Você está na página 1de 44

Introduction to MPLS

and Traffic Engineering


Outline
 Traditional IP Routing
 Forwarding and routing
 Problems with IP routing
 Motivations behind MPLS
 MPLS Terminology and Operation
 MPLS Label, LSR and LSP, LFIB Vs FIB
 Transport of an IP packet over MPLS
 More MPLS terminology
 Traffic Engineering [with MPLS]
 Nomenclature
 Requirements
 Examples

2
Outline
 Traditional IP Routing
 Forwarding and routing
 Problems with IP routing
 Motivations behind MPLS
 MPLS Terminology and Operation
 MPLS Label, LSR and LSP, LFIB Vs FIB
 Transport of an IP packet over MPLS
 More MPLS terminology
 Traffic Engineering [with MPLS]
 Nomenclature
 Requirements
 Examples

3
Forwarding and routing
 Forwarding:
 Passing a packet to the next hop router
 Routing:
 Computing the “best” path to the destination
 IP routing – includes routing and forwarding
 Each router makes the forwarding decision
 Each router makes the routing decision
 MPLS routing
 Only one router (source) makes the routing decision
 Intermediate routers make the forwarding decision

4
IP versus MPLS routing
 IP routing
 Each IP datagram is routed independently
 Routing and forwarding is destination-based
 Routers look at the destination addresses
 May lead to congestion in parts of the network
 MPLS routing
 A path is computed “in advance” and a “virtual
circuit” is established from ingress to egress
 An MPLS path from ingress to egress node is
called a label switched path (LSP)

5
How IP routing works

Searching
Longest
Prefix Match
in FIB (Too
Slow)
6
Problems with IP routing
 Too slow
 IP lookup (longest prefix matching) “was” a
major bottleneck in high performance routers
 This was made worse by the fact that IP
forwarding requires complex lookup operation
at every hop along the path
 Too rigid – no flexibility
 Routing decisions are destination-based
 Not scalable in some desirable applications
 When mapping IP traffic onto ATM

7
IP routing rigidity example
D
1 1
A
S A B
B
1 2
C

 Packet 1: Destination A
 Packet 2: Destination B
 S computes shortest paths to A and B; finds D as next hop
 Both packets will follow the same path
 Leads to IP hotspots!
 Solution?
 Try to divert the traffic onto alternate paths

8
IP routing rigidity example
D
1 4
A
S A B
B
1 2
C

 Increase the cost of link DA from 1 to 4


 Traffic is diverted away from node D
 A new IP hotspot is created!
 Solution(?): Network Engineering
 Put more bandwidth where the traffic is!
 Leads to underutilized links; not suitable for large networks

9
Motivations behind MPLS
 Avoid [slow] IP lookup
 Led to the development of IP switching in 1996
 Provide some scalability for IP over ATM
 Evolve routing functionality
 Control was too closely tied to forwarding

 Evolution of routing functionality led to some


other benefits
 Explicit path routing
 Provision of service differentiation (QoS)

10
IP routing versus MPLS routing
Traditional IP Label
Multiprotocol Routing
Switching (MPLS)
1 2

S D

3 4 5

MPLS allows overriding shortest paths!

11
Outline
 Traditional IP Routing
 Forwarding and routing
 Problems with IP routing
 Motivations behind MPLS
 MPLS Terminology and Operation
 MPLS Label, LSR and LSP, LFIB Vs FIB
 Transport of an IP packet over MPLS
 More MPLS terminology
 Traffic Engineering [with MPLS]
 Nomenclature
 Requirements
 Examples

12
MPLS label
 To avoid IP lookup MPLS packets carry
extra information called “Label”
 Packet forwarding decision is made using
label-based lookups
Label IP Datagram

 Labels have local significance only!


 How routing along explicit path works?

13
Routing along explicit paths
 Idea: Let the source make the complete routing
decision
 How is this accomplished?
 Let the ingress attach a label to the IP packet and let
intermediate routers make forwarding decisions only
 On what basis should you choose different paths
for different flows?
 Define some constraints and hope that the constraints
will take “some” traffic away from the hotspot!
 Use CSPF instead of SPF (shortest path first)

14
Label, LSP and LSR
 Label 01234567890123456789012345678901
Label | Exp|S| TTL

Label = 20 bits
Exp = Experimental, 3 bits
S = Bottom of stack, 1bit
TTL = Time to live, 8 bits

 Router that supports MPLS is known as label


switching router (LSR)
 Path which is followed using labels is called LSP

15
LFIB versus FIB
 Labels are searched in LFIB whereas normal IP
Routing uses FIB to search longest prefix match
for a destination IP address
 Why switching based on labels is faster?
 LFIB has fewer entries
 Routing table FIB has larger number of entries???

 In LFIB, label is an exact match


 In FIB, IP is longest prefix match

16
MPLS Flow Progress

D
R1 LSR4 R2
LSR1
D

LSR6 destination

LSR3
LSR2 R1 and R2 are
regular routers
LSR5

1 - R1 receives a packet for destination D connected to R2

17
MPLS Flow Progress

D
R1 LSR4 R2
LSR1
D

LSR6 destination

LSR3
LSR2

LSR5

2 - R1 determines the next hop as LSR1 and forwards the packet


(Makes a routing as well as a forwarding decision)

18
MPLS Flow Progress

R1 LSR4 R2
LSR1
31 D
D

LSR6 destination

LSR3
LSR2

LSR5

3 – LSR1 establishes a path to LSR6 and “PUSHES/ATTACHES” a label


(Makes a routing as well as a forwarding decision)
19
MPLS Flow Progress

R1 LSR4 R2
LSR1
D

LSR6 destination

LSR3 17 D

LSR2

LSR5 Labels have local


signifacance!
4 – LSR3 just looks at the incoming label
LSR3 “SWAPS” with another label before forwarding

20
MPLS Flow Progress

R1 LSR4 R2
LSR1
D

LSR6 destination

LSR3 17 D

LSR2
Path within MPLS cloud
is pre-established:
LSR5
LSP (label-switched path)

5 – LSR6 looks at the incoming label


LSR6 “POPS/REMOVES” the label before forwarding to R2

21
MPLS and explicit routing recap
 Who establishes the LSPs in advance?
 Ingress routers
 How do ingress routers decide not to always take
the shortest path?
 Ingress routers use CSPF (constrained shortest path
first) instead of SPF
 Examples of constraints:
 Do not use links left with less than 7Mb/s bandwidth

 Do not use blue-colored links for this request

 Use a path with delay less than 130ms

22
CSPF
 What is the mechanism? (in typical cases!)
 First prune all links not fulfilling constrains
 Now find shortest path on the rest of the topology
 Requires some reservation mechanism
 Changing state of the network must also be
recorded and propagated
 For example, ingress needs to know how much bandwidth
is left on links
 The information is propagated by means of routing
protocols and their extensions

23
More MPLS terminology

Upstream Downstream

172.68.10/24
LSR1 LSR2

Data

24
Label advertisement
 Always downstream to upstream label
advertisement and distribution

Upstream Downstream
Use label 5 for destination
171.68.32/24
171.68.32/24

MPLS Data Packet LSR2


LSR1 with label 5 travels

25
Label advertisement
 Label advertisement can be downstream
unsolicited or downstream on-demand
Upstream Sends label Downstream
Without any Request
171.68.32/24

LSR1 LSR2

Upstream Sends label ONLY after Downstream


receiving request
171.68.32/24

LSR1 LSR2
Request For label

26
Label distribution
 Label distribution can be ordered or unordered
 First we see an example of ordered label distribution

Label

Egress LSR
Ingress LSR

27
Label distribution
 Label distribution can be ordered or unordered
 Next we see an example of unordered label distribution

Label

Label

Egress LSR
Ingress LSR

28
Label operations
 Advertisement
 Downstream unsolicited
 Downstream on-demand
 Distribution
 Ordered
 Unordered

29
Outline
 Traditional IP Routing
 Forwarding and routing
 Problems with IP routing
 Motivations behind MPLS
 MPLS Terminology and Operation
 MPLS Label, LSR and LSP, LFIB Vs FIB
 Transport of an IP packet over MPLS
 More MPLS terminology
 Traffic Engineering [with MPLS]
 Nomenclature
 Requirements
 Examples

30
Traffic Engineering
Traffic Engineering with MPLS
(Application of CSPF)
What is traffic engineering?
 Performance optimization of operational networks
 optimizing resource utilization
 optimizing traffic performance
 reliable network operation
 How is traffic engineered?
 measurement, modeling, characterization, and
control of Internet traffic
 Why?
 high cost of network assets
 service differentiation

32
Traffic engineering
 Recall the IP hotspot problem

 The ability to move traffic away from the


shortest path calculated by the IGP (such as
OSPF) to a less congested path

 IP: changing a metric will cause ALL the traffic


to divert to the less congested path
 MPLS: allows explicit routing (using CSPF) and
setup of such explicitly computed LSPs

33
MPLS-TE: How to do it?
 LSPs are set up by LSRs based on information
they learn from routing protocols (IGPs)
 This defeats the purpose!
 If we were to use “shortest path”, IGP was okay

34
MPLS TE: How we actually do it?
 MPLS TE Requires:
 Enhancements to routing protocols
 OSPF-TE
 ISIS-TE
 Enhancement to signaling protocols to allow
explicit constraint based routing
 RSVP-TE and CR-LDP
 Constraint based routing
 Explicit route selection
 Recovery mechanisms defined

35
Signaling mechanisms
 RSVP-TE
 Extensions to RSVP for traffic engineering
 BGP-4
 Carrying label information in BGP-4
 CR-LDP
 A label distribution protocol that distributes labels
determined based on constraint based routing

 RSVP-TE and CR-LDP both do label distribution


and path reservation – use any one of them!

36
RSVP-TE

Basic flow of LSP set-up using RSVP

37
RSVP-TE PATH Message

 PATH message is used to establish state and


request label assignment
 R1 transmits a PATH message addressed to R9

38
RSVP-TE RESV Message

 RESV is used to distribute labels after reserving resources


 R9 transmits a RESV message, with label=3, to R8
 R8 and R4 store “outbound” label and allocate an “inbound” label.
They also transmit RESV with inbound label to upstream LSR
 R1 binds label to forwarding equivalence class (FEC)

39
Rerouting LSP tunnels
 When a more “optimal” route/path
becomes available
 When a failure of a resource occurs along
a TE LSP

 Make-before-break mechanism
 Adaptive, smooth rerouting and traffic
transfer before tearing down the old LSP
 Not disruptive to traffic

40
Recovering LSP tunnels
LSP Set-up

41
Protection LSP set up

42
Protection LSP

43
References
 http://www.cisco.com/univercd/cc/td/doc/product/s
oftware/ios120/120newft/120limit/120s/120s5/mpls
_te.htm
 RFC 2702 “Requirements for Traffic Engineering
Over MPLS”
 RFC 3031 “Multiprotocol Label Switching
Architecture”
 RFC 3272 “Overview and Principles of Internet
Traffic Engineering”
 RFC 3346 “Applicability Statement for Traffic
Engineering with MPLS”
 MPLS Forum (http://www.mplsforum.org)

44

Você também pode gostar