Adam M IP MPLS For R4 Final Print

Cisco IP/MPLS for R4/5
Adam MacHale October 2005
amachale@cisco.com
Cisco2005 CiscoNDA 2005 Cisco Systems, subject to Systems, Inc. All rights reserved. Inc. All rights reserved.
Customer proof of concept labs CPOC Bedfont UK
2005 Cisco Systems, Inc. All rights reserved.
Agenda
What next for IP/MPLS backbone QoS Availability and convergence ATM enabled RNC/MGW IP enabled MGW Deployment examples
Agenda
Mobile operators proliferation of disparate networks

Voice
2G Voice 3G Voice Enterprise Retail stores Call centres Corporate access
Data
2.5G GPRS 3G GPRS Enterprise Retail stores Call Centres Corporate access Management WAP V110 Dial WLAN Billing IPSec Signalling Paging
Stage 1 migrate disparate networks to single MPLS core
Gn
Site 2 Site 4
Gi Billing Internet
Site 1 Site 5 Site 3 Site 4 Site 2 Site 5 Site 1
IP/MPLS Core
Site 6
Site 3
Site 5
Many networks on common sites with different edge devices and transmission
Single network over high capacity transmission carrying all services

6
The business case for a converged network
By reducing the number of networks

Lower transmission costs E.g. cost of 9xE1 = 1xSTM1 Use alternates such as GE/10GE Less maintenance contracts Single management solution Quickly deploy new services
Stage 2 reducing complexity and overlap
2G/3G R99
GGSN PDSN GRX
2G/3G R4/5
MGW
RNC SGSN
FR
IP
BTS
SS7
MSC
ATM
BTS
SS7
MSC MGW
RNC SGSN PDSN GGSN GRX
SONET/SDH
IP/MPLS
Dark Fibre
DWDM/Dark Fibre
What next for IP/MPLS core in mobile
Next stage driving deployment is the R4 split architecture with VoIP Operators looking to retire existing MSCs early to reduce OPEX and cap investment Availability of R4 IP deployments is accelerating Voice drives bandwidth with mobile providers looking to upgrade optical networks to provide additional bandwidth 10Gig+
9
2G voice
Ran Edge
BTS BSC MSC
Ran Core
PSTN
GMSC
BTS
BSC MSC
SDH
10
2G and R99 voice

Ran Edge
BTS BSC MGW NodeB RNC MSC
Ran Core
PSTN
GMSC
BTS
BSC MGW MSC
NodeB
RNC
ATM SDH
11
R4 with ATM MGW voice interconnect

Ran Edge
BTS BSC MSC MGW NodeB RNC
Ran Core
PSTN
GMSC
ATM
BTS BSC
MSC Server
MSC MGW NodeB RNC
ATM
ATM IP/MPLS SDH

12
R4 with IP MGW voice interconnect

Ran Edge
BTS BSC MGW NodeB RNC
Ran Core
PSTN
GMSC
IP/MPLS ATM
MSC Server
BTS
BSC MGW
NodeB
RNC
ATM ATM
IP/MPLS SDH
13
R5+
Ran Edge
Ran Core
PSTN
GMGW
IP/MPLS
MSC Server
BTS
BSC MGW
NodeB
RNC
IP/MPLS DWDM
14
Converged backbone platform positioning

Infrastructure Transport Edge Core Services Edge Services
WLAN GGSN PDSN Gn IPSec
SGSN
Iu-PS Gn Iu-CS Iu-PS Iur P
Internet
RNC
IP/MPLS
GRX
MGW
Iu-CS P SS7 ITP
3rd Party
HLR SCP
Hosting
12000 Edge PE
12000 CRS
7600 Services PE
Data centre
Managed and secure

15
Agenda
16
Deploying Tight-SLA services on an IP Backbone

Number of tools are available to enabled tight SLA services
Physical network design and topology
Capacity planning and active monitoring Diffserv: per-hop congestion management Traffic engineering: avoid aggregation on shortest path Convergence FRR Protection Tuning IGP Convergence
17
Solution Diffserv Architecture: RFC2475

Diffserv Domain
VoIP
Bus
BestEffort VoIP
Classification and conditioning (meter, marking, policing) at EDGE
Packet colour In DSCP
Bus BestEffort
Aggregate PHBs in backbone (EF, AF, DF, CSn)
18
Simple view of deploying QoS

Correctly mark all traffic entering the network Deploy a single queuing scheme based on proportional differentiation model on all links Capacity planning essential to ensuring adequate provisioning of bandwidth Strictly manage the volume of traffic in classes with a low drop tolerance
19
Marking at the edge

CPE
Frame Relay DS0s presented on CH STM1 DLCI 1004 user traffic DLCI 1005 control traffic 2.5G SGSN ATM STM1/4 IuCS up VPI 10 VCI 20 AAL2 CBR IuCS cp VPI 10 VCI 30 SSCOP/AAL5 nrt VBR IuPS up VPI 11 VCI 100 IP/AAL5 UBR Prec 1 IuPS up VPI 11 VCI 100 IP/AAL5 UBR Prec 0 3G RNC Gigabit Ethernet or 802.1Q VLAN trunk VLAN1 IP Prec 4 VLAN1 IP Prec 2 VLAN2 IP Prec 4 VRF 1 Exp 4 VRF 1 Exp 2 VRF 2 Exp 4 FRoMPLS Exp 2 FRoMPLS Exp 6
12000 PE
Cell ATMoMPLS Exp 5 Cell ATMoMPLS Exp 6 VRF 1 Exp 1 VRF 1 Exp 0
TE tunnel
LLQ
LFIB
Outbound Interface POS or GE
2.5/3G GGSN
20
QoS Mapping for 3G Data Roaming (From GSMA PRD IR.34)

3GPP QoS Information Traffic Class THP Diffse rv PHB QoS Requirement on GRX DSCP Max Delay 101110 100010 011010 010010 001010 000000 20ms 40ms 250m s 300m s 350m s 400m s Max Jitter 5ms 5ms N/A N/A N/A N/A Packe t Loss 0.5% 0.5% 0.1% 0.1% 0.1% 0.1% SDU Error Ratio 10
-6
Service Example
Conversational Streaming
N/A N/A 1
EF AF41 AF31 AF21 AF11 BE
10-6 10-8 10-8 10-8 10-8
VoIP, Video Conferencin g Audio/Video Streaming Transaction al Services Web Browsing Telnet E-mail Download
Interactive
2 3
Background
N/A
21
3GPP to Diffserv marking the edge
3GPP Class Routing/signalling Conversational AAL2 voice Streaming Gold Interactive Gold Interactive Silver Interactive Bronze Best Effort
Diffserv AF EF AF AF AF AF Default BE
DSCP 48 40 32 24 16 8 0
IP Prec 6 5 4 3 2 1 0
MPLS Exp 6 5 4 3 2 1 0
22
3GPP to Diffserv and MPLS Exp marking in the core
Class Routing/signaling Routing/signalling Conversational IP/AAL2 voice AAL2 voice Business Best Effort
Diffserv AF EF AF Default
DSCP 48 40 32 0
IP Prec 6 5 4 0
MPLS Exp 6 5 4 0
23
MQC queuing setup
Priority / AAL2 Routing/signalling
FIFO
Priority max 80%
6
WRED
90%
Business
WRED
Interface
4 0
Exhaustive Queuing
4 4 6 5
4
Tx WFQ
Best Effort
WRED
10%
% is of remaining bandwidth after priority queue
24
Measuring SLA Metrics
Availability and performance

Service Assurance Agent probes
Traffic type and volume

Netflow Traffic monitoring
25
Availability and measurement

Service Assurance Agent (SAA)
SAA Source
Network
SAA Target
Multiple Test Packets
Equal Time Spacing
SAA actively sends user defined probes into the network Used to measure delay, jitter application responsiveness HTTP,DNS etc SAA is available in all IOS images at no additional cost
26
Cisco Service Assurance Agent reporting
27
Cisco Service Assurance Agent reporting
28
Traffic type and volume

Netflow traffic monitor
Netflow can be enabled on IOS routers in order to monitor traffic flows 7 unique keys define a flow: Source IP address Destination IP address Source port Destination port Layer 3 protocol type TOS byte (IP DSCP + ECN bits) Input logical interface (ifIndex)
Exported Data
29
Netflow Report example

Cisco NetFlow Performance Reporting Protocol Distribution and Top Talkers AS Distribution and Top Talkers Performance Detail Drilldown
30
Agenda
31
Define requirements for each service

What is the service availability requirement
Often quoted 5-9s target equal to 5 minutes down time Is end application resilient
What is the convergence requirement in event of failure

For IP user data an outage of 3 seconds may be acceptable For signalling an outage of 60 seconds may be acceptable if using diverse paths (SCTP will recover) or 800ms unprotected For user voice an outage of less than 300-500ms may be required If targeting SDH/SONET protection may require sub 50ms for all services
32
Platform availability
12406 availability
availability non redundant 99.985 % or 79 minutes availability with redundant power 99.993% or 37 minutes availability with redundant power and GRP 99.996% or 21 minutes
Transmission availability
Typical figure offered by a service provider 99.94% or 5 hrs
This is a predicted MTBF (Mean Time Between Failures) that is based on the Telcordia (formerly Bellcore) Parts Count Method.
33
Network availability
Based on worst path with all 12406 routers

Network availability 99.99980% Or 1 minute downtime per year
34
Convergence
Fast convergence protocols
ISIS, OSPF, LDP, BGP, VPN Typically convergence in seconds MPLS TE FRR Link, node, path protection Target sub 50ms Covered in examples
35
Good operation and maintenance increases uptime IP Solution Center

A family of network-intelligent element management applications for managing MPLS and Metro Ethernet networks.
MPLS Virtual Private Network Management Application (ISC:MPLS) Layer 2 Virtual Private Network and Metro Ethernet Application (ISC:L2VPN, ISC:METRO) Traffic Engineering Management Application (ISC:TEM)
36
The Cisco Approach Intelligent MPLS Instrumentation and Cisco MPLS Diagnostics Expert (GA Dec 2005)
Type in Simple details e.g. Customer Edge IP addresses and press OK to start
Simple GUI telling you where problem is, what is underlying root cause and recommended action 100+ potential failure scenarios checked automatically repeatable process
37
Agenda
38
R4 with ATM MGW voice interconnect

Ran Edge
BTS BSC MGW NodeB RNC MSC Server BTS BSC MGW NodeB RNC
Ran Core
PSTN
GMSC
ATM IP/MPLS SDH

39
L2transport
L2 connections Pseudowires
ATMoMPLS Pseudo wire EoMPLS Pseudo wire FRoMPLS Pseudo wire
L2 connections
ATM
ATM
Ethernet
Ethernet
PE
IP/MPLS
FR
PE
FR
Any Transport over MPLS (AToM) Supports Ethernet (port, VLAN), ATM VC/VP, FR DLCI, PPP, HDLC Directed LDP used for VC-Label Negotiation, withdrawal and Error Notification Martini encapsulation for PDU transport Integrates with Diffserv QoS, TE and FRR
40
Deployment considerations for ATM enabled MGW over and MPLS backbone
Resilience Optimization
2004 Cisco Systems, Inc. All rights 2005 Cisco Systems, Inc. All rights reserved. reserved.
41
42
R4 with ATM enabled MGW

ATM service carries voice as AAL2 and data as AAL5
ATM VPI/VCI Active
ATM
IP/MPLS
ATM
RNC
ATM VPI/VCI Standby
MGW
Diverse VPI/VCI paths AToM single cell relay Convergence requirement 500ms for link and node failure
43
Example ATMoMPLS deployment

ATM service carries voice as AAL2 and data as AAL5
P P
PE1 ATM P
PE3
ATM PE2 PE3
RNC
MGW
P
Diverse VPI/VCI paths AToM single cell relay Convergence requirement 500ms for link and node failure
44
One hop primary tunnels link protection

One hop tunnels reduces total number of tunnels
P P
PE P
PE
PE
PE
MGW
MGW
P
Autoroute used to direct traffic down tunnels All tunnels are static
Primary one hop tunnel

45
FRR link protection of all links sub 50ms restoration
Primary
PE P
PE
PE
PE
RNC Backup
P
MGW
46
Link failure detection

POS
AIS alarm is used to trigger FRR protection, detected within a few ms SDH/SONET has end to end signalling
GE
LOS based GE triggers FRR when GE interface goes down Can be as fast as POS but should only be deployed over dark fibre or optical network with end to end signalling
Yes Yes with signalling No, BFD will help but not as fast
SDH
SDH
47
Traffic directed down full mesh of end to end tunnels

Full mesh of tunnels needs to managed carefully
P P
PE P
PE
PE
PE
RNC
MGW
P
Data path
48
Node protection for P node failure sub 50ms restoration
Primary
PE P PE
PE
PE
RNC Backup
P
MGW
Data path
49
Failures not protected by FRR

Can build HA into platform but will not achieve platform availability of 5 9s
P P
PE P
PE
PE
PE
MGW
MGW
P
50
AToM resilience for PE failure
OAM RDI is sent out on ATM VC and MGW uses redundant path 12.0.26S MR-APS 12.0.30S Backup Pseudowire 12.0.31S
51
52
ATM over MPLS

ATM Cell
Tunnel header AToM header ATM Header ATM header
H E C
Payload
5 bytes
48 bytes
Payload
AToM
HDLC header
4 bytes 8 bytes
4 bytes
ATM header
48 bytes
Payload FCS and Flag
POS
PPP Tunnel header header
AToM header
3 bytes 2 bytes 4 bytes 8 bytes
4 bytes
48 bytes
3 - 5 bytes
ATM Cell AToM PPP HDLC
ATM over MPLS 22 bytes per ATM Cell

VC, VP and port mode
53
Cell packing on 4 port STM1 ATM Engine 3
Effect of cell packing on throughput & latency for 100Mbps of ATMoMPLS

Latency us/Bandwith Mbps 350 300 250 200 150 100 50 0 1 2 3 4 10 20 28 Cell pack
Latency us Bandwidth
54
ATM Cell Packing Maximum Cell Packing Timers (MCPT)

int atm 7/1 atm mcpt-timer 50 100 300 atm pvp 1 l2transport xconnect 122.122.102.102 101 encapsulation mpls cell-packing 5 mcpt-timer 1 In microseconds Example Configuration for packing 5 cells with an MCPT of 50 microseconds (mcpt-timer 1)
1 MCPT timeout group per interface; 3 independent timers Values set in microseconds in 50 microsecond increments up to 25 milliseconds Maximum Cells set on the VC, VP, Port Cell-packing command specifies timer 1 is to be used.
55
Agenda
56
R4 with IP MGW voice interconnect

Ran Edge
Ran Core
PSTN
GMSC
IP/MPLS
MSC Server
BTS
BSC MGW
NodeB
RNC
ATM
IP/MPLS SDH
57
Deployment considerations for IP enabled MGW
Resilience Call admission control Overheads
58
Resilience Call admission control Overheads
59

PE or CE or hybrid? ATM, GE or POS connected? Dual homed?
Multiple IP addresses Single destination IP address with mulitple physical connections Multiple ATM VCs (L2)
Is the MGW redundant path aware and does it converge? What are the effects of loss of control/user plain connections? Clocking and synchronisation?
60
Simple deployment of two MGW to be considered

MSC Server
IP/MPLS MGW1 MGW2
Control plane User plane

61
Example of a MGW connected via multiple HSRP groups

MSC Server not shown
Tunnel1
HSRP .2 .1
PE PE P
HSRP .2
Active
Active
.4
PE PE
.4
.3 .3
.1
MGW1 MGW
MGW MGW2
12.0.30S supports Fast HSRP hellos 50ms hello, 250ms dead
Tunnel2
User and control plane may use separate HSRP groups
Data path Primary tunnel Backup tunnel

62
Link protection of all links
Primary
HSRP .2 .1
PE PE P
HSRP .2
Active
Active
.4
PE PE
.4
.3 .3
.1
MGW1 MGW
MGW MGW2
Backup
P

63
Node protection for P node failure
Primary
HSRP .2 .1
PE PE P
HSRP .2
Active
Active
.4
PE PE
.4
.3 .3
.1
MGW1 MGW
MGW MGW2
Backup
P

64
FRR does not protect from PE failure

MGW1 converges quickly Return path is the problem
P
Tunnel1
HSRP .2 .1
PE P PE
HSRP .2
Active
.4
PE PE
.4
.3 Active .3
.1
MGW1
MGW2
HSRP fails over to PE2 in 250ms
Tunnel2

65
FRR attempts a fix

Tunnel1
HSRP .2 .1
PE P PE
HSRP .2
Active
.4
PE PE
.4
.3 Active .3
.1
MGW1 MGW
MGW2 MGW
Tunnel2 FRR attempts to reroute round link failure


66
MPLS VPN eventually converges

Tunnel1
HSRP .2 .1
PE P PE
HSRP .2
Active
.4
PE PE
.4
.3 Active .3
.1
MGW1 MGW
MGW2 MGW
L3 VPN converges
Tunnel2

67
Convergence for PE node failure Feb 2005

PE power off failure far end convergence
5000 traffic loss (msec) 4000 3000 2000 1000 0 999990 0% 50% 100% average
All VPNv4
999992 999994 prefix nr 999996 999998
90%
68
Convergence for PE node failure Aug 2005

PE power off failure far end convergence
traffic loss (msec)
2000 1500 1000 500 0

0 2 4 6 8
0% 50% 90% 100%
prefix nr
69
Resilience 5 9s Call admission control Overheads
70
In R4 split MSC the user plain and control plane are separate
MSC Server
IP/MPLS MGW1
PE
PE
MGW2
There may be enough bandwidth to signal a call but not to carry it
Control plane User plane

71
Capacity planning real time traffic based on known topology

30Mbps CBR Only real time priority queue considered in this example and no over booking Max call Capacity 90Mbps
STM1 POS
MGW
PE
30Mbps CBR
MGW
PE
PE
MGW
30Mbps CBR
MGW
PE
STM1 ATM STM1 POS

72
Capacity planning real time traffic using TE

30Mbps CBR MPLS TE uses RSVP to reserve bandwidth ensuring traffic is well distributed and priority queue is not over booked
MGW
PE
30Mbps CBR
30Mbps TE tunnels
IP/MPLS core
Max call Capacity 90Mbps
MGW
PE
PE
MGW
30Mbps CBR
MGW
PE
STM1 ATM STM1 POS

73
ISC:TEM Major Features

Discovery, Audit and Provisioning Manage change Provision primary tunnels FRR link and node protection Bandwidth Protection
Compute placement of Fast Reroute backup tunnels to protect critical network elements Protect bandwidth against link, node, or SRLG (Shared Risk Link Group) failures
Can take into account delay figures for links from other sources ie find a path with 10Mbps and 5ms delay for primary and backup.
74
The CAC problem

Can us static TE tunnels across network to ensure required capacity is available but what traffic load do you base the tunnels on?
Busiest hour of the year (new years eve at 2359hrs) Over provisioning can be expensive As is often the case plan for 95th percentile relying on graceful degradation on other occasions In an IP based solution degradation may not be graceful What failure cases do you include? Link or node failure could double bandwidth on links Will this failure occur at the busiest hour?
75
Example traffic profile from red MGW even distribution of calls

MGW1 MGW3
0.33G
PE1
PE3
1G
0.33G
0.33G
PE2 PE4
2.5G
MGW2
MGW4
Each MGW can support up to 1G of calls, assume even distribution

76
Traffic profile from all MGW even distribution

MGW1 0.33G 0.G 0.33G 0.66G MGW3
PE1
PE3
1G
0.33G 0.66G 0.33G 0G
0.66G 0.33G
0G 0.33G
PE2 PE4
2.5G
0G 0.33G 0.66G 0.33G MGW2 MGW4

77
Loading on links from all MGW even distribution

MGW1 MGW3
1.33G
PE1
PE3
1G
1.33G
1.33G
PE2 PE4
2.5G
1.33G
MGW2
MGW4

78
Loading on links during failure

MGW1
Congestion
MGW3
2.66G
PE1
PE3
1G
1.33G
1.33G
PE2 PE4
2.5G
MGW2
MGW4

79
Resilience 5 9s Call admission control Overheads
80
Voice over IP encapsulation

G729 packet
20 bytes
RTP
12 bytes 60 bytes
UDP
8 bytes
IP Header
20 bytes
G729 packet
20 bytes 24 bytes
cRTP
2-4 bytes
cRTP reduces voice bandwith requirement by 60%
G.729 no cRTP / no VAD: (20 + 40) * 8 * 50 = 26.4kbps G.729 cRTP / no VAD: (20 + 4) * 8 * 50 = 9.6kbps
*Sampling interval 20ms
81
VoIP over MPLS VPN with TE

MPLS/POS
22 bytes
G729 packet
20 bytes
RTP
12 bytes
UDP
8 bytes
IP Header
20 bytes
20 byte voice becomes 82 bytes

82x8x50 = 33kbps
In an attempt to reduce latency sampling may be 100 times a second

72x8x100 = 58kbps
82
Agenda
83
Customer example Vodafone UK
84
Presentation_ID
85

Adam M IP MPLS For R4 Final Print

Diunggah oleh

Informasi Dokumen

Judul Asli

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

Adam M IP MPLS For R4 Final Print

Diunggah oleh

Hak Cipta:

Format Tersedia

Cisco IP/MPLS for R4/5

Adam MacHale October 2005

Customer proof of concept labs CPOC Bedfont UK

2005 Cisco Systems, Inc. All rights reserved.

2005 Cisco Systems, Inc. All rights reserved.

2005 Cisco Systems, Inc. All rights reserved.

Mobile operators proliferation of disparate networks

2005 Cisco Systems, Inc. All rights reserved.

Stage 1 migrate disparate networks to single MPLS core

Single network over high capacity transmission carrying all services

The business case for a converged network

By reducing the number of networks

2005 Cisco Systems, Inc. All rights reserved.

Stage 2 reducing complexity and overlap

RNC SGSN PDSN GGSN GRX

2005 Cisco Systems, Inc. All rights reserved.

What next for IP/MPLS core in mobile

2005 Cisco Systems, Inc. All rights reserved.

2G and R99 voice

BSC MGW MSC

R4 with ATM MGW voice interconnect

MSC MGW NodeB RNC

ATM IP/MPLS SDH

R4 with IP MGW voice interconnect

Converged backbone platform positioning

Iu-PS Gn Iu-CS Iu-PS Iur P

Iu-CS P SS7 ITP

Managed and secure

2005 Cisco Systems, Inc. All rights reserved.

Deploying Tight-SLA services on an IP Backbone

Physical network design and topology

Solution Diffserv Architecture: RFC2475

Classification and conditioning (meter, marking, policing) at EDGE

Packet colour In DSCP

2005 Cisco Systems, Inc. All rights reserved.

Aggregate PHBs in backbone (EF, AF, DF, CSn)

Simple view of deploying QoS

Marking at the edge

Outbound Interface POS or GE

2005 Cisco Systems, Inc. All rights reserved.

QoS Mapping for 3G Data Roaming (From GSMA PRD IR.34)

EF AF41 AF31 AF21 AF11 BE

10-6 10-8 10-8 10-8 10-8

2005 Cisco Systems, Inc. All rights reserved.

3GPP to Diffserv marking the edge

2005 Cisco Systems, Inc. All rights reserved.

3GPP to Diffserv and MPLS Exp marking in the core

2005 Cisco Systems, Inc. All rights reserved.

MQC queuing setup

Priority / AAL2 Routing/signalling

Priority max 80%

% is of remaining bandwidth after priority queue

2005 Cisco Systems, Inc. All rights reserved.

Measuring SLA Metrics

Availability and performance

Traffic type and volume

2005 Cisco Systems, Inc. All rights reserved.

Availability and measurement

Multiple Test Packets

Equal Time Spacing

2005 Cisco Systems, Inc. All rights reserved.

Cisco Service Assurance Agent reporting

2005 Cisco Systems, Inc. All rights reserved.