS1 originates prefix d
• AS1 exports d to AS2
• AS2 imports
• AS2 exports d to AS3
• AS3 imports
• AS3 exports d to AS5
• AS5 imports

8
Routing policy example (Cont.)

• AS5 also imports d from AS4
• Which route does it prefer?

9
Elements of routing policy

• Written documentation policy document
• Handle different types of routes, differently
• Customer Inbound
• Customer Outbound
• Transit/Peer Inbound
• Transit/Peer Outbound

10
Customer Routes, Inbound

• Accept by prefix or by AS
• Maximum prefix size (/24)
• Customer Routes-Highest local preference
• MEDs are usually listened to (to compare update
  metric, the lower metric will be selected to make
  a decision inside the AS)
• MED a hint to external neighbors about the
  preferred path into an AS.
• Use a route-map or policy-statement to set
  appropriate communities on INGRESS.

11
Customer Routes, Outbound

• Full routes, partial routes, default route only
• Full routes can be sent with or without summary
  aggregates-providers do not aggregate towards
  customer
• Some providers allow you to choose amount of
  aggregation.
• Customer can do their own filtering, when ISP
  providing customer with rich communities

12
Peer and Transit Provider Routes Inbound

• Not usually filtered-sometimes filtered on prefix
  length
• For peers, some set limits on number of prefixes
• MEDs are not usually listened to
• Local Preference Hierarchy
• customer, private peering, public peering,
  transit, AS-Path length
• Use a route map or policy statement to set
  appropriate communities on INGRESS

13
Peer and Transit Provider Routes Outbound

• Usually filtered on the /24 boundary
• MEDs are usually sent but are rarely listened to
• Aggregates are normally announced for your own
  blocks
• For peers, usually only customer and internal
  routes are announced

14
Complexity of routing policy

• Problems?
• Policy can easily get very complex and result in
  even more complex router configuration

15
Issues to be considered in implementing router
policy

• Redundancy
• Providing multiple alternative paths for the
  traffic
• Usually having multiple connections to one or
  more ASes
• Symmetry
• Traffic that leaves the AS from an exit point
  return through the same point
• Load balancing
• The capability to divide traffic optimally over
  multiple link
• Must consider trade-offs in implementing routing
  policies

16
Redundancy

• Redundancy aims to warranty uninterrupted
  connectivity.
• Connectivity problems occur at the router, power,
  cabling, physical access line.
• Human errors, software errors, physical error, or
  bad weather can affect connectivity.
• Advantage
• Improve connectivity/reliability
• Disadvantage
• May reduce symmetry
• Make the traffic more unpredictable

17
Redundancy and routing information

• Because redundancy refers to the existence of
  alternative routes, additional routing
  information needs to be kept in routing tables.
• One way to reduce routing overhead is to
  designate default routes.
• Default routing provides backup routes in case
  primary connection fails.

18
Set default routes

• Default routes Traffic to destinations that is
  unknown to the router is sent to the default
  outlet.
• Default route represented by network mask
  combination 0.0.0.0/0.0.0.0 (0/0)
• Dynamically learned defaults
• Statically set defaults

19
Dynamically learned defaults

• The route can be exchanged as a dynamic
  advertisement between routers.
• Dynamic defaults can be learned via BGP or IGP
• Destination is down, the route disappear
• For redundancy purpose, we should receive
  defaults from multiple sources.
• Give a degree of local preference over which
  default is primary and which is backup.

20
Statically set defaults

• Filter dynamically learned defaults to avoid
  undesirable situation. (traffic ends up)
• Provide more control over routing behaviors
• The 0/0 static route can point to a network
  number, a gateway address, or a physical
  interface as the default path.

21
Symmetry

• Symmetry means having traffic that leaves the AS
  from a certain exit point returns through the
  same point.
• Easy to achieve if there is only a single exit
• Traffic tends to be asymmetric due to redundancy
  and multiple connections
• Lack of symmetry means lack of control over how
  traffic flows into and out of ASes

22
More about symmetry?

• Customers and providers like to see traffic come
  back close to or at the same point it left the AS
  to minimize potential delays
• On the other hand, traffic may go as far as
  possible on the network to avoid latency or
  congestion on the peer network
• Asymmetry is unavoidable
• asymmetric traffic is often acceptable, depending
  on the applications
• usually asymmetric traffic is not a significant
  problem

23
To accommodate symmetry

• One should designate a primary link and make the
  utmost effort to direct the majority of traffic
  to flow on this link
• Redundancy and symmetry are in conflict with each
  other

24
Load balancing

• Load balancing is the capability to distribute
  traffic optimally over multiple links
• not to distribute traffic equally over
  connections
• Load balancing tries to achieve a traffic
  distribution pattern that optimally utilizes the
  multiple links that provide redundancy

25
What affects load balancing?

• Two types of traffic, incoming and outgoing
• Incoming traffic is affected by how the AS
  advertises its networks to the outside world
• Outgoing traffic is affected by the routing
  updates coming in from outside ASes.

26
Incoming and Outgoing traffic

• We can affect inbound traffic by applying
  attributes to outbound routing announcement
  because that is how our routes are learned by
  others.
• We can affect outbound traffic by applying
  attributes to inbound routing announcements
  because how our network learns routes affects
  outbound traffic.

27
Designing redundancy, symmetry and load balancing

• Responsibility of the operator to choose and
  configure the correct attributes and filtering to
  achieve the desired outcome

28
Scenarios

• Depending on how many links external to the local
  network
• Single homing
• Multihoming

29
Example 1 dynamically learned defaults

• RTA originates a default route 0.0.0.0/0 toward
  RTC only. IBGP neighbors such as RTF, will not
  get the default.

30
Example 1 (cont)

• Router bgp 3
• no synchronization
• network 172.16.1.0 mask 255.255.255.0
• neighbor 172.16.20.1 remote-as 1
• neighbor 172.16.20.1 default-originate
• no auto-summary
• The default-originate option of the neighbor
  router subcommand will cause 0/0 to be sent
  toward RTC.

31
Example 2 statically set defaults

• Instead of dynamically learning the 0/0 default,
  a router can set its own default statically. RTC
  set default to point toward network 192.78.0.0/16

32
Example 2 (cont)

• Router bgp 1
• network 192.68.11.0
• neighbor 172.16.20.2 remote-as 3
• neighbor 192.68.6.1 remote-as 2
• no auto-summary
• ip route 0.0.0.0 0.0.0.0 192.78.0.0
• Route with shorter distance are preferred over
  routes with a longer distance

33
Single-homing

• Customer connects to the Internet via a single
  connection to an ISP
• Customer usually can be served by pointing
  defaults towards the provider
• Provider can install static routing to reach the
  customer

34
Single-homing

• Advantage
• Least expensive and more effective
• Reduce memory usage and processing overhead
• Static default configuration
• Disadvantage
• Poor reliability

35
Multihoming

• Multihoming Definition
• More than one link external to the local network
• Multihoming Scenarios
• Multihoming to a single provider
• Multihoming to different providers
• Customers of the same provider with a backup link
• Customers of different providers with a backup
  link

36
Multihoming to a single provider

• Definition Two or more links to the same ISP
• Case
• Default only, one primary, and one backup

37
Example Default only, one primary, and one
backup (single provider)

• AS1 provider
• AS3 customer multihomed to AS1.
• RTA running default toward AS1

38
The policies apply to this scenario

• Outbound traffic from AS3 always go on the NY
  link unless it fails.
• Configuring two static routes in RTA pointing the
  defaults toward the two links.
• Inbound traffic to AS3 always go on the NY link
  unless it fails
• sending difference metrics toward AS1 on both
  links with a lower metric on the NY link.
• Prevent any BGP updates from coming into AS3
• configuring a route map or prefix list

39
RTA configuration

• router bgp 3
• network 172.16.220. 0 mask 255.255.255.0
• neighbor 172.16.20.1 remote-as 1
• neighbor 172.16.20.1 route-map BLOCK in
• neighbor 172.16.20.1 route-map SETMETRIC1
  out (RTC)
• neighbor 192.68.9.2 remote-as 1
• neighbor 172.68.9.2 route-map BLOCK in
• neighbor 172.68.9.2 route-map SETMETRIC2 out
  (RTD)
• no auto-summary
• ip route 0.0.0.0 0.0.0.0 172.16.20.1 50 (RTC SF
  backup)
• ip route 0.0.0.0 0.0.0.0 192.68.9.2.40 (RTD NY
  primary)
• route-map SETMETRIC permit 10 (RTC)
• set metric 100
• route-map SETMETRIC permit 10 (RTD)
• set metric 50

40
Multihoming to different ISP providers

• Definition Two or more links to different ISP
• Case
• Default, primary, and backup, plus full and
  partial routing

41
Example Default, primary and backup,
full/partial (Multiple providers)

• AS3 multihomed to AS1, AS2

42
The policies apply to this scenario (Multiple
providers)

• AS3 accept AS1 partial routes only via the SF
  link. All other internet routes accepted via the
  NY link.
• AS3 accept a default route from AS1 in case NY
  link failure.
• AS3 cannot be a transit network from AS1 and AS2.

43
RTF configuration (Multiple providers)

• router bgp 3
• no synchronization
• network 172.16.1.0 mask 255.255.255.0
• network 172.16.220.0 mask 255.255.255.0
• neighbor 172.16.1.1 remote-as 3
• neighbor 172.16.1.1 next-hop-self
• neighbor 192.68.5.2 remote-as 2
• neighbor 192.68.5.2 route-map PREPEND_PATH out
  (to RTD)
• no auto-summary
• ip as-path access-list 2 permit
• Access-list 1 permit 172.16.220.0 0.0.0.255
• route-map PREPEND-PATH permit 10
• match ip address 1
• set as-path prepend 3 (3 is specified AS
  number)
• route-map PREPEND-PATH permit 20 (advertised)
• match as-path 2

44
Customers of the same provdier with a backup link

• The private link can be used as a backup link
  when an Internet link fails

45
Example private link used as backup (customers
of the same provider)
46
The policies apply to this scenario (customers of
the same provider)

• AS3 offers services to AS1 and AS2.
• AS1 and AS2 have private link as backup
• Assume AS1 and AS2 getting full internet routes,
  AS1 and AS2 advertise each others routes to AS3.
• AS3 can reach AS1network via AS2 and AS2s
  network via AS1.
• Handled by the BGP default behavior
• Using BGP policies, by manipulating the
  LOCAL_PREF

47
RTC configuration (customers of the same provider)

• router bgp 1
• network 192.68.11.0
• neighbor 172.16.20.2 remote-as 3
• neighbor 172.16.20.2 route-map PREF_FROM_AS3 in
• neighbor 192.68.6.1 remote-as 2
• neighbor 172.16.20.1 route-map PREF_FROM_AS2 in
• no auto-summary
• ip as-path access-list 1 permit _2_
• route-map PREF_FROM_AS3 permit 10
• match as-path 1
• set local-preference 100 (default from AS2 and
  AS3)
• route-map PREF_FROM_AS3 permit 20
• set local-preference 300 (all other updates)
• route-map PREF_FROM_AS2 permit 10
• set local-preference 200 (from AS2)

48
Customers of different providers with a backup
link

• The community approach
• Dealing with adding and removing customers with
  dynamically set the customers policies
• The AS path manipulation approach
• An alternative to the community approach

49
Example The community approach (Customers of
different providers)
50
The policies apply to this scenario (Customers of
different providers)

• AS1 and AS2 are getting their service from
  provider AS4 and AS3 respectively, unless a link
  to a provider fails
• AS1 and AS2 have a private link, if it goes down,
  the customer should be able to talk to one
  another via the provider

51
RTC configuration (Customers of different
providers)

• router bgp 1
• network 192.68.11.0 mask 255.255.255.0
• neighbor 172.16.20.2 remote-as 1
• neighbor 172.16.20.2 send-community
• neighbor 172.16.20.2 route-map setcommunity out
• neighbor 172.16.20.2 filter-list 10 out
• neighbor 192.68.6.1 remote-as 2
• no auto-summary
• ip as-path access-list 2 permit _2_

52
RTC configuration (Customers of different
providers)

• ip as-path access-list 10 permit
• ip as-path access-list 10 permit 2
• (prevent AS4 learning AS1)
• route-map setcommunity permit 10
• match as-path 2
• setcommunity 440
• (via AS2 match_2_, set community 40)
• Route-map setcommunity permit 20
• (do not have community set)

53
Conclusion

• Routing policy acts as a traffic scheduler
• Routing policy decides one route from a set of
  routable path
• It aims to optimize the network performance

54
The end

• Question?