04 MPLS Te PDF

GTER 19
São Paulo – 3-5 Julho, 2005
Ativando MPLS Traffic Engineering
Alexandre Longo – alongo@cisco.com

Cisco Systems
GTER 19 © 2005 Cisco Systems, Inc. All rights reserved. 1

Some Assumptions
• You understand basic IP routing

• You understand MPLS concepts and operation
• You understand how a link-state protocol works
• Some knowledge of QoS is useful
• You will still be awake at the end of this
GTER © 2005 Cisco Systems, Inc. All rights reserved. 2

A Blatant Plug
• Traffic Engineering
with MPLS
ISBN: 1-58705-031-5
• Now available in
Portuguese and
Chinese!

Agenda
• Traffic Engineering Overview

• Traffic Engineering Theory
• Configuration (*)
• Protection
• Design and Scalability (*)
• Summary

TRAFFIC ENGINEERING
OVERVIEW

Network vs. Traffic Engineering
• Network engineering
Build your network to carry your predicted traffic
• Traffic engineering
Manipulate your traffic to fit your network
• Traffic patterns are impossible to accurately predict

• Symmetric bandwidths/topologies, asymmetric load
• TE can be done with IGP costs, ATM/FR, or MPLS

Motivation for Traffic Engineering
• Increase efficiency of bandwidth resources

Prevent over-utilized (congested) links whilst other links
are under-utilized
• Ensure the most desirable/appropriate path for
some/all traffic
Override the shortest path selected by the IGP
• Replace ATM/FR cores
PVC-like traffic placement without IGP full mesh and
associated O(N^2) flooding
• The ultimate goal is COST SAVING
Service development also progressing

The “Fish” Problem (Shortest Path)
R3
R8
R4
R5
R2
R1
R6 R7
• IP uses shortest path destination-based routing

• Shortest path may not be the only path
• Alternate paths may be under-utilized
• Whilst the shortest path is over-utilized
Shortest Path and Congestion
20Mbps 60Mbps 26Mbps
traffic to R5 aggregate R3 drops!
R8
R4
OC3
E3 (155Mbps)
OC3 (34Mbps) R5
R2 (155Mbps)
GigE
R1 (1Gbps)
GigE
(1Gbps)
R6 GigE R7
(1Gbps)
40Mbps
traffic to R5

The TE Solution
20Mbps 20Mbps traffic
traffic to R5 R3 to R5 from R8
R8
R4
R5
R2
40Mbps traffic
to R1 from R8
R1
R6 R7
40Mbps
traffic to R5
• MPLS Labels can be used to engineer explicit paths
• Tunnels are UNI-DIRECTIONAL
Normal path: R8 Î R2 Î R3 Î R4 Î R5
Tunnel path: R1 Î R2 Î R6 Î R7 Î R4
Terminology
• Constrained-Based Shortest Path First (CSPF)

MPLS-TE uses CSPF to create a shortest path based
on a series of constraints:
Bandwidth
Affinity/link attributes
…or an explicitly configured path
• Tunnels are UNI-DIRECTIONAL!
HEADEND MIDPOINT TAILEND
Upstream Tunnel Direction Downstream

TRAFFIC ENGINEERING
THEORY

Traffic Engineering Components
• Information distribution
• Path selection/calculation
• Path setup
• Trunk admission control
• Forwarding traffic on to tunnel
• Path maintenance

Information Distribution
• Need to flood TE information (Resource Attributes)

across the network
Available bandwidth per priority level, a few other things
• IGP extensions flood this information

OSPF uses Type 10 (area-local) Opaque LSAs
ISIS uses new TLVs
• Basic IGP: {self, neighbors, cost to neighbors}

• TE extensions: {self, neighbors, cost to neighbors,
available bandwidth to neighbors}
• TE bandwidth is a control-plane number only

Path Calculation
• Once available bandwidth information and attributes are

flooded, router may calculate a path from head to tail
Path may be explicitly configured by operator
• TE Headend does a “Constrained SPF” (CSPF)

calculation to find the best path
• CSPF is just like regular IGP SPF, except
Takes required bandwidth and attributes into account
• Looks for best path from a head to a single tail

(unlike OSPF)
• Minimal impact on CPU utilization using CSPF
• Path can also be explicitly configured

Path Setup
• Once the path is calculated, it must be signaled

across the network
Reserve any bandwidth to avoid “double booking” from
other TE reservations
Priority can be used to pre-empt low priority existing
tunnels
• RSVP used to set up TE LSP
PATH messages (from head to tail) carries
LABEL_REQUEST
RESV messages (from tail to head) carries LABEL
• When RESV reaches headend, tunnel interface = UP
• RSVP messages exist for LSP teardown and
error sig
Trunk Admission Control
• On receipt of PATH message

Router will check there is bandwidth available to honour
the reservation
If bandwidth available then RSVP accepted
• On receipt of a RESV message

Router actually reserves the bandwidth for the TE LSP
If pre-emption is required lower priority LSP are torn down
• OSPF/ISIS updates are triggered

Path Setup Example
R8 R9
R3
R4 10
30
PATH
Message R2 RESV
Pop message
20Mbps
R5
R1 60
80 50
70 32
40
60
49 27 R6 R7
80
100
22
RSVP PATH: R1 Î R2 Î R6 Î R7 Î R4 Î R9
RSVP RESV: Returns labels and reserves
bandwidth on each link
80 Bandwidth available
49 Returned label via RESV message
Forwarding Traffic to a Tunnel
• Static routing
• Policy routing
Global table only—not from VRF at present
• Autoroute
• Forwarding Adjacency
Static, autoroute, and forwarding adjacency get you

unequal-cost load-balancing

Autoroute
• Used to include TE LSP in SPF calculations

• IGP adjacency is NOT run over the tunnel!
• Tunnel is treated as a directly connected link
to the tail
When tunnel tail is seen in PATH list during IGP SPF,
replace outgoing physical interface with tunnel interface
Inherit tunnel to all downstream neighbors of said tail

Autoroute Topology (OSPF and ISIS)
R8
R3
R4
R2
R1
R5
R6 R7
Tunnel1: R1 ÎR2 Î R3 Î R4 Î R5
Tunnel2: R1 ÎR6 Î R7 Î R4

Autoroute Topology (OSPF and ISIS)
R8
R3
R4
R2
R1
R5
R6 R7
From R1 router perspective:

Next hop to R4 and R8 is Tunnel1
Next hop to R5 is Tunnel2
All nodes behind tunnel routed via tunnel

Forwarding Adjacency
• Autoroute does not advertise the LSP into the IGP

• There may be a requirement to advertise the
existence of TE tunnels to upstream routers
Like an ATM/FR PVC—attract traffic to a router regardless
of the cost of the underlying physical network cost
• Useful as a drop-in replacement for ATM/FR

(and during migration)
• Can get suboptimal forwarding (NOT loops) if
you’re not careful

Forwarding Adjacency
R3 R4 All links use
cost of 10
R2 R5
R1
R9
R8
R6
IGP Cost = 40
R7
Tunnel: R2 ÎR3 Î R7 Î R4 Î R5
R1 shortest path to R9 via IGP
Tunnel at R2 is never used as R1 can’t see it

Advertise TE Links into IGP
R3 R4 FA IGP Cost = 10
R2 R5
R1
R9
R8
R6
IGP Cost = 30
R7
Tunnel: R2 ÎR3 Î R4 Î R5
R1 shortest path to R9

Load Balancing Across FA
IGP Cost = 30
R2 R5
R1
R9
R8
R6
R7
Tunnel: R2 ÎR3 Î R4 Î R5
R1 shortest path to R9

Unequal Cost Load Balancing
• IP routing has equal-cost load balancing, but not

unequal cost*
• Unequal cost load balancing difficult to do while
guaranteeing a loop-free topology
• Since MPLS doesn’t forward based on IP header,
permanent routing loops don’t happen
• 16 hash buckets for next-hop, shared in rough
proportion to configured tunnel bandwidth or load-
share value
*EIGRP Has ‘Variance’, but That’s Not as Flexible

Unequal Cost: Example 1
Router F
Router A 40MB
Router E
Router G
20MB
gsr1#show ip route 192.168.1.8

Routing entry for 192.168.1.8/32
Known via "isis", distance 115, metric 83, type level-2
Redistributing via isis
Last update from 192.168.1.8 on Tunnel0, 00:00:21 ago
Routing Descriptor Blocks:
* 192.168.1.8, from 192.168.1.8, via Tunnel0
Route metric is 83, traffic share count is 2
192.168.1.8, from 192.168.1.8, via Tunnel1
Route metric is 83, traffic share count is 1

Unequal Cost: Example 1
Router F
Router A 40MB
Router E
Router G
20MB
gsr1#sh ip cef 192.168.1.8 internal

………
Load distribution: 0 1 0 1 0 1 0 1 0 1 0 0 0 0 0 0 (refcount 1)
Hash OK Interface Address Packets Tags imposed
1 Y Tunnel0 point2point 0 {23}
2 Y Tunnel1 point2point 0 {34}
………
Note That the Load Distribution

Is 11:5—Very Close to 2:1, but Not Quite!

Path Maintenance
• Steady-state information load is low

Especially with refresh reduction (RFC2961)
• Path re-optimization
Process where some traffic trunks are rerouted to new
paths so as to improve the overall efficiency in bandwidth
utilization
For example, traffic may be moved to secondary path
during failure; when primary path is restored traffic
moved back
• Path restoration
Comprised of two techniques; local protection
(link and node) and path protection
Discussed later in protection section

CONFIGURATION
RST-2603
9866_05_2004_c2 © 2004 Cisco Systems, Inc. All rights reserved. 31
Prerequisite Configuration (Global)
ip cef [distributed]
mpls traffic-eng tunnels

• OSPF
mpls traffic-eng router-id loopback0
mpls traffic-eng area ospf-area
• ISIS
mpls traffic-eng router-id loopback0
mpls traffic-eng level-x
metric-style wide

• On each physical interface

interface pos0/0
ip rsvp bandwidth Kbps (Optional)
mpls traffic-eng attribute-flags attributes (Opt)

Build a Tunnel Interface (Headend)
interface Tunnel0
ip unnumbered loopback0
tunnel destination RID-of-tail
tunnel mode mpls traffic-eng
tunnel mpls traffic-eng bandwidth 10

Tunnel Attributes
interface Tunnel0
tunnel mpls traffic-eng bandwidth Kbps
tunnel mpls traffic-eng priority pri [hold-pri]
tunnel mpls traffic-eng affinity properties [mask]
tunnel mpls traffic-eng autoroute announce

Path Calculation
• Dynamic path calculation
int Tunnel0
tunnel mpls traffic-eng path-option # dynamic
• Explicit path calculation

int Tunnel0
tunnel mpls traffic path-opt # explicit name foo
ip explicit-path name foo
next-address 1.2.3.4 [loose]
next-address 1.2.3.8 [loose]

Multiple Path Calculations
• A tunnel interface can have several path options,

to be tried successively
tunnel mpls traffic-eng path-option 10 explicit name foo
tunnel mpls traffic-eng path-option 20 explicit name bar
tunnel mpls traffic-eng path-option 30 dynamic
• Path-options can each have their own bandwidth

tunnel mpls traffic-eng path-option 10 explicit name foo
bandwidth 100
tunnel mpls traffic-eng path-option 20 explicit name bar
bandwidth 50
bandwidth 0
LSP Attributes
Configure on Tunnel: • Attribute list options

tunnel mpls traffic-eng path- affinity
option 10 dynamic attributes
foo auto-bw
bandwidth
lockdown
priority
Attribute list ‘foo’ is defined at: protection
mpls traffic-eng lsp record-route

attributes foo
bandwidth 25
priority 2 2

Static and Policy Routing Down a Tunnel
• Static routing
ip route prefix mask Tunnel0
• Policy routing (Global Table)

access-list 101 permit tcp any any eq www
interface Serial0
ip policy route-map foo
route-map foo
match ip address 101
set interface Tunnel0

Autoroute and Forwarding Adjacency
interface Tunnel0
OR
tunnel mpls traffic-eng forwarding-adjacency
isis metric x level-y (ISIS)
ip ospf cost ospf-cost (OSPF)

Summary Configuration (1/2)
ip cef (distributed}
interface Tunnel0
tunnel mode mpls traffic-eng
ip unnumbered Loopback0
tunnel destination RID-of-tail

Summary Configuration (2/2)
! Configure in IGP
mpls traffic-eng router-id Loopback0
mpls traffic-eng area ospf-area (OSPF)
mpls traffic-eng level-x (ISIS)
metric-style wide
!
! On Physical interface
interface POS0/0
ip rsvp bandwidth Kbps

PROTECTION

Protection
• Mechanism to minimize packet loss during a failure

• Pre-provisioned protection tunnels that carry traffic
when a protected link or node goes down
• MPLS TE protection also known as FAST REROUTE
(FRR)
• FRR protects against LINK FAILURE
For example, Fibre cut, Carrier Loss, ADM failure
• FRR protects against NODE FAILURE

For example, power failure, hardware crash, maintenance
• Fast failure recovery due to local repair

Link Protection
• TE Tunnel A Î B Î D Î E
Router A Router B Router D Router E
Router C

Link Protection
• B has a pre-provisioned backup tunnel to the other end of the

protected link (Router D) B Î C Î D
• FRR relies on the fact that D is using global
label space
Protected Link
Fast ReRoute
Backup Tunnel
Router C

Link Protection
• When B Î D link fails, A Î E tunnel is encapsulated

in B Î D tunnel
• Backup tunnel is used until A can re-compute tunnel path as
A Î B Î C Î D Î E (~5-15 seconds or so)
A B D E
Original
Tunnel Router C
Fast ReRoute
GTER © 2005 Cisco Systems, Inc. All rights reserved. Backup Tunnel 48
Link Protection Example
Primary Path Pop

R8 R9
R2 14 R3 Tail End for
primary path
37
Protected Link
Fast Reroute path Pop
R1 17 R5
Head End for R6 R7

primary path 22
Primary path: R1 Î R2 Î R3 Î R9
Fast Reroute path: R2 Î R6 Î R7 Î R3

Normal TE Operation
Pop 14
R8 Swap 37 with 14
R9
R3
R2 R3
Push 37
R1 R5
R6 R7
IP 14
37

Fast Reroute Link Failure
Pop 14
R8 Swap 37 with 14
R9
R2 R3
Push 37
R1 R5
Push 17 Pop 22
R6 R7
Swap 17 with 22
IP 14 17
37 22

Link Protection Configuration
interface Tunnel0
tunnel destination Router D
… explicit-path R2-R3-R4
no tunnel mpls traffic-eng autoroute announce
interface POS0/0
mpls traffic-eng backup-path Tunnel0
interface Tunnel0 Router C

tunnel destination Router E
.. etc ...
tunnel mpls traffic-eng fast-reroute
Node Protection
• What if Router D failed?

• Link protection would not help as the backup tunnel
terminates on Router D (which is the NHop of the
protected link)
Protected Link
NHop
Fast ReRoute
Backup Tunnel
Router C

Node Protection
• SOLUTION: NODE PROTECTION (If network topology allows)

• Protect tunnel to the next hop PAST the protected link
(NNhop)
Protected Node
NNHop
Router C Fast ReRoute

Backup Tunnel

Node Protection
• Node protection still has the same convergence

properties as link protection
• Deciding where to place your backup tunnels is a
much harder problem to solve on a large-scale
• For small-scale protection, link may be better
• Configuration is identical to link protection,
except where you terminate the backup tunnel
(NNHop vs. NHop)

Link and Node Protection Times
• Link and Node protection are very similar

• Protection times are commonly linear to number of
protected items
• One nationwide provider gets ~35ms of loss
• New code on GSR E3 linecards gets a prefix-
independent 2ms-4ms loss

DESIGN AND SCALABILITY

Design Approach and Scalability
Two Methods to Deploy MPLS-TE
• Tactical
As needed to clear up congestion
You only have tunnels when there is a problem (and you
must remember to remove them
• Strategic
Mesh of TE tunnels between a level of routers
Typically P to P but can be PE to PE in smaller networks
N(N-1) LSPs (one in each direction)

Design Spectrum
Tactical Strategic TE
IGP
TE Online Offline
Control
Less Greater
Complexity

Tactical: Large ISP Case Study
700Mpbs of data
100Mbps dropped!
Router A
Router B Router C
• All links are OC12 (622Mbps)

• A has consistent ±700Mbps to
send to C
• ~100Mbps constantly dropped!
Router D Router E

Multiple TE and Unequal Cost
Load Balancing
175Mbps on this 525Mbps on this

tunnel tunnel
Router A
Router B Router C
• Tunnels with bandwidth in 3:1
(12:4) ratio
• 25% of traffic sent the long way
• 75% sent the short way
• No out-of-order packet issues—
CEF’s normal per-flow hashing
is used!
Router D Router E

Tactical
• As needed—Easy, quick, but hard to track over time

• Easy to forget why a tunnel is in place
• Inter-node BW requirements may change, tunnels
may be working around issues that no longer exist

Strategic
• Full mesh of TE tunnels between routers

• Initially deploy tunnels with 0 bandwidth
• Monitor tunnel interface statistics
~Bandwidth used between router pairs
TE tunnels have interface MIBs
Make sure that Σtunnel <= Σnetwork BW
• As tunnel bandwidth is changed, tunnels will find

the best path across the network

Strategic: Physical Topology
Router A
Router B Router C
Router D Router E

Strategic: Logical Topology
• Total of 20 tunnels in this network

• Each link is actually 2 unidirectional tunnels
Router A
Router B Router C
Router D Router E

Strategic
• N routers, N*(N-1) tunnels

• Routing protocols do not run over a TE tunnel
Unlike an ATM/FR full mesh!
• Tunnels are unidirectional

This is a good thing
Can have different bandwidth reservations in two different
directions

SUMMARY

MPLS TE
• Helps optimize network utilization (strategic)

• Assists in handling unexpected congestion
(tactical)
• Provides fast reroute for link and node failures
• TE is only part of a method of guaranteeing
bandwidth
It is a control plane mechanism only
Must be used with traditional QoS mechanisms

OBRIGADO !!

04 MPLS Te PDF

Uploaded by

Copyright:

Available Formats

04 MPLS Te PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

04 MPLS Te PDF

Uploaded by

Copyright:

Available Formats

GTER 19

São Paulo – 3-5 Julho, 2005

Ativando MPLS Traffic Engineering

Alexandre Longo – alongo@cisco.com

GTER 19 © 2005 Cisco Systems, Inc. All rights reserved. 1

• You understand basic IP routing

GTER © 2005 Cisco Systems, Inc. All rights reserved. 2

GTER © 2005 Cisco Systems, Inc. All rights reserved. 3

• Traffic Engineering Overview

GTER © 2005 Cisco Systems, Inc. All rights reserved. 4

GTER © 2005 Cisco Systems, Inc. All rights reserved. 5

• Traffic patterns are impossible to accurately predict

GTER © 2005 Cisco Systems, Inc. All rights reserved. 6

• Increase efficiency of bandwidth resources

GTER © 2005 Cisco Systems, Inc. All rights reserved. 7

• IP uses shortest path destination-based routing

GTER © 2005 Cisco Systems, Inc. All rights reserved. 9

• Constrained-Based Shortest Path First (CSPF)

HEADEND MIDPOINT TAILEND

Upstream Tunnel Direction Downstream

GTER © 2005 Cisco Systems, Inc. All rights reserved. 11

GTER © 2005 Cisco Systems, Inc. All rights reserved. 12

GTER © 2005 Cisco Systems, Inc. All rights reserved. 13

• Need to flood TE information (Resource Attributes)

• IGP extensions flood this information

• Basic IGP: {self, neighbors, cost to neighbors}

GTER © 2005 Cisco Systems, Inc. All rights reserved. 14

• Once available bandwidth information and attributes are

• TE Headend does a “Constrained SPF” (CSPF)

• Looks for best path from a head to a single tail

GTER © 2005 Cisco Systems, Inc. All rights reserved. 15

• Once the path is calculated, it must be signaled

• On receipt of PATH message

• On receipt of a RESV message

• OSPF/ISIS updates are triggered

GTER © 2005 Cisco Systems, Inc. All rights reserved. 17

Static, autoroute, and forwarding adjacency get you

GTER © 2005 Cisco Systems, Inc. All rights reserved. 19

• Used to include TE LSP in SPF calculations

GTER © 2005 Cisco Systems, Inc. All rights reserved. 20

GTER © 2005 Cisco Systems, Inc. All rights reserved. 21

From R1 router perspective:

GTER © 2005 Cisco Systems, Inc. All rights reserved. 22

• Autoroute does not advertise the LSP into the IGP

• Useful as a drop-in replacement for ATM/FR

GTER © 2005 Cisco Systems, Inc. All rights reserved. 23

GTER © 2005 Cisco Systems, Inc. All rights reserved. 24

GTER © 2005 Cisco Systems, Inc. All rights reserved. 25

GTER © 2005 Cisco Systems, Inc. All rights reserved. 26

• IP routing has equal-cost load balancing, but not

*EIGRP Has ‘Variance’, but That’s Not as Flexible

GTER © 2005 Cisco Systems, Inc. All rights reserved. 27

gsr1#show ip route 192.168.1.8

GTER © 2005 Cisco Systems, Inc. All rights reserved. 28

gsr1#sh ip cef 192.168.1.8 internal

Note That the Load Distribution

GTER © 2005 Cisco Systems, Inc. All rights reserved. 29

• Steady-state information load is low

GTER © 2005 Cisco Systems, Inc. All rights reserved. 30