Prod Presentation0900aecd8031280f

QoS in MPLS Networks
RST-1607 Santiago lvarez saalvare@cisco.com CCIE 3621
RST-1607 Networkers 2004
2003 Cisco Systems, Inc. All rights reserved.
Prerequisites
Basic understanding of MPLS (L3VPN, L2VPN, TE) Basic understanding of QoS (DiffServ)
Copyright 2003, Cisco Systems, Inc. All rights reserved. Printed in USA. Presentation_ID.scr
Agenda
Technology Overview Backbone Infrastructure IP Services Layer-2 Services Management
MPLS QoS Technology Overview
MPLS QoS Architectures

MPLS does NOT define new QoS architectures MPLS QoS uses Differentiated Services (DiffServ) architecture defined for IP QoS DiffServ Architecture defined in RFC2475 MPLS support for DiffServ defined in RFC3270
Differentiated Services Architecture

DiffServ Domain
Ingress Node Interior Node Egress Node
TCA PHB
PHB
TCA PHB
Traffic Conditioning Agreement (TCA)

Classification/Marking/Policing/Shaping
Per-Hop Behavior (PHB)

Queuing/Dropping
Whats Unchanged in MPLS Support of DiffServ

Functional components (TCA/PHB) and where they are used
Classification, marking, policing, and shaping at network boundaries Buffer management and packet scheduling mechanisms used to implement PHB
PHB definitions
Expedited Forwarding (EF): low delay/jitter/loss Assured Forwarding (AF): low loss Default (DF): No guarantees (best effort)
Whats New in MPLS Support of DiffServ

How aggregate packet classification is conveyed (E-LSP vs. LLSP) Interaction between MPLS DiffServ info and encapsulated DiffServ info (e.g. IP DSCP)
EXP-Inferred-PSC LSP (E-LSP)

Frame Encapsulation
Layer-2 Header Layer-2 Header Label Stack Label Header Label Header Label Header Label Header Payload Payload
0
MPLS Shim Header

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 Label
EXP S EXP S EXP
TTL TTL
Class & Drop Precedence
Packet Class and drop precedence inferred from EXP (3-bit) field RFC3270 does not recommend specific EXP values for DiffServ PHB (EF/AF/DF) Used for frame-based MPLS
Label-Only-Inferred-PSC LSP (L-LSP)

Frame Encapsulation
Layer-2 Header Layer-2 Header Label Stack Label Header Label Header Label Header Label Header Payload Payload
MPLS Shim Header

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 S
TTL TTL
Class
Drop Precedence
Cell Encapsulation
GFC GFC VPI VPI VPI VPI VCI VCI
Label CLP
VCI VCI PTI CLP VCI PTI HEC HEC
Label VCI
Class
Drop Precedence
Packet Class inferred from label Drop precedence inferred from EXP or ATM CLP Can be used for frame-based and cell-based MPLS
10
E-LSP vs. L-LSP

An E-LSP may carry multiple classes (max 8, in real life less than that) An L-LPS carries one class Both E-LSP and L-LSP can use LDP or RSVP for label distribution Cisco products currently support E-LSP for frame-mode MPLS No demand for L-LSP support with frame-mode MPLS yet
11
MPLS Support of DiffServ All done with Modular QoS CLI (MQC)
class-map [match-any | match-all] class-name Enters configuration sub-mode for class definition class-map match-all REAL-TIME match mpls experimental topmost 5 class-map match-all PREMIUM match mpls experimental topmost 1 2 ! ! policy-map OUT-POLICY class REAL-TIME priority percent 25 class PREMIUM bandwidth remaining percent 50 random-detect class class-default random-detect ! interface POS1/0 ip address 10.150.1.1 255.255.255.0 service-policy output OUT-POLICY !
policy-map policy-name
Enters configuration sub-mode for policy definition (marking, policing, shaping, queuing, etc.)
service-policy {input | output} policy-name
Command in interface configuration sub-mode to apply QoS policy for input or output traffic
Template-based command syntax for QoS Separates classification engine from QoS functionality Platform-independent CLI for QoS features
12
MPLS QoS Backbone Infrastructure
RST-1607 Networkers 2004 Presentation_ID
2003, Cisco Systems, Inc. All rights reserved. Cisco Systems, Inc. All rights reserved. 2003
13
QoS in my Backbone?
Do you think my core looks like this? Of course not (or does it?)
14
Backbone Requirements
Growing trend: MPLS as selected choice for next gen multiservice network
IP VPN
IPv6 ATM
MPLS QoS architecture must fit mutiservice strategy

Frame Relay
PPP
IP/MPLS
Architecture must be flexible and scalable
Internet VoIP
Ethernet
PSTN
15
Some Theory First Delay and Loss vs. Utilization

Traffic patterns still debated (markovian, self-similar, others) Queuing maths not fully cooked but:
if utilization* kept below x%, performance is excellent as utilization* approaches 100%, performance degrades
x% 100% Link utilization*
Delay/Loss
* Measured on a large timescale

16
Selecting Utilization Level (x%)

Target Utilization Level (x%) is a function of:
Target QoS guarantees (delay, jitter, loss) Failure handling policies (link, node, SLRG) School of thoughts for queuing theory Heuristics Risk tolerance Testing Politics Technology religion, etc.
* Measured on a large timescale
Delay/Loss
x%
100%
Link utilization*
17
Enforcing Utilization Level (x%)

Aggregate capacity planning
Adjust link capacity to expected link load
MPLS DiffServ
Adjust class capacity to expected class load
MPLS Traffic Engineering

Adjust link load to actual link capacity
MPLS DiffServ-Aware TE (DS-TE)

Adjust class load to actual class capacity
18
Lets Review Technology options MPLS DiffServ

DiffServ Domain
Ingress Node Interior Node PHB Egress Node TCA PHB Per-Hop Behavior (PHB)
Queuing/Dropping
TCA PHB
Traffic Conditioning Agreement (TCA)

Classification/Marking/Policing/Shaping
Already discussed in overview section Only PHB relevant in the backbone

19
MPLS TE Overview
Introduces explicit routing Supports constrained-based routing Supports admission control Protection capabilities RSVP-TE to establish LSPs ISIS and OSPF extensions to advertise link attributes Lots more in session RST-2603
TE LSP
IP/MPLS
20
How MPLS TE Works

Information Distribution
ISIS-TE IP/MPLS OSPF-TE
Path Calculation (CSPF) Path Setup (RSVP-TE) Forwarding Traffic down Tunnel
Auto-route Static
Mid-point Head end Tail end
PBR CBTS Forwarding Adjacency Tunnel select
21
DiffServ-Aware Traffic Engineering (DS-TE)

Brings per-class dimension to MPLS TE
Per-class constrainedbased routing Per-class admission control
IP/MPLS
Low-Latency TE LSP with Reserved BW Best-Effort TE LSP
22
DiffServ-Aware Traffic Engineering (DS-TE)

Control Plane Link BW distributed in pools or Bandwidth Constrains (BC)
Maximum Reservable Bandwidth
DS-TE BW Allocation
Up to 8 BW pools Different BW pool models Unreserved BW per TE class computed using BW pools and existing reservations
Forwarding Plane
Unreserved BW per TE class advertised via IGP

Link/shaper rate
DiffServ BW Allocation
23
DS-TE Bandwidth Pools Maximum Allocation Model (MAM)

BW pool applies to one class Sum of BW pools may exceed MRB Sum of total reserved BW may not exceed MRB
BC0
Class1
BC1
Class2
Maximum All Reservable classes Bandwidth (MRB)
BC2
Class3
BC0: 20% Best Effort BC1: 50% Premium BC2: 30% Voice
24
DS-TE Bandiwdth Pools Russian Dolls Model (RDM)

BW pool applies to one or more classes
BC0 All classes (class1 Maximum Reservable + Class2 Bandwidth + (MRB) Class3)
Global BW pool (BC0) equals MRB

BC1
BC0..BCn used for computing unreserved BW for class n
Class2 + Class3
BC2
Class3
BC0: MRB Best Effort + Premium + Voice BC1: 50% Premium + Voice BC2: 30% Voice
25
DS-TE Bandiwdth Pools Why Russian Dolls Model?

Good match for common bandwidth allocation in forwarding plane
VoIP gets priority treatment and is unaffected by other traffic: use BC2 Business Data gets preferential access to link vs. BE: use BC1 Best effort may use MRB if other classes not fully used, but should be reduced if lots of VOIP or Business Data: use BC0
BC0 Maximum All Reservable classes Bandwidth (MRB)
BC1
Class1 + Class2
BC2
Class2
Good isolation between classes, efficient use of bandwidth

26
Class-Based Tunnel Selection CBTS

T1 T2 Dst1
T3 T4 T5 T6 T7
EXP-based selection between multiple tunnels to same destination Local mechanism to head-end Tunnels configured with EXP values to carry Tunnels may be configured as default
Dst2
Dst3
FIB
Dst1, exp 4 Dst1, * Dst2, exp 4 Dst2, exp 2 Dst2, * Dst3, exp 4 Dst3, * Tunnel1 Tunnel2 Tunnel3 Tunnel4 Tunnel5 Tunnel6 Tunnel7
No IGP extensions VRF aware Simplifies use of DS-TE tunnels Similar operation to ATM/FR VC Bundles
* Wildcard EXP value

27
Dealing with Failure Scenarios

Load vs Capacity in the absence of failure
During a failure:
Are you missing your SLA? For how long?
Load
Capacity
Link failure may have 2x impact on load Node / SRLG failure may have a 4x impact on load
Load vs Capacity during failure
Failure impact and duration dependent on:

Load Capacity
Network topology backbone QoS design
28
MPLS TE Fast Re-Route (FRR)

Subsecond recovery against node/link failures Scalable 1:N protection
PE PE
PE
IP/MPLS
Bandwidth protection Greater protection granularity Cost-effective alternative to optical protection
Primary TE LSP Backup TE LSP
29
How MPLS TE FRR Works

Next-hop backup tunnel for Link Protection Next-next-hop backup tunnel for Node Protection
PE PE
PE
IP/MPLS
Point of Local Repair (PLR) swaps label and pushes backup label Local repair in msecs Failure detection critical for total repair time
Point of Local Repair (PLR)
Merge Point (MP)
PLR sends PathErr to head end triggering global reoptimization
Primary TE LSP Backup TE LSP
30
What should I Use in my Backbone?

Nothing MPLS TE MPLS DiffServ MPLS DiffServ + MPLS TE MPLS DiffServ + MPLS DS-TE Any of the above + MPLS TE FRR
31
Backbone with Nothing No MPLS DiffServ and No MPLS TE

Service Differentiation
A solution when:
No differentiation required No optimization required
Capacity planning as QoS tool Link over-provisioning to meet all SLAs

Resource Optimization
Nothing
Adjust link capacity to expected link load
Load
Capacity
32
Backbone with MPLS TE

A solution when:
No differentiation required Optimization required
Full mesh or selective deployment to avoid oversubscription

TE Resource Optimization
Increased network utilization Adjust link load to actual link capacity
Load
Capacity
33
Backbone with MPLS DiffServ

A solution when:
Differentiation required Optimization required
Per-class capacity planning

DiffServ
Same or lower number of classes than edge

Adjust class capacity to expected class load
Class2
Load Load Capacity Capacity
Class3 Load
Capacity
34
Backbone with MPLS DiffServ and MPLS TE

A solution when:
Differentiation required
DiffServ + TE
Optimization required
Adjust class capacity to expected class load Adjust class load to actual class capacity for one class
Alternatively, adjust link load to actual link capacity
Class2 Class1
Load Capacity Load Capacity
Class3 Load
Capacity
35
Backbone with MPLS DiffServ and MPLS DS-TE

DiffServ + DS-TE
A solution when:
Strong differentiation required Fine optimization required
Adjust class capacity to expected class load Adjust class load to actual class capacity
Resource Optimization Class2 Class1
Load Capacity Load Capacity
Class3 Load
Capacity
36
Bringing MPLS TE FRR into the Mix

FRR
Increases resiliency regardless of backbone QoS design

DiffServ + DS-TE DiffServ + TE FRR DiffServ FRR
Stronger SLAs during single failure conditions (link, node, SLRG) Optimization of backup resources
FRR Nothing
FRR TE
37
What Model to Use?

Take your pick !!! As sophisticated as necessary, but not more
FRR Operational Complexity
IPv6 ATM DiffServ + DS-TE DiffServ + TE FRR DiffServ FRR Nothing FRR TE Internet VoIP FRR
IP VPN
PPP
IP/MPLS
Frame Relay
Ethernet
PSTN
38
MPLS QoS IP Services
RST-1607 Networkers 2004 Presentation_ID
2003, Cisco Systems, Inc. All rights reserved. Cisco Systems, Inc. All rights reserved. 2003
39
QoS for IP Services

Elaborate DiffServ Edge implementation
Access link capacity controlled by customer (prone to congestion) Trust boundary (SLA enforcement)
CE CE
CE
PE IP/MPLS PE
PE
Applies to both IPv4 and IPv6 Backbone must be able to support customer SLA Per-customer QoS policies only at the edge
CE
PE
CE
CE
40
Site IP SLA
Typically between 3 and 5 classes (real time, video, interactive, business, BE) Delay, jitter and loss guarantees for conforming real-time traffic Combination of delay and loss guarantees for data traffic Sum of committed bandwidth (per-class CIR) not to exceed link/shaper rate Additional classes not visible to customer may exist (e.g. management, control traffic)
Class1 Class2 Class4 Class2 Class5 Link/shaper rate
Class Real time Interactive Business Best Effort
Committed BW Delay X Y Z NA Low Low NA NA
Jitter Low NA NA NA
Loss Low Low Low NA
41
IP SLA between Sites

Site-to-network (point-tocloud) guarantees for conforming traffic Each site may send x% of class n to network per SLA Each site may receive x% of class n from network per SLA No site-to-site (point-to-point) guarantees
CE CE CE
CE
PE
PE
IP/MPLS
PE PE
42
IP SLA Enforcement
Managed vs. unmanaged IP service Trust boundary on PE for unmanaged service Trust boundary on CE for managed service Trust boundary defines SLA enforcement point Different QoS design options
Site 1 Site 2
Unmanaged CE
IP/MPLS
PE PE
Managed CE
43
Lets See how SLA enforcement is done IP QoS Managed Service

CE output and PE output policies enforce SLA Traffic classification and marking on CE No input QoS policies generally needed Explicit-null encapsulation may be used on CE to avoid remarking customer traffic
Managed CE
PE
CE Output Policy Classification / Marking LLQ WRED [Shaping] [LFI / cRTP]
PE Output Policy LLQ WRED [Shaping] [LFI / cRTP]
44
IP QoS Unmanaged Service

PE input and PE output policies enforce SLA Traffic classification and markings on PE
Unmanaged CE
PE
CE Output Policy <NOT SP controlled >
PE Input Policy Classification / Marking Policing PE Output Policy LLQ WRED [Shaping] [LFI / cRTP]
CE policies require coordination with PE policies (e.g. LFI, cRTP, end-to-end latency)
45
Sample PE Input Policy Unmanaged Service

Real Time
Excess real time (voice) usually dropped Excess data marked down
Video
Interactive
Dropping excess data at policer would affect many TCP sessions Limited bandwidth sharing between classes with aggregate sub-rate Voice and video will benefit from admission control
Policing
Business
Best Effort
Classifier
46
Sample CE Output Policy Managed Service

Real Time Video Interactive TD Business WRED Best Effort WRED
Priority Queue Tail drop
TD
Classifier
Congestion Management Congestion Avoidance
Shaping
Link Fragmentation and Interleaving (LFI)
LFI used in slow links to reduce delay and jitter for real-time traffic WRED used for TCP-friendly packet dropping
47
How DiffServ Markings Interact DiffServ Tunneling Modes
PE
MPLS IP
Several models (modes) of interaction between these markings RFC2983 defines models (uniform/pipe) for DiffServ with IP tunnels RFC3270 defines models (uniform/pipe/short-pipe) for MPLS
EXP
POP PUSH
DSCP
Only relevant where pop or push operations take place (both on IP or MPLS packets) Explicit NULL label may be used for managed services
What is their relationship?
48
MPLS DiffServ Tunneling Modes

Uniform Pipe Short Pipe
IP
IP/MPLS
CE1 PE1 PE2 CE2
IP
49
Uniform Mode
LSP DiffServ Marking
IP
IP/MPLS
CE1 PE1 PE2 CE2
IP
Push
Packet remarked
Pop Packet served on LSP DiffServ Marking (propagated down)
Label Label
Label Label
IP or MPLS packet
50
Pipe Mode
LSP DiffServ Marking Tunneled DiffServ Marking
IP
IP/MPLS
CE1 PE1 PE2 CE2
IP
Push
Packet remarked
Pop Packet served on LSP DiffServ Marking*
Label Label
Label Label
IP or MPLS packet
* Pipe mode precludes Penultimate Hop Popping (PHP)

51
Short Pipe Mode

LSP DiffServ Marking Tunneled DiffServ Marking
IP
IP/MPLS
CE1 PE1 PE2 CE2
IP
Push
Packet remarked
Pop Packet served on Tunneled DiffServ Marking
Label Label
Label Label
IP or MPLS packet
52
Local Packet Marking
PE
MPLS IP
QoS Group Id and Discard class for local packet marking Always an input feature (before label POP) Used to implement uniform and pipe mode Recommended semantics
QoS Group Id
QoS group identifies class

Output Policy
EXP
Input Policy
POP
Discard Class
Discard class identifies drop precedence
Discard class can drive WRED Not all classes will have a drop precedence (e.g. EF, best effort)
53
DiffServ Tunneling Modes Keep present
PE
MPLS IP
When input policy defines EXP to be imposed, value applies to all imposed labels If no imposition EXP defined, IP precedence copied to all imposed labels EXP maintained during label swaps
EXP
POP PUSH
DSCP
EXP not propagated down by default during disposition Pipe mode precludes PHP
What is their relationship?
54
Some Advanced Configurations QoS Policy Propagation via BGP (QPPB)

Despite the name, no policies are really propagated Input packet marking (IP precedence, QoS Group Id) based on
Community
AS65001 Set community 65172:1 AS65000
PE
CE
IP/MPLS
CE
PE
PE PE
CE RR Mark EF if: community 65172:1 or AS65000 CE
AS Path IP Prefix
eBGP iBGP
Packet marking happens before input QoS policy Supports IPv4 and VPNv4 addresses Could add intelligence to IP SLA between sites
BGP Update BGP Table RIB Rx Packet
Prefix1 marking1 Prefix2 marking2 Prefix3 marking3
FIB Switch and Mark
Tx Packet
55
IP QoS in Inter-Provider Environments

Carrier Supporting Carriers (CsC)
IP VPN Customer Customer Carrier Backbone Carrier A-PE1 B-PE2 B-CE2 IP/MPLS B-PE1 B-CE1 Customer IP VPN Carrier Customer
Complex coordination between providers

Number of classes Markings SLAs
A-PE2 C-PE2 B-CE2 IP/MPLS C-PE1 B-CE1
IP/MPLS
End user may receive least common denominator MPLS DiffServ tunnel modes supports CsC hierarchies Tunnel modes my differ at different levels in a hierarchy
IP VPN Customer
Carrier A
Carrier B
IP VPN Customer
Inter-AS
A-PE2 A-CE2 IP/MPLS A-PE1 A-CE1
B-PE2 B-CE2 IP/MPLS B-PE1 B-CE1
56
MPLS QoS Layer-2 Services
Presentation_ID
2003, Cisco Systems, Inc. All rights reserved.
57
QoS for Layer-2 Services

Well-defined SLAs for Frame Relay/ATM Differentiation for Ethernet services Point-to-point SLA with exception of VPLS Backbone must be able to support customer SLA TE-enabled backbone attractive
PPP ATM Ethernet
IP/MPLS
Frame Relay
Ethernet
HDLC
58
Layer-2 SLA Enforcement

User interface vs network interface Trust boundary on PE for user interface Trust boundary on access network for network interface Trust boundary defines SLA enforcement point Different QoS design options
Site 1 Site 2
User Interface
IP/MPLS
PE PE
Network interface
59
Lets See how SLA enforcement is done Layer-2 QoS User Interface
PE input and PE output policies enforce SLA Drop precedence may be marked for FR / ATM / Ethernet Output drop precedence (e.g. ATM CLP, FR DE) marking when input marking not possible Ethernet may support multiple classes (802.1p bits)
User Interface
CE
PE
PE Input Policy Policing [Marking] PE Output Policy Queuing (LLQ) WRED [Marking] [shaping]
60
Layer-2 QoS Network Interface

Network Interface
SP enforces SLA on access network PE may only need simple aggregate policies
CE
PE
Access Network Input Policy Policing [Marking] Access Network Output Policy Queuing (LLQ) Dropping (WRED) [Shaping]
PE Input Policy [Marking]
PE Output Policy <optional>
61
Encapsulation Details Layer-2 QoS Frame Relay

DLCI
Output Policy
EXP
PUSH
Input Policy
FR DE
PE
MPLS Frame Relay
Incoming traffic classified by DE or DLCI for DLCI-to-DLCI mode Input policer may exclude DEmarked frames from CIR metering Several classes of service may be implemented
CIR (EIR=0) CIR+EIR CIR=EIR=0
QoS Group Id
EXP
Input Policy
POP
Discard Class
Output Policy
FR DE
Output DE marking when input marking not possible FECN/BECN marking supported on egress PE only Control word carries original DE/FECN/BECN values
62
Layer-2 QoS ATM

Incoming traffic classified by CLP Support for all service categories (CBR, rt-VBR, nrt-VBR, ABR, UBR) Different traffic conformance supported (CBR.1, VBR.1, VBR.2, VBR.3, UBR.1, UBR.2) ATM TM 4.0 metering parameters converted to MQC (token-bucket) policer parameters
CIR = SCR*53*8 PIR = PCR*53*8 bc/be = CDVT*(CIR+53)*8 bc = MBS*PCR/SCR Output Policy
EXP
PUSH
Input Policy
CLP
PE
MPLS ATM
Output queuing handled by ATM hardware Cell-relay transport for delay sensitive traffic Control word carries original CLP and EFCI values
EXP
Input Policy
POP
Discard Class
Output Policy
CLP
63
Layer-2 QoS Ethernet

VLAN ID Output Policy
EXP
PUSH
Input Policy
Incoming traffic classified by CoS (802.1p) or VLAN Id for 802.1Q encapsulation Service characteristics being proposed at the Metro Ethernet Forum (BW Profile: CIR, CBS, EIR, EBS, CF, CM) Site-to-network (point-tocloud) SLA for VPLS Control word does not carry any CoS (802.1p) info
CoS
PE
MPLS Ethernet
QoS Group Id
EXP
Input Policy
POP
Discard Class
Output Policy
CoS
64
Layer-2 QoS PPP/HDLC

No layer-2 marking to set or classify on No standard service definition but classes of service are possible
Output Policy
EXP
PUSH
Input Policy
PE
MPLS PPP/HDLC
QoS Group Id
EXP
Input Policy
POP
Discard Class
Output Policy
65
Coupling Layer-2 Services with MPLS TE Tunnel Selection

Static mapping between pseudo-wire and TE Tunnel on PE Implies PE-to-PE TE deployment TE tunnel defined as preferred path for pseudowire Traffic will fall back to peer LSP if tunnel goes down
CE CE
CE
PE
IP/MPLS
PE
CE
PE
CE
TE LSP Layer 2 Circuit Layer 2 Circuit
66
Are We There Yet?
Were pretty much done!
67
MPLS QoS Management
Presentation_ID
68
Some Monitoring Tools Monitoring Utilization Level (x%)

Measuring internal and external traffic matrix
Interface MIB
AS65001 AS65002 AS65003
MPLS LSR MIB Cisco Class Based QoS MIB Netflow

NetFlow BGP Next Hop TOS Aggregation MPLS-Aware Netflow
PE P PE P PE PE
BGP Policy Accounting

Communities AS path IP prefix
PE
P POP
P POP
PE
Server Farm
Server Farm
69
Cisco Class-Based Qos MIB

Primary per-link accounting mechanism for QoS:
Classification (cbQosMatchStmtStats/ cbQosClassMapStats) Marking (cbQosClassMapStats) Policing (cbQosPoliceStats) Shaping (cbQosTSStats) Congestion management (cbQosQueueingStats) Congestion avoidance (cbQosREDClassStats)
Management Station
QoS policy must be applied to interface/PVC for accounting to happen Read access to configuration and statistical information for MQC
70
NetFlow MPLS Features Overview

Traditional NetFlow (IP to MPLS) MPLS-Aware NetFlow (MPLS to MPLS) Egress MPLS NetFlow (MPLS to IP)
IP/MPLS
PE
PE
Output Sampled NetFlow (MPLS to IP, IP to IP)
Lots of detailed info in session NMS-2032

71
BGP Policy Accounting

Assign counters (traffic-index) to IP traffic based on:
Community AS Path IP Prefix
CE Set community 65172:1 AS65000
PE
CE
IP/MPLS
CE
PE
PE PE
AS65001 eBGP iBGP RR Count packets if: community 65172:1 or AS65000 CE
Up to 64 counters (traffic-index) Supports IPv4 and VPNv4 addresses Similar in concept/operation to QPPB, but accounting instead of marking
BGP Update BGP Table RIB Rx Packet
Prefix1 traffic-idx1 Prefix2 traffic-idx2 Prefix3 traffic-idx3
FIB Switch and count
Tx Packet
72
Performance Monitoring Cisco IOS SAA Today

Applications
Availability Network Performance Monitoring VoIP Monitoring Service Level Agreement (SLA) Monitoring Network Assessment MPLS Monitoring Trouble Shooting
Measurement Metrics
Round Trip Time Packet Loss Network Jitter Uni-directional Measurements Connectivity
Operations
Jitter FTP Path Jitter DNS DHCP DLSW ICMP UDP TCP HTTP VoIP LSP
Soon
H323 RTP
Defined Packet Size, Spacing COS and Protocol Source MIB Data Source MIB Data SAA
IP Server
SAA Destination Active Generated Traffic Responder

73
SAA
SAA Jitter Operation Example

1. Send train of packets with constant Interval 2. Receive train of packets at Interval impacted by Network
Source SAA
Responder
4. Compute: Per-direction OWD (one-way delay) Per-direction inter-packet delay (Jitter) Per-direction packet loss
3. Time stamp when Rxed Increment Rx Count Delta Time
74
SAA Reaction Conditions

Threshold Violation 100ms 50ms Threshold violation Resolution Alert No Alert Threshold Violation Alert
SAA event triggers Connection loss / timeout Latency (one way, round trip) Jitter (one way, round trip) Loss (one way, round trip) MOS
Trigger threshold definitions Immediate Average Consecutive X out of Y times
SAA Triggers can generate SNMP trap or another SAA probe

75
Cisco SAA MPLS Embedded Management
NMS
MPLS L3 VPN MPLS L2 VPN
Echo and Path Tracing RTT, Packet Loss, Jitter Statistics
Auto-Config / CLI / MIB
Traps Syslog
SAA Probe Generation
Reaction Configuration Thresholds
SAA MIB Data Storage
MPLS LSP Ping/Trace
VCCV
ECMP Tree Trace
76
Key Cisco Partners

NetFlow SAA
Flow-Tools
Cisco NetFlow Collector
IP Solution Center Internetworking Performance Monitor (IPM)

77
IP Solution Overview
Unified mgmt for MPLS VPN, L2VPN, Security and MPLS TE
78
ISC QoS Management Features

QoS provisioning on access link (both CE and PE) Internal constrain matrix check software and hardware dependencies Support for pre-MQC QoS functionality QoS provisioning on backbone links using Smart Template utility
CE CE CE
PE
PE
IP/MPLS
PE PE
classification marking policing shaping congestion management congestion avoidance LFI cRTP
CE
79
Summary
Technology Overview
MPLS support of DiffServ MQC
Backbone Infrastructure
DiffServ TE/FRR DS-TE IPv6 ATM IP VPN
IP Services
SLAs DiffServ Tunneling Modes PPP
IP/MPLS
Frame Relay
Layer-2 Services
Frame Relay ATM Ethernet PPP/HDLC Internet VoIP Ethernet
Management
MIBs Netflow BGP Policy Accounting SAA
PSTN
80
Q and A
Presentation_ID
81
Reference Materials
MPLS QoS Q&A
http://www.cisco.com/warp/public/cc/pd/iosw/tech/mpotc_qp.htm
Cisco IOS 12.0S/12.2S QoS Configuration Guide and Command References

http://www.cisco.com/
Multi-Protocol Label Switching (MPLS) Support of Differentiated Services RFC3270

http://www.ietf.org/rfc/rfc3270.txt
82
Associated Sessions
RST-1601 - Introduction to Multiprotocol Label Switching (MPLS) RST-3606 - Troubleshooting MPLS VPNs RST-2602 - Deploying MPLS VPNs RST-2603 - Deploying MPLS Traffic Engineering RST-2606 - Understanding Convergence in MPLS VPN Networks RST-2T09 - Advanced Concepts and Developments in MPLS RST-3605 - Troubleshooting MPLS Networks RST-4607 - Advanced Topics and Future Directions in MPLS NMS-4012 - MPLS Embedded Management Tools NMS-2032 - Netflow for Accounting, Analysis and Attack
83
Recommended Reading
IP Quality of Service
ISBN 1578701163
Traffic Engineering with MPLS

ISBN 1587050315
MPLS and VPN Architectures

ISBN 1587050021
84
Presentation_ID
85

Prod Presentation0900aecd8031280f

Transféré par

Informations du document

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Prod Presentation0900aecd8031280f

Transféré par

Droits d'auteur :

Formats disponibles

QoS in MPLS Networks

RST-1607 Santiago lvarez saalvare@cisco.com CCIE 3621

RST-1607 Networkers 2004

2003 Cisco Systems, Inc. All rights reserved.

RST-1607 Networkers 2004

2003 Cisco Systems, Inc. All rights reserved.

RST-1607 Networkers 2004

2003 Cisco Systems, Inc. All rights reserved.

MPLS QoS Technology Overview

RST-1607 Networkers 2004

2003 Cisco Systems, Inc. All rights reserved.

MPLS QoS Architectures

RST-1607 Networkers 2004

2003 Cisco Systems, Inc. All rights reserved.

Differentiated Services Architecture

Traffic Conditioning Agreement (TCA)

Per-Hop Behavior (PHB)

RST-1607 Networkers 2004

2003 Cisco Systems, Inc. All rights reserved.

Whats Unchanged in MPLS Support of DiffServ

RST-1607 Networkers 2004

2003 Cisco Systems, Inc. All rights reserved.

Whats New in MPLS Support of DiffServ

RST-1607 Networkers 2004

2003 Cisco Systems, Inc. All rights reserved.

EXP-Inferred-PSC LSP (E-LSP)

MPLS Shim Header

EXP S EXP S EXP

Class & Drop Precedence

Label-Only-Inferred-PSC LSP (L-LSP)

MPLS Shim Header

VCI VCI PTI CLP VCI PTI HEC HEC

E-LSP vs. L-LSP

RST-1607 Networkers 2004

2003 Cisco Systems, Inc. All rights reserved.

MPLS QoS Backbone Infrastructure

RST-1607 Networkers 2004 Presentation_ID

RST-1607 Networkers 2004

2003 Cisco Systems, Inc. All rights reserved.

MPLS QoS architecture must fit mutiservice strategy

Architecture must be flexible and scalable

RST-1607 Networkers 2004

2003 Cisco Systems, Inc. All rights reserved.

Some Theory First Delay and Loss vs. Utilization

* Measured on a large timescale

Selecting Utilization Level (x%)

Enforcing Utilization Level (x%)

MPLS Traffic Engineering

MPLS DiffServ-Aware TE (DS-TE)

RST-1607 Networkers 2004

2003 Cisco Systems, Inc. All rights reserved.

Lets Review Technology options MPLS DiffServ

Traffic Conditioning Agreement (TCA)

Already discussed in overview section Only PHB relevant in the backbone

RST-1607 Networkers 2004

2003 Cisco Systems, Inc. All rights reserved.

How MPLS TE Works

PBR CBTS Forwarding Adjacency Tunnel select

RST-1607 Networkers 2004

2003 Cisco Systems, Inc. All rights reserved.

DiffServ-Aware Traffic Engineering (DS-TE)

Low-Latency TE LSP with Reserved BW Best-Effort TE LSP

RST-1607 Networkers 2004

2003 Cisco Systems, Inc. All rights reserved.