Transcript Mpls Basics And Applications

MPLS
Basics and
Applications
Peter Tomsu
Senior Consultant Cisco Systems EMEA
[email protected]
Presentation_ID
© 1999, Cisco Systems, Inc.
1
MPLS Basics
Presentation_ID
© 1999, Cisco Systems, Inc.
www.cisco.com
2
MPLS Peer Model
OSPF, BGP
OSPF, BGP
PNNI
Overlay Model
Peer Model
eg Classical IP, MPOA, NHRP
Routers and Switches totally isolated
Routers have no idea of ATM Topo
IP features must be approximately
mapped into ATM
eg MPLS
Routers and Switches totally integrated
Routers & Switches share topology
IP features directly supported by ATM
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
3
Peer vs Overlay
Overlay Model:
IP Intelligence
Around
Peer Model:
IP Intelligence
at every hop
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
4
MPLS Switching - Overview
MPLS Domain
Label Edge Router
egress LER
Label Switch Router
LSR
Label Edge Router
ingress LER
128.89
I/f 0
Label Edge Router
egress LER
I/f 1
I/f 4
Unlabeled Data
2
Labeled Data
2
Labeled Data
171.69
Unlabeled Data
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
5
MPLS Switching—Example
Local Remote Address
Lbl
Lbl
Prefix Interface
X
1
128.89
1
X
2
171.69
1
..
…
…
Local Remote Address
Lbl
Lbl
Prefix Interface
Label Information
Base
1
7
128.89
0
2
5
171.69
4
3
…
…
128.89
0
I/f 1
171.69.12.1 Data
I/f 4
2 171.69.12.1 Data
Unlabeled Data
5 171.69.12.1 Data
171.69
171.69.12.1 Data
CEF Forwarding Table Populated
with Routing Topology Information
Unlabeled Data
Each Route/Prefix Mapped to a Label Value
Switching Decision Then Only ‘Label-Swaps’ via
the Label Information Base (LIB)
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
6
MPLS Switching— FECs
FEC1
3
Dest: B
3
Dest: C
B
FEC1
1
Dest: B
1
Dest: C
2
Dest: D
2
Dest: E
intf 0
LSR Y
LSR Z
A
LSR V
FEC2
C
intf 1
LSR X
4
Dest: D
4
Dest: E
FEC2
D
E
The ingress router can use additional information
LIB LSR X
when it is assigning packets to a FEC, like
IN
OUT
INTF
•incoming port
1
3
0
•ToS bits
2
4
1
•source address
FEC … Forwarding Equivalent Class
•any arbitrary information
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
7
Generic Label Encapsulation
L2 Header
(PPP/Ethernet/...)
Generic Encapsulation/
Shim Header
L2 Header
Lbl Stack
Layer 3 Header
Label (0)
Exp S
TTL
20 Bits
3
1
Bits Bits
8 Bits
EXP … Experimental Use (used as QoS bits)
S ……. Bottom of Stack (set to 1 for last entry, o for all other label stack entries)
TTL … Time to Live
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
8
Label Stack
L2 Header
Label (0)
Exp S
Lbl Stack
TTL
Layer 3 Header
Label (1)
Exp S
TTL
...
The Label Stack consists of a sequence of
Label Stack Entries equal or greater 1
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
9
ATM Label Encapsulation
ATM Cell Header
oebb_update_062k
© 1999, Cisco Systems, Inc.
GFC
VPI
VCI
PTI CLP HEC
DATA
Lbl
Lbl
Top Label
encoded in
VPI/VCI fields
Top Label and subsequent
Labels (if present) are also
encoded with generic
encapsulation (+CoS, +TTL
fields)
www.cisco.com
10
Label Allocation
“Downstream on Demand”
Packets with
Label n
1. Label Request
Message for Label n
Upstream LSR
2. Label Mapping
Message for Label n
Downstream LSR
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
11
Label Distribution
OSPF, IS-IS, etc ...
Layer 3 Routing Protocol
LDP, RSVP, mp-BGP-4, etc ...
Label Distribution Protocol
LSR X
oebb_update_062k
© 1999, Cisco Systems, Inc.
ATM, PPP, Ethernet, PoSIP, etc
Data Link Technology
www.cisco.com
LSR Y
12
MPLS Example:
Routing Information
In
Lbl
Address
Prefix
128.89
171.69
...
Out Out
I’face Lbl
In
Lbl
In
I/F
Address
Prefix
1
1
...
128.89
171.69
...
Out Out
I’face Lbl
In
Lbl
In
I/F
0
1
...
Address
Prefix
128.89
0
...
...
0 128.89
1
1
0
2
You can reach 128.89 and
171.69 through me
Out Out
I’face Lbl
You can reach 128.89
through me
1
171.69
Routing Updates
(OSPF, IS-IS, …)
oebb_update_062k
© 1999, Cisco Systems, Inc.
You can reach 171.69
through me
www.cisco.com
13
MPLS Example:
Requesting Labels
In
Lbl
Address
Prefix
128.89
171.69
...
Out Out
I’face Lbl
In
Lbl
In
I/F
Address
Prefix
1
1
...
128.89
171.69
...
Out Out
I’face Lbl
In
Lbl
In
I/F
0
1
...
Address
Prefix
128.89
0
...
...
1
1
I need a Lbl for 128.89
I need a Lbl for 171.69
0
2
3
Out Out
I’face Lbl
0
128.89
I need a Lbl for 128.89
I need another Lbl for 128.89
1 I need a Lbl for 171.69
171.69
Label Distribution
Protocol (LDP)
I need a Lbl for 128.89
(Downstream on Demand Allocation)
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
14
MPLS Example:
Assigning Labels
In
Lbl
Address
Prefix
-
128.89
171.69
...
Out Out
I’face Lbl
1
1
...
4
5
In
Lbl
In
I/F
Address
Prefix
4
8
5
2
3
2
128.89
128.89
171.69
Out Out
I’face Lbl
In
Lbl
In
I/F
Address
Prefix
9
10
7
9
10
1
1
128.89
0
128.89
...
0
...
0
0
1
Use Lbl 4 for 128.89
Use Lbl 5 for 171.69
0
2
-
0 128.89
1
1
Out Out
I’face Lbl
Use Lbl 9 for 128.89
Use Lbl 10 for 128.89
3
1
Use Lbl 7 for 171.69
171.69
Use Lbl 8 for 128.89
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
15
MPLS Example:
Packet Forwarding
In
Lbl
Address
Prefix
-
128.89
171.69
...
Out Out
I’face Lbl
1
1
...
4
5
In
Lbl
In
I/F
Address
Prefix
4
8
5
2
3
2
128.89
128.89
171.69
Out Out
I’face Lbl
0
0
1
9
10
7
In
Lbl
In
I/F
Address
Prefix
9
10
1
1
128.89
0
128.89
...
0
...
1
1
Out Out
I’face Lbl
0
-
128.89
0
2
128.89.25.4 Data
9 128.89.25.4 Data
128.89.25.4 Data
4 128.89.25.4 Data
1
171.69
Each label defines
a different LVC
oebb_update_062k
© 1999, Cisco Systems, Inc.
LSR forwards based
on label
www.cisco.com
16
MPLS on ATM
In
I/F
In
Lbl
Address
Prefix
Out
I/F
1
Labels act as the VC
identifier for ATM switches 2
(Label VC or LVC)
...
5
8
...
128.89
128.89
...
0
0
...
Out
Lbl
3
7
Labels change between
...
switches - LVCs are
not end-to-end.
Cells
5
5
5
5
1
0
Packet
128.89
3
ATM Cell Header
MPLS “partition”
allocated for each link GFC
VPI
VCI
(no per-VC
bandwidth reservation).
3
3
PTI
3
CLP
HEC
DATA
Label
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
17
VC Merge
Cells
5
Packet
Packet
8
5
8
5
8
In
I/F
In
Lbl
Address
Prefix
Out
I/F
Out
Lbl
1
2
...
5
8
...
128.89
128.89
...
0
0
...
3
3
...
1
0
2
3
5
8
128.89
3
3
3
3 3
• With a ATM switch supporting VC-Merge:
Can reuse outgoing Label
Hardware prevents cell interleave
Fewer Labels required , For very large networks
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
18
MPLS-VPN
What is a VPN ?
• An IP network infrastructure delivering private
network services over a public infrastructure
Use a layer 3 backbone
Scalability, easy provisioning
Global as well as non-unique private address
space
QoS
Controlled access
Easy configuration for customers
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
19
MPLS Applications
Presentation_ID
© 1999, Cisco Systems, Inc.
www.cisco.com
20
MPLS Traffic Engineering
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
21
Traffic Engineering: Motivations
• Reduce the overall cost of operations by
more efficient use of bandwidth resources
by preventing a situation where some parts of
a service provider network are over-utilized
(congested)
while other parts under-utilized
The ultimate goal is cost saving
and maximized performance!
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
22
Traffic Engineering’s Job
• Construct routes for traffic streams
within a service provider network
to avoid causing some parts of the
provider’s network to be over-utilized
while others parts remain under-utilized
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
23
Traffic Engineering With
Overlay
R2
R3
R1
PVC for R2 to R3 traffic
PVC for R1 to R3 traffic
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
24
MPLS Traffic Engineering
R8
R3
R4
R2
R5
R1
R6
R7
MPLS LSP for R8 to R5 traffic
MPLS LSP for R1 to R5 traffic
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
25
TE Example Deployment
Find route & set-up tunnel for 20 Mb/s from POP1 to POP4
Find route & set-up tunnel for 10 Mb/s from POP2 to POP4
WAN area
POP4
POP1
POP
POP2
POP
POP
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
26
MPLS TE Components (1)
• Link Attribute Flooding
Link state IGP protocols enhanced to advertise Link
Resource Attributes
• Constraint based Routing
SPF computation enhanced to compute path which
satisfies the resource Constraints (bandwidth, policy) for
a TE tunnel
• TE Tunnel establishment
RSVP signaling extended (eg label binding) to set-up the
LSP along the route computed by Constraint Base
Routing
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
27
MPLS TE Components (2)
• MPLS Forwarding
LFIB handles the forwarding “as usual”
only - LFIB has been populated by another Control
module than Destination Based LDP
• Routing Traffic over TE Tunnels
IGP enhanced on tunnel Head-ends to “route” IP
packets “into” TE tunnels
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
28
Constrained Based Routing
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
29
Path Computation
Input:
– constraints imposed by TE tunnel to be
routed
– resource attributes of every link (bandwidth,
Resource Class affinity, metric)
available from IS-IS or OSPF
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
30
Path Computation
• Prune links if:
insufficient resources (e.g., bandwidth)
violates policy constraints
• Compute shortest distance path
R3 uses its own metric
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
31
LSP Tunnel Setup
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
32
TE Tunnel Setup
• Initiated at the head-end of a trunk
• Uses Explicit Route calculated by Constraint
Based Routing or configured manually by
operator
• Uses RSVP (with few extensions) to establish
Label Switched Paths (LSPs) for TE tunnel
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
33
Fast Restoration
Handling link failures - two
complementary mechanisms:
• Path protection
• Link protection
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
34
Link Protection for R2-R4 Link
R9
R8
R4
R2
R5
R1
R6
R7
Setup: Path (R2->R6->R7->R4)
Labels Established on Resv message
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
35
TE Tunnel Prior to Link Failure
R9
R8
R4
R2
R5
R1
R7
R6
Setup: Path (R1->R2->R4->R9)
Labels Established on Resv message
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
36
Link Protection Active
R9
R8
R4
R2
R5
R1
R7
R6
On failure of link from R2 -> R4, R2 simply changes outgoing
Label Stack from <Label1> to <Label2, Label1>
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
37
MPLS VPN QoS
And
Traffic Engineering
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
38
MPLS VPN QoS and
Traffic Engineering
•MPLS VPN service unchanged:
MPLS VPN QoS SLA exactly as defined earlier
•Traffic Engineering in core to reduce cost
MPLS TE
WAN area
POP4
POP1
POP
POP2
POP
POP
Question: How many MPLS labels ???
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
39
MPLS VPN QoS and
Traffic Engineering
iBGP
LDP
LDP
WAN area
LDP
POP4
POP1
RSVP
POP
POP2
POP
POP
User IP Packet
Answer: 3 labels
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
40
Carrying Service Class Information:
Packet Media
IPv4 Header
Layer 2 Header
IPv4 Header
Payload
Type of Service field (old definition)
Diffserv field (expanded definition)
IPv6 Header
Layer 2 Header
IPv6 Header
Payload
Diffserv field (supercedes the Traffic
Class octet)
Packet-based MPLS
Layer 2 Header
MPLS Header
L3 Header & Payload
Different labels to each destination
for different Classes
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
41
Carrying Service Class Information:
ATM
ATM Cell Header
GFC
VPI
VCI
PTI
CLP
HEC
DATA
Label
Different LVCs to each destination for
different Classes.
• LVCs have DiffServ service types, not ATM Forum CBR,
UBR, VBR or ABR
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
42
Carrying Service Class Information:
ATM
PVC/SVC Traffic
ATMF Queues
?
IP Traffic
Traditional ATM Switch:
No IP Awareness
PVC/SVC Traffic
IP Traffic
ATMF Queues
IP Queues
PVC/SVC Traffic
IP Traffic
MPLS+DiffServ model: Separate
DiffServ Queues & Policies on the ATM
switch
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
43
Differentiated Service on a Link:
Two Classes
Bandwidth
Spare
Estimated premium
traffic
Premium Traffic
But premium traffic is
guaranteed access to
most of the bandwidth,
if it needs it.
Best Effort Traffic
Best effort: little guaranteed
Time
• Premium traffic can have extra bandwidth allocated to it,
which it will use only if needed.
• Premium traffic gets excellent QoS, as if it has bandwidth
over-engineered for it
• ‘Best Effort’ traffic gets access to bandwidth unused by
premium traffic: little or no wasted bandwidth.
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
44
MPLS VPN QoS Model
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
45
How It Feels for a CPE:
Routing Viewpoint
Layer 2 VPN
Layer 2 VPN : Physical View
Layer 2 VPN : Logical View
MPLS VPN
MPLS VPN : Physical View
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
MPLS VPN : Logical View
46
How It Feels for a CPE:
Routing Viewpoint
• Routing Adjacencies:
Before MPLS VPN:
point-to-point to all remote sites
With MPLS VPN:
point-to-cloud
• “Point-to-Cloud” is key to MPLS VPN
benefits from Routing Viewpoint
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
47
How It Feels for a CPE:
QoS Viewpoint
Layer 2 VPN
Layer 2 VPN : Logical View
Layer 2 VPN : Physical View
MPLS VPN
MPLS VPN : Physical View
oebb_update_062k
© 1999, Cisco Systems, Inc.
MPLS VPN : Logical View
www.cisco.com
48
How It Feels for a CPE:
QoS Viewpoint
• QoS Commitment:
Before MPLS VPN
point-to-point to all remote sites
With MPLS VPN:
point-to-cloud
this is exactly the Diff-Serv model
• “Point-to-Cloud” is key to MPLS VPN benefits from
QoS Viewpoint
scalability in SP Backbone
simplicity for Customer
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
49
Benefits of the
“Point-to-Cloud” Model
• Any to any connectivity ...
• … without requiring the customer to know or
specify its traffic matrix
Changes in traffic matrix accommodated by SP
without changes in the QoS contract
• Preserves MPLS VPN scalability
no “per- VPN-Site” awareness in SP backbone
• Resource Allocation by SP is at very aggregate level
per COS
easier, higher statistical gain
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
50
How to Build
“Point-to-Cloud” Service?
• Scenario 1:
– Constrained access
– Unconstrained Backbone
Best-Effort o IP
Diff-Serv o IP
Diff-Serv o IP
MPLS VPN
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
51
How to Build
“Point-to-Cloud” Service?
• Scenario 2:
– Constrained access
– Constrained Backbone (or requirement for tightest
possible delay) Diff-Serv o MPLS
Diff-Serv o IP
Diff-Serv o IP
MPLS VPN
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
52
How to Build
“Point-to-Cloud” Service?
• Scenario 3:
MPLS VPN QoS does not “require”, but can
benefit from, MPLS Traffic Engineering
– Constrained access
– Constrained Backbone (or
requirement for tightest possible
delay)
Does not change the “Point-to-Cloud” model
Opportunity to reduce cost
Opportunity to improve QoS target (eg.
delay)
– Requirement to maximise use of
backbone resources
Diff-Serv o IP
Diff-Serv o MPLS
Traffic Engineering o MPLS
Diff-Serv o IP
MPLS VPN
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
53
MPLS VPN QoS - Conclusions
• Key MPLS VPN QoS Service is “point-to-cloud”
• MPLS QoS number one goal is to support Diff-Serv, the
whole of Diff-Serv and nothing but Diff-Serv
• For Service Provider, MPLS Diff-Serv deployment is
virtually the same as IP Diff-Serv deployment
activate Diff-Serv queuing/dropping
perform Diff-Serv capacity planning
on ATM PVCs Model is IP QoS and not Layer 2 QoS
no per-VPN QoS
rather, per Class QoS
each VPN can use multiple Classes
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
54
DiffServ over MPLS
Standardization Update
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
55
IETF Progress
• draft-ietf-mpls-diff-ext-03.txt
• Working Group document
• (optimistic) goal: Last Call at April
Adelaide meeting
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
56
Diff-Serv over MPLS:
“Colouring” MPLS Frames
• Two methods are possible
– Single LSP per FEC
•use EXP field in MPLS header to select Diff-Serv
queue
–E-LSP
– Multiple LSPs per FEC
•use label to select Diff-Serv queue
–L-LSP
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
57
Yet More Terminology
• E-LSP
behavior (queue & drop) inferred from EXP bits
only
Allows up to 8 BAs on an LSP
• L-LSP
behavior inferred from Label (and perhaps EXP
bits too)
for AFxy, label determines the queue, EXP bits
determine drop preference
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
58
E-LSPs and L-LSPs
• MPLS over PPP and LAN:
both E-LSPs and L-LSPs allowed
• MPLS over ATM/FR:
only L-LSPs possible (EXP is not seen)
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
59
Using the EXP Bits: E-LSP
0
1
2
3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|
Label
| EXP |S|
TTL
|
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
• Mapping of IP DSCP into MPLS EXP
MPLS
Diff-Serv Domain
Non-MPLS
Diff-Serv Domain
IPv4 Packet
MPLS
DSCP= xxxxxx
oebb_update_062k
© 1999, Cisco Systems, Inc.
Hdr
MPLS
EXP= yyy
www.cisco.com
DSCP= xxxxxx
60
Using the EXP bits: E-LSP
LSR
LDP
LDP
E-LSP
• LDP or RSVP establishes one
E-LSP per FEC
• Queue is selected based on EXP
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
61
Using Multiple LSPs: L-LSPs
LDP
LSR
LDP
L-LSPs
• LDP or RSVP establishes multiple
L-LSPs per FEC
• Queue is selected based on label
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
62
MPLS COS Phase 2
COS Translation
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
63
COS Translation for
DiffServ IP Transport
IP with Full Diff-Serv
6-bit DS
IP with Full Diff-Serv
6-bit DS
MPLS VPN
• Allows operations of Diff-Serv IP over MPLS backbone (VPN or non-VPN)
• only max 8 COS supported by the MPLS cloud
--> if more than 8 COS (BAs) supported in IP clouds they have to be
mapped onto the MPLS backbone 8 COS
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
64
COS Translation
• Developed as flexible translation:
– COS={Prec, DS, EXP, CLP}
– COS translation =
Translation from any* to any
* except from CLP
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
65
MPLS Guaranteed
Bandwidth
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
66
MPLS as the MultiService Infrastructure:
Layer Collapsing
Applications
IP
Hard Pt-2-Pt QoS
Soft Pt-2-Cloud QoS
ATM
SDH
MPLS
IP
MPLS
Admission Control
Traffic Engineering
WDM
Fast Restoration
WDM
Transport
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
67
MPLS as the MultiService Infrastructure:
Layer Collapsing
Applications
IP
Hard Pt-2-Pt QoS
Soft Pt-2-Cloud QoS
ATM
SDH
+ MPLS
Guaranteed
Bandwidth
IP
MPLS
Admission Control
Traffic Engineering
WDM
Fast Restoration
WDM
Transport
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
68
MPLS Guaranteed Bandwidth:
The Service
• Provisioned Diff-Serv COS is fine for many endcustomer application’s requirements
• Special services (voice, bandwidth trading, Carrier’s
Carrier…) need guarantees and tighter QoS
• Massive over-provisioning cannot always be assumed
everywhere in network
• MPLS Guaranteed Bandwidth:
offers Layer-2-like point-to-point QoS commitments
while preserving MPLS/IP scalability
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
69
MPLS Guaranteed Bandwidth:
The Service
• MPLS Guaranteed Bandwidth Service
unidirectional Point-to-point Bandwidth with
commitment on QoS parameters
N2 Mb/s
Guarantee
CE
10.2
11.5
N1 Mb/s Guarantee
CE
CE
11.6
oebb_update_062k
CE 10.1
© 1999, Cisco Systems, Inc.
www.cisco.com
70
MPLS Guaranteed Bandwidth:
The Mechanisms
MPLS Guaranteed Bandwidth =
Diff-Serv
Traffic Conditioning on Edge +
Queues/PHBs in Core +
MPLS TE with COS awareness
COS-aware Routing +
COS-aware Admission Control
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
71
MPLS Guaranteed Bandwidth:
The Mechanisms
Diff-Serv
MPLS
Diff-Serv Traffic Conditioning:
(on a per e2e service basis)
- Classification
- Metering
- Marking
- Policing
MPLS Traffic Engineering for GB:
(aggregated: one GB Tunnel for
multiple services)
- 150 Mb/s from P_in to P_out
- COS aware Routing
- COS aware Admission Control
P_in
50 Mb/s
P_out
100 Mb/s
Diff-Serv PHB:
Diff-Serv (even more aggregated: one Diff-Serv queue)
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
72
MPLS Guaranteed Bandwidth:
The Mechanisms
More on MPLS Traffic Engineering for GB:
50 Mb/s
P_in
P_out
100 Mb/s
IGP advertises non-reserved
bandwidth on every link
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
73
MPLS Guaranteed Bandwidth:
The Mechanisms
More on MPLS Traffic Engineering for GB:
P_in performs Constraint Based Routing:
finds a Path with sufficient non-reserved bandwidth for GB
50 Mb/s
P_in
P_out
100 Mb/s
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
74
MPLS Guaranteed Bandwidth:
The Mechanisms
More on MPLS Traffic Engineering for GB:
P_in sends MPLS signalling for establishment
of GB Tunnel along computed path
50 Mb/s
P_in
P_out
100 Mb/s
admission control performed on every link
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
75
MPLS Guaranteed Bandwidth
for Voice
PSTN
GW
GW
Call
Agent
PSTN
GW
GB Tunnel
oebb_update_062k
© 1999, Cisco Systems, Inc.
ensures that Voice Load is below
configured X% on EVERY link
(--> Guaranteed QoS)
www.cisco.com
76
MPLS Guaranteed Bandwidth
for Voice
PSTN
GW
GW
Call
Agent
PSTN
GW
GB Tunnel
oebb_update_062k
© 1999, Cisco Systems, Inc.
explicit rejection of new Tunnels if
there is no path that can meet QoS
(--> explicit knowledge that extra
resources required)
www.cisco.com
77
MPLS Guaranteed Bandwidth
for Voice
PSTN
GW
GW
Call
Agent
PSTN
GW
GB Tunnel
oebb_update_062k
© 1999, Cisco Systems, Inc.
Voice Traffic distributed over
alternate path if required:
“Traffic Engineering” of Voice
www.cisco.com
78
MPLS Guaranteed Bandwidth
for Voice
PSTN
GW
GW
Call
Agent
PSTN
GW
GB Tunnel
oebb_update_062k
© 1999, Cisco Systems, Inc.
MPLS Fast Reroute:
Voice calls not affected by failure
www.cisco.com
79
MPLS VPNs
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
80
Managed IP Services Scale to
Large and Small Customers
Separately engineered
customer private IP
networks
Vs.
Single carrier network
supporting multiple
customer IP VPNs
BGP/MPLS
VPN Network
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
81
MPLS—Foundation for L3 VPNs
• VPNs uniquely defined
via Label + VPN ID
decoupling forwarding
from IP addressing
• Data privacy via
logically separated
label switched paths
• Quaility-of-Service
(Label CoS)
• Provides IP address
uniqueness
Enterprise B
Enterprise A
Internet Backbone—
“VPN 0”
Intranet
VPN 10
Extranet
VPN 20
Enterprise B
Enterprise A
• Eliminates tunnel
mesh
oebb_update_062k
© 1999, Cisco Systems, Inc.
Enterprise C
www.cisco.com
82
VPN-Aware Network
Routing Architecture
iBGP
1. SP network uses an
IGP to exchange
local reachability
2. CEs (customer
edge) and PEs
(provider edge)
exchange routing
info (IP)
3. PEs exchange VPN
routing info and tag
bindings (VPN-IP)
IGP (e.g.
via mBGP (RFC2283) OSPF)/TDP
4. LDP is used to
bind tags to routes
in the core
oebb_update_062k
© 1999, Cisco Systems, Inc.
www.cisco.com
PE
eBGP/
Static/RIP
CE
83
MPLS VPN—Network
Formation
Cust A
10.1.1
VPN 15
Controlled Route
Distribution via
Selective Advertisement
(15)10.1.1
(15)10.2.1
Internet
Scale
VPN
Private View
Cust A
10.2.1
VPN 15
(354)128.24.1
(15)10.3.1
Public View
(354)128.24.2
Cust A
10.3.1
VPN 15
Private View
Forwarding Examples
Cust B
128.24.1
VPN 354
oebb_update_062k
© 1999, Cisco Systems, Inc.
IN
OUT
(15)10.2.1
(15)10.1.1
(15)10.3.1
(354)128.24.2
(354)128.24.1
www.cisco.com
Cust B
128.24.2
VPN 354
84
Presentation_ID
© 1999, Cisco Systems, Inc.
www.cisco.com
85