CS672: MPLS Architecture, Applications and FaultTolerance

1
CS672 MPLS Architecture, Applications and
Fault-Tolerance
2
Stub-AS Routing Options

• Provider may use one of the following options to
  learn about customer routes
• Static routing
• IGP
• BGP
• Customer may use one of the following options
• Default routing
• IGP
• BGP

3
Multihomed Stub AS Routing

• Provider may use one of the following options to
  learn about customer routes
• IGP
• BGP (preferable for better load balancing)
• Customer may use one of the following options
• IGP
• BGP (preferable for better load balancing)

4
Transit AS Routing

• In addition to advertising its own routes, a
  transit AS also advertises routes learned from
  other ASs.
• Upon learning routes advertised by a transit AS,
  other ASs direct traffic toward transit AS for
  certain routes.
• A transit AS uses BGP-4 for routing between other
  ASs

5
Transit AS
destinations (d7,d8,d9)
AS 3
(d4,d5,d6)
BGP
(d1,d2,d3)
BGP
(d1,d2,d3,, d8,d9)
destinations (d1,d2,d3)
destinations (d4,d5,d6)
AS 1
AS 2
6
Guidelines for BGP Usage

• The Internet may be considered as an
  interconnection of transit, multihomed
  (non-transit), and stub ASs.
• Single- and Multihomed Stub Autonomous Systems
• may use protocols other than BGP (e.g., EGP, IGP,
  static routes)
• however, use of BGP for advertising reachability
  information is recommended
• Transit Autonomous Systems
• use BGP-4 for distributing routing information
  between multiple transit ASs.
• Using BGP inside an AS
• BGP is mainly used for exchanging routing
  information between autonomous systems (inter-AS)
• However, BGP can be used to carry routing
  information across inside the AS

7
Introduction to BGP-4

• Border Gateway Protocol (BGP) is an an exterior
  gateway protocol
• main function of BGP is to allow exchange of
  routing information between BGP peers in
  different ASs.
• routers that run BGP are known as BGP speakers.
• BGP version 4 (BGP-4) is the de-facto standard
  for inter-AS routing
• BGP-4 obsoletes the earlier versions 1-3
• BGP can also be used inside of an AS (e.g., as a
  pipe) to carry external routing information
  across the AS.
• Use of BGP inside an AS protects non-BGP speakers
  from potentially large number of external routing
  information.
• Provides better control in selecting the exit and
  entrance points for the ASs traffic.

8
Motivations for BGP-4

• Existing IGP (e.g., OSPF, IS-IS, RIP) protocols
  are not designed for routing in large global
  networks. Some of the factors include
• Lack of mechanism to partition large networks
  into politically independent technical
  administrations (e.g., governments, institutions
  etc.)
• Count to infinity limitation of distance vectors
  protocols such as RIP
• Excessive control traffic generated by periodic
  updates of entire routing table (e.g., RIP).

9
Motivations for BGP-4

• Lack of mechanism to suppress excessive control
  traffic generated due unstable links flaps (i.e.,
  up/down events) in link state protocols (e.g.,
  OSPF)
• The processing and path recompilations due to
  excessive control traffic can easily overload
  control processor in the router
• Lack of support for route aggregation and CIDR
• Growth of table size becomes unmanageable in
  large networks
• Lack of effective mechanism to enforce policies
  (e.g., selection of routes,control of route
  redistribution)

10
BGP Routing Model

• BGP is a path vector protocol because it
  exchanges complete AS path information between
  peers.
• A BGP speaker can constructs a graph of
  autonomous systems interconnection based on the
  exchanged routing information
• The use of full path information enables
• Detection of BGP speaker routing loops
• Elimination of count-to-infinity problems
  associated with distance vector protocols
• Enforcement of policies based on configurations
• Provides effective mechanisms for route selection
  and controlling redistribution of routing
  information.
• BGP supports CIDR and aggregation of routing
  information
• Reduction of table size
• Scalability

11
AS Topology Graph
Path
AS2
AS1
AS4
Route (AS1, AS2, AS3)
AS3
Routing Flow
Traffic flow
Vertex represents an AS Path (or edge)
represents interconnection between two AS. Route
collection of path information form a route.
12
BGP Session

• BGP uses TCP for reliable delivery of its
  messages
• For establishing connection with neighbors, BGP
  uses TCP port 179
• Before any routing exchange can take place, a TCP
  connection must be established between two BGP
  peers
• To establish TCP connection, BGP uses TCP port
  179
• After TCP connection establishment, BGP session
  parameters are negotiated.
• After BGP session parameters have been
  negotiated
• during the initial update the complete entire
  routing table information is exchanged between
  neighbors
• afterwards, only incremental updates take place.

13
External BGP Session

• BGP speakers may reside in the same AS or
  different ASs
• A BGP session between two speakers in different
  ASs is known as external BGP (eBGP) session.
• An eBGP session requires that the two BGP peers
  be directly connected (e.g., shared data between
  two ASs)
• Cisco routers have removed the requirement of
  physical connectivity for eBGP session.

14
Internal BGP Session

• A session between two speakers in the same AS is
  known as internal BGP (iBGP) session.
• An eBGP session can be established between
  directly or indirectly connected BGP speakers.

15
eBGP and iBGP Sessions

• How does a BGP speaker know whether a BGP session
  internal or external?
• During BGP session establishment, BGP peers
  exchange AS numbers.
• Therefore, determination of a BGP session as iBGP
  or eBGP is based on comparison of AS numbers.
• If the AS numbers are same, the session is iBGP.
• Otherwise, eBGP.

16
(No Transcript)
17
BGP Messages

• Each BGP message has a fixed size header that
  contains following fields
• Marker
• Length
• Type (i.e., BGP message type)
• BGP defines following message types
• Open
• KeepAlive
• Notification
• Update

18
BGP Message Fixed Header
Total size of the fixed header 1621 19 Byte
19
BGP Open Message

• After establishing TCP connection, each speaker
  sends a BGP Open message
• BGP Open contains fields such as
• Version
• My Autonomous System
• Hold Time
• BGP Identifier
• Optional Parameters
• If the values of the parameters are acceptable,
  the receiving side sends back KeepAlive message
  as an acknowledgement.
• Upon receipt of KeepAlive, other messages (e.g.,
  Update, KeepAlive etc.) are exchanged.

20
BGP Open Message
21
BGP KeepAlive Message

• BGP peers send periodic KeepAlive messages to
  avoid Hold timer expiration
• For example, every 1/3 of Hold Timer period a
  KeepAlive message is sent
• The two BGP peers may negotiate to not send any
  KeepAlive. For example,
• If Hold Time 0, KeepAlive is never sent.
• KeepAlive message only contains fixed length
  header (19 bytes)

22
BGP Notification Message

• Whenever an error condition is encountered, BGP
  sends a Notification message.
• Following transmission of a notification message,
  the associated TCP connections is immediately
  closed.
• In addition to the fixed sized header,
  Notification message contains following fields
• Error Code (e.g., 1 Message Hdr Error, 2Open
  Msg Error, )
• Error Subcode
• Data to help troubleshooting the error.

23
(No Transcript)
24
BGP Update Message

• Update is the most important BGP message because
  it is used for exchanging routing information
  between BGP peers.
• Based on routing information exchanged via Update
  messages, BGP speakers construct graph of the
  ASs topology. In short, BGP Update contains all
  the required information that enables BGP
  speakers to perform loop-free-routing.
• In addition to mandatory fixed-sized header, BGP
  Update may contain one or more of the following
  optional fields
• Network Layer Reachability Information (NLRI)
• Withdrawn Routes
• Path Attributes

25
(No Transcript)
26
BGP Update Message

• In order to be able to exchange Update message,
  BGP session must be in established state
• Update message may simultaneously
• Advertise single feasible route
• But withdraw multiple routes
• What is a route?
• A route is an association of the path attributes
  with one or more IP destination prefixes
  contained in the NLRI.
• It is important to realize that all path
  attributes (carried in the Path Attribute field)
  apply to all IP destination prefixes listed in
  the NLRI field.
• This means, Update message cannot advertise more
  than one route

27
BGP Update Message

• How come more than one routes can be withdrawn in
  a single message?
• A BGP speaker receiving an Update message can
  unambiguously can identify a route to be
  withdrawn based on an IP destination address
  listed in the withdrawn field.
• As a result , it is possible to withdraw multiple
  routes simultaneously.

28
AS Topology Graph
Path
AS2
AS1
AS4
Route (AS1, AS2, AS3)
AS3
Routing Flow
Traffic flow
Vertex represents an AS Path (or edge)
represents interconnection between two AS. Route
collection of path information form a route.