QoS support in IPv6

environments
Location, country
Date

Speaker name (or email address)

 Copy Rights
This slide set is the ownership of the 6DISS project via its
partners

The Powerpoint version of this material may be reused and

modified only with written authorization

Using part of this material must mention 6DISS courtesy

PDF files are available from www.6diss.org

Looking for a contact ?

Mail to : [email protected]
Or [email protected]
 Contributions
Main author
Athanassios Liakopoulos, GRNET, Greece

Contributors
Dimitrios Kalogeras, GRNET, Greece
Dimitrios Primpas, CTI, Greece
 Presentation Outline
The need for QoS
QoS Terminology & Metrics
QoS Architectures
IPv6 header & QoS
Configuration Examples
Performance Tests in IPv6 environments
Conclusions
 The need for QoS (1/2)
QoS developments in IP networks is inspired by
new types of applications:
VoIP, VoIP trunks
Low bit rate variability, strict delay requirements, jitter sensitive
Audio/video streaming
Low/medium bit rate variability, elastic delay and jitter
requirements
Networked virtual environments, interactive gaming
Medium bit rate variability, delay intolerant, jitter sensitive, error
intolerant
Mission critical, control traffic (signalling), tele-immersion
Delay requirements, error sensitive
 The need for QoS (2/2)
QoS developments in IP networks is inspired by
new types of applications:
Videoconferencing, high quality video distribution
High bit rate variability, strict delay requirements, jitter sensitive
Interactive-transactional applications, e.g. e-commerce
Delay requirements, error sensitive
Elastic IP applications
Delay, jitter and error tolerant
GRIDs & Collaborative Environments
Long lived connections, bulk data transfers, delay and error
tolerant
 Presentation Outline
The need for QoS
QoS Terminology & Metrics
QoS Architectures
IPv6 header & QoS
Configuration Examples
Performance Tests in IPv6 environments
Conclusions
 Terminology
Quality-of-Service (QoS): A set of service
requirements to be met by the network while
transporting a flow.

Class of Service (CoS): The definition of the

semantics and parameters of a specific type of QoS.

Service Level Agreement (SLA): A negotiated

agreement between a customer and a service
provider on levels of service characteristics and the
associated set of metrics. The content of the SLA
varies depending on the service offering and includes
the attributes required for the negotiated agreement.
 QoS Metrics
Bandwidth
Affected by the slowest link capacity, congestion
mechanisms, device forwarding capabilities, queue
scheduling, etc.
Delay
Consisting of queuing, transmission, propagation, and
switching delays.
Inter-packet Delay Variation - Jitter
Caused by traffic multiplexing, variations in queue
lengths, queue scheduling, etc.
Packet loss
Caused by buffer exhaustion, congestion control
mechanisms, etc.
 Presentation Outline
The need for QoS
QoS Terminology & Metrics
QoS Architectures
IPv6 header & QoS
Configuration Examples
Performance Tests in IPv6 environments
Conclusions
 QoS Architectures
Best Effort Internet
Integrated Services
Differentiated Services
 Best Effort Internet
All packets are treated equally.
Does not provide any performance
guarantees to traffic.
Unpredicted bandwidth, delay, jitter

Service guarantees may be provided

via over-provisioning!
 Integrated Services (IntServ)
Performance guarantees to traffic and
resource reservations are provided on per-
flow basis.
Explicit admission control via signalling:
RSVP
RSVP Path messages specify resource
requirements of the applications.
RSVP Resv messages specify the reserved
resources
 IntServ Services
Guaranteed Service
Provides guarantees for bandwidth (i.e. negligible
packet loss) and delay but not for jitter.
Emulates a virtual circuit.
Suitable for non-elastic applications
Controlled-load Service
Provides a better-than-best-effort service, similar to
services provides in a lightly-loaded network.
Does not provide any strict guarantees on bandwidth,
delay or jitter.
Suitable for elastic applications
 IntServ Scaling Issues
Each routers maintains per-flow state
information.
Data structures are created and maintained for
each active flow.
Each incoming packet is classified, policed
(token bucket) and forwarded based on the
flow state information.
Processing power is proportional of the concurrent
active flows.
Signalling overheads
RSVP refresh messages
Differentiated Services (DiffServ)
Performance guarantees are provided to traffic
aggregates rather than to flows.
Traffic classification is based on the Differentiated
Service Codepoint (DSCP) field in the IP header.
Different Per-Hop Behaviours (PHB) are defined for
each traffic class.
DiffServ functionality is pushed at the domain
boundaries.
Classification, policing, marking, etc are performed by
edge routers.
Core routers forward traffic based only on DSCP
values
Basic QoS Mechanisms in DiffServ
Classification
Metering / Colouring
Policing
Shaping
Queue management
Queue scheduling
Average
Token Rate

Maximum
Burst Size

Packet Rate
 DiffServ Per-Hop Behaviours
Expedite Forwarding PHB (EF-PHB)
Provide guaranteed bandwidth (a.k.a. negligible
loss), low delay and jitter.
Strict admission control: Non-conformant traffic is
dropped or shaped.
EF traffic should not be influenced by the other
traffic classes.
Assured Forwarding PHB (AF-PHB)
Four classes are defined that provide different
forwarding guarantees. Within each class, there
are three drop precedence.
Non-conformant traffic is remarked.
 Service examples in DiffServ
Premium IP
Based on Expedited Forwarding PHB (EF-PHB).
Gives absolute priority over any other class and
provides low delay/jitter plus negligible packet loss
guarantees.
Flavors: PIP Virtual Wire (source & destination
aware), PIP VoIP (destination unaware).
Less than Best Effort (LBE)
Exploits network resources without (negative) impact
other traffic classes
Suited for specific scavenger applications
Uses low priority queue 1% of capacity guaranteed
DiffServ Limitations & Challenges
Although performance guarantees are provided
on traffic aggregates, individual flows may
experience low grade services.
DiffServ lacks any signalling protocol for
resource allocation (admission control) and QoS
mechanisms control.
Inter-domain QoS service provisioning can be
difficult as no standardized class of service
mapping exist between peer providers.
 Presentation Outline
The need for QoS
QoS Metrics
QoS Architectures
IPv6 header & QoS
Configuration Examples
Performance Tests in IPv6 environments
Conclusions
 IPv6 & IPv4 Header Comparison
The IPv6 header is IPv4 Header IPv6 Header
redesigned. Version IHL
Type of
Service
Total Length Traffic
Version Flow Label
Class
Minimize header Fragment
Identification Flags
overhead and reduce the Offset
Payload Length
Next
Header
Hop Limit

header process for the Time to

Live
Protocol Header Checksum

majority of the packets. Source Address

Source Address
Less essential and Destination Address
optional fields are moved Options Padding

to extension headers - Fields name kept from IPv4 to IPv6

- Fields removed in IPv6 Destination Address
- Name & position changed in IPv6
- New field in IPv6

IPv6 and IPv4 headers are not interoperable.

 QoS fields in IPv6 Header
Traffic Class
An 8-bit field used to distinguish packets from different
classes or priorities.
Provides the same functionality as the type of service
field in the IPv4 header.
Flow label
A 20-bit field defining the packets of the flow.
Selected by the source and never modified in the
network.
Fragmentation or encryption is not anymore problem,
as in IPv4.
 Presentation Outline
The need for QoS
QoS Metrics
QoS Architectures
IPv6 header & QoS
Configuration Examples
Performance Tests in IPv6 environments
Conclusions
 Configuration steps in MQC
Define Class Map
Separate traffic into classes based on access lists
(ACLs), DSCP/ToS, MPLS EXP, protocol, etc. or
combinations of those criteria
class-map [match-any | match-all] class-name

Define Policy Map (Service Policy)

Associate a class map with one or more QoS
policies, e.g. bandwidth allocation, queue
management, (re)-marking
policy-map policy-map-name
 Configuration steps in MQC
Apply a Service Policy to an interface
Associate a policy map to an physical or logical
interface at input or output.
service-policy {input | output} policy-map-name
 Configuration examples
class-map match-any
ip_premium_out IP Premium
match ip dscp 46 classification
class-map
match ip dscp 47
match ip dscp 40
match mpls experimental 5

class-map match-any lbe_out

LBE
match ip dscp 8 classification
class-map
match mpls experimental 1
 Configuration examples
policy-map QoS_out
QoS policy
class ip_premium_out
definition
priority policy-map
class lbe_out
bandwidth percent 1
class class-default
exit
exit
Apply
interface POS 0/1
service policy
service policy output QoS_out to an interface
 Presentation Outline
The need for QoS
QoS Metrics
QoS Architectures
IPv6 header & QoS
Configuration Examples
Performance Tests in IPv6 environments
Conclusions
IPv6 QoS Performance Tests
Objectives
Perform diverse set of tests on hardware- and
software-based routers
Validate the performance of basic QoS
mechanisms with IPv6 traffic and identify missing
functionality
Test environment
IPv6-only 6NET research network
Dual stack GRNET production network
 Services in 6NET & GRNET
Premium IP (PIP)
Best Effort (BE)
Less than Best Effort (LBE)
 6NET Network
Gain experience of IPv6
deployment. Extensively test a
variety of new IPv6 services and
applications.
Technical specifications
IPv6 only network!
PoPs in sixteen European
countries.
STM-1 core links, up to
1Gbps access links.
Cisco GSR12400 series
routers in the core and 7200
series routers in the access.

http://www.6net.org
 6NET test network

Software-based traffic generators iperf, mgen tools

Qualitative tests - Validate that PIP traffic experience better
services than other traffic
Packet loss for PIP traffic (UDP)
Use mixture of PIP and BE
traffic.
Gradually increase PIP traffic
(UDP). 10

Create different levels of 80Mbps 120Mbps

congestion with 80Mbps and 8

P a c k e t los s (% )
120Mbps background BE
traffic. 6

Activate policing at 5% of the 4

core links (STM-1).
Results 2

Almost zero packet loss in both

0
congestion scenarios. 1 2 3 4 5 6 7 8
Classification separates traffic Transmitted traffic (Mbps)
into different queues
Priority queues protect
Premium IP traffic.
Policing drops exceeding
Premium IP traffic.
 Jitter for PIP traffic (UDP)
Use mixture of PIP and BE traffic.
Gradually increase PIP traffic
(UDP). 8
Create different levels of 80Mbps 120Mbps
congestion with 80Mbps and 7

120Mbps background BE traffic.

J itte r (m s e c )
6
Activate policing at 5% of the
core links (STM-1). 5

Results
4
Jitter remains the same in both
congestion scenarios. PIP traffic 3
in priority queue is not affected 1 2 3 4 5 6 7 8

from BE traffic. Throughput (Mbps)

Jitter is reduced as PIP rate is

increased. A higher transmission
rate leads to packets arriving
closer together at the destination.
 GRNET Network
GRNET is the Greek National
Research and Education
Network (NREN).
Technical specifications
Dual stack network!
12 PoPs in major cities.
STM-16 core links, up to 1Gbps
access links.
Cisco GSR12400 series with
4xGE (Eng3) and 10xGE
(Eng4+) line cards.
5Gbps connection to upstream
provider (GANT).
 GRNET2 Testbed

Hardware-based traffic generators Smartbit 600.

Collect accurate time-related statistics.
 CPU Impact
Generate IP traffic at 2Gbps
for a 30min period
Use IPv4, IPv6, and mixture
of IPv4/6 traffic.
Use Eng3 and Eng4+ cards.
500Mbps IPv4 production
traffic 12% average load.
Results
No impact with IPv4 traffic.
7% absolute increase with
mixture of IPv4 & IPv6
traffic.
11%(26%) absolute
increase for 5min (1min)
intervals for IPv6-only traffic.
Routing problem impacted
network connectivity.
 Packet loss for BE traffic
in Eng3/Eng4+ cards
Create bidirectional IPv6 and
IPv4 flows.
Gradually increase traffic 80
load.
Use Eng3 and Eng4+ cards.
60
Results

Pac k e t los s (% )
Different IPv6 switching 40
capabilities for Eng3
(hardware-based) and
Eng4+ (software-based) 20
GigEthernet cards.
IPv6 and IPv4 traffic 0
experience the same packet 10% 25% 40% 55% 70% 85% 100%
loss in Eng3 card (direction Load (%)
1->2). IPv6: 1 -> 2 IPv4: 1 -> 2 IPv6: 2 -> 1 IPv4: 2 -> 1
IPv4 packet loss in Eng4+
card (direction 2->1)
Latency and packet loss for PIP traffic
in Eng3/Eng4+ cards
Create bidirectional IPv6 15,000

and IPv4 flows.

Latency (usec)
2% of load is PIP traffic. 10,000

Increase load gradually. 5,000

Results
Latency for PIP traffic in Eng3 0
10% 25% 40% 55% 70% 85% 100%
card is very low provided zero Load (%)

packet loss (<85% load). When PIP IPv6: 1 -> 2

IPv6: 1 -> 2
PIP IPv4: 1 -> 2
IPv4: 1 -> 2
PIP IPv6: 2 -> 1
IPv6: 2 -> 1
PIP IPv4: 2 -> 1
IPv4: 2 -> 1
there is packet loss (100% load),
PIP latency is increased but still 45

remains thousand times smaller

than (IPv4/6) BE latency. PIP

Latency (usec)
40

traffic experiences no packet

loss. 35

PIP IPv6 experiences higher

packet loss and latency than PIP 30

IPv4 in Eng4+ card. Probably, 10% 25% 40% 55% 70% 85% 100%
Load (%)
PIP IPv6 traffic is handled as BE. PIP IPv6: 1 -> 2 PIP IPv4: 1 -> 2 IPv6: 1 -> 2 IPv4: 1 -> 2
 Presentation Outline
The need for QoS
QoS Metrics
QoS Architectures
IPv6 header & QoS
Configuration Examples
Performance Tests in IPv6 environments
Conclusions
 Conclusions
The IPv6 protocol, in terms of QoS support, is neither
superior nor inferior to IPv4 counterpart. However, the
flow label field in the IPv6 header is expected to ease
provision of services in the future.
Routers under test allowed the definition of a common
QoS policy for IPv6 and IPv4 traffic. This simplifies the
delivery of QoS in production networks.
New hardware / software does not do impose limitations
in the support of IPv6 QoS
Achieved performance for IPv6/v4 traffic is identical.
Old hardware / software may either lacks some pieces
of functionality or provide lower level services to IPv6
compared to IPv4 traffic.
Testing is needed.
 Bibliography
Miras D. et al., A Survey of Network QoS Needs of Advance Internet
Applications, Internet2 Working Group, 2002
ITU-T Rec Y.1451, Network Performance Objectives for IP Based Services, May
2002.
Gozdecki J. et al., Quality of Service Terminology in IP Networks, IEEE
Communications Magazine, March 2003
Crawley E. et al., A Framework for QoS-Based Routing in the Internet, IETF
RFC2386, August 1998.
IETF IP Performance Metrics (IPPM) WG,
http://www.ietf.org/html.charters/ippm-charter.html
ITU-T Rec. Y.1540, IP Packet Transfer and Availability Performance
Parameters, December 2002.
Braden R., D. Clark, and S. Shenker. Integrated Services in the Internet
Architecture: An Overview. RFC 1633, Internet Engineering Task Force, June
1994.
Wroclawski, J., "The Use of RSVP with IETF Integrated Services", RFC2210,
September 1997.
S. Shenker, C. Partridge, R. Guerin, Specification of the Guaranteed Quality of
Service, RFC 2212, September 1997.
J. Wroclawski, Specification of the Controlled-Load Network Element Service,
RFC 2211, September 1997.
 Bibliography
J. Wroclawski, Specification of the Controlled-Load Network Element Service,
RFC 2211, September 1997.
Steven Blake, David Black, Mark Carlson, Elwyn Davies, Zheng Wang, and
Walter Weiss. An Architecture for Differentiated Services RFC2475, December
1998.
B. Davie et al., An expedited forwarding PHB, RFC 3246, March 2002.
J. Heinanen, F. Baker, W. Weiss, and J. Wroclawski, Assured forwarding PHB,
RFC 2597, 1999.
Rudolf R. et al. IP QoS Across Multiple Management Domains: Practical
Experiences from Pan-European Experiments, IEEE Communications Magazine,
January 2003.
Liakopoulos Ath. et al. QoS Experiences in native IPv6 GRNET and 6NET
networks International Conference on Telecommunication Systems, Modeling
and Analysis 2005 (ICTSM2005).
S.Deering, R.Hinden, Intenet Protocol, Version 6 (IPv6), RFC2460, December
1998.
J.Rayahalme et. al., IPv6 Flow Label Specification, RFC3697, March 2004.
 Revision Questions!
What are the difference related to QoS
between the IPv6 and IPv4 headers? Is there
any improvement in the IPv6 and why?
Shall we expect different performance
guarantees for IPv6 and IPv4 traffic? Under
which conditions?
Is there any functionality limitations or security
consideration in the deployment QoS services
in a production network?