QoS for Media Networks

Cisco Confidential 1 © 2011 Cisco and/or its affiliates. All rights reserved.

Amine Choukir, Abderrahim Maroufi

June 2013

© 2011 Cisco and/or its affiliates. All rights reserved. Cisco Confidential 2

•  Cisco & Innovation

•  Medianet: Architecture for pervasive media experience •  Problem Statement

•  Today (What are we executing on)

•  Tomorrow •  Academia

•  Prototype

•  Product status

•  Conclusion


Worldwide leader in networking that transforms how people connect, communicate and collaborate

70,000 partners

90% of world’s internet data travels on Cisco gear

63,000+ employees

470+ offices

$40+ Billion sales

165+ countries

12,000 patents

22,000 CCIE’s

$5.8 Billion Annual R&D budget



Innovation Customer Driven

Market Transitions


2.5% 13.5% 34% 34% 16%

Area under the curve is number of

customers



R&D: $ 5-ish B 24,000 employees

Partner w/ Other Companies

Microsoft, SAP, Citrix VMWare, PacketAction …

Design New Products

CRS-1, IOS XR, Nexus, UCS

Spin in Investments

Andiamo Nuova Impresa

Pursue Acquisitions

180+ Acquisitions

Start New Business Models

WebEX Cloud Wireless for SMB Wireless SP wholesale

Develop New Technology

TelePresence, IP Telephony Converged Wireline/Wireless Medianet Cisco ONE






•  Prototype


•  Conclusion


CAGR: Compound Annual Growth Rate


Coding

Transmission

Decode


15 fps

30 fps

I Frame 1024–1518

Bytes

I Frame 1024–1518

Bytes

P and B Frames 128–256 Bytes

600 Kbps

32 Kbps

Variability of Video Coders

I Frame Complete Frame Encoded

P Frame Ball Encoded with Motion Vector from I frame

B Frame Only Motion Encoded Ball Bi-directionally from I & P


•  A compressed video is a series of bytes, so the first step is to split it into packets.

•  A video stream is then often coupled with audio that has to be synchronized (and sometimes subtitles and other info are added).

•  Typical streaming techniques are: IP/UDP/TS IP/UDP/RTP

IP/UDP/SRTP IP/UDP/TS/RTP

•  But you can come across something like that: IP/TCP/RTP

IP/TCP/HTTPS/RTP…

•  Recently ABR is getting traction (http)

•  NAT, buffering, real-time vs. recorded


Traffic Profiles and Requirements

Voice

§  Smooth §  Benign §  Drop sensitive §  Delay sensitive §  UDP priority

Video-Conf

§  Bursty §  Greedy §  Drop sensitive §  Delay sensitive §  UDP priority

Surveillance

§  Bursty §  Drop sensitive §  Delay sensitive §  Jitter sensitive

Media Streaming

§  Bursty §  Drop sensitive §  Delay and Jitter

insensitive

Data

§  Smooth/bursty §  Benign/greedy §  Drop insensitive §  Delay insensitive §  TCP retransmits

TelePresence

§  Bursty §  Drop sensitive §  Delay sensitive §  Jitter sensitive §  UDP priority


•  Massive (exclusive) use of IP protocol also for voice and video!

•  IP was not designed for multimedia

•  Examples of challenges and real problems for multimedia networks:

IPTV

§ Audio/Video sync (lip sync)

§ Channel change time

§  Latency

§ Scale

Enterprise networks

§ Quality of Service

§ Bandwidth Reservation

§ Authentication

§ Security






•  Prototype


•  Conclusion


•  Service quality

•  Ability to provide different priorities to different applications, users, or data flows

•  Guarantee a certain level of performance to a data flow. E.g. bit rate, delay, jitter…

•  QoS refers to the capability of a network to provide better service to selected network traffic

Resource Reservation DSCP

Control over resources

Access Control QoE

Most of the rest of the talk will focus on solutions for improved QoS on media networks


Video Solutions

Optimize User Experience

Video Monitoring

Media Optimization

Media Aware Routing

Auto-configuration

Enable Video Solutions

Video Demand Acceleration § Collaboration, security, information § Any device, anywhere § Optimized experience

IT Efficiency § Multiple applications § Network optimization § Resource control § Visibility and management

Medianet Service Interface APIs


Identify Media Classify Schedule Provision/ Resource Control

Monitor, Troubleshoot,

Optimize

Network Management

WA

N

Flow metadata Media Monitoring

EF

AF41

Current Medianet Features

Metadata


•  Metadata is set of attributes, describes the properties of traffic flow. Example: application identification (e.g. the flow is a webex-meeting stream)

codec information (e.g. the flow is an RTP video with 90KHz clock frequency)

•  Metadata can be produced by: Application Network Proxies

•  Once produced, Metadata is transported via flow path towards destination

•  Metadata is then consumed by various services in the network.


1. Operator configures QoS policy on router Define a traffic class “telepresence” (class-‐map with match applica,on telepresence) Configure preferential treatment for this class (policy-‐map with DSCP setting or queuing)

2. Application signals metadata through MSI [flow = <e1,p1,e2,p2, udp> , application-name = ”telepresence”]

3. Metadata arrives at the router Program the data plane: flow <e1,p1,e2,p2, udp> is telepresence class

4. Application starts sending media traffic Media flow receives preferential treatment at the router

1.

2. 3.

4.

Media Monitoring


•  Hard to pinpoint root cause

•  What path?

•  Quantify quality (what is good/bad video quality?)

Quality Issues: Resolution Artifacts Frozen Frame …

CTMS

ISP B

ISP A

Firewall Router

Router L3Switch

Router

Router

L3Switch

L2Switch

L3Switch

L3Switch

Router

Router

L2Switch

WAAS

L3Switch


CTMS

ISP B

ISP A

Firewall Router

Router L3Switch

Router

Router

L3Switch

L2Switch

L3Switch

L3Switch

Router

Router

L2Switch

WAAS

L3Switch

§  For TCP, RTP provides: Fault-isolation, problem ownership assignment Accelerated troubleshooting SLA validation

§  Identifies and measures user traffic on routers, switches and endpoints network contribution (loss, jitter) to media stream Applied on interface: inbound and/or outbound


Media Flow

Mediatrace Initiator & local Responder

Mediatrace Responder



PM

PM PM

PM

1. Configure 2. Pull 3. Clean-up 4. Teardown






•  Prototype


•  Conclusion


•  Main Goal: Improve QoE of media delivery

•  Initial focus: Multiparty video conference (more later)

•  Map Network Metric to QoE

•  Inference and resource estimation Effective use of feedback from network (PerfMon, Mediatrace)

•  Multipath streaming Based on network feedback, choose the appropriate path for media (Metadata) Other forms of coding (e.g., SVC) to be considered

•  Network processing Network nodes can help coders (typically at endpoint) with basic tasks E.g., combining flows (network coding)


Preliminary ideas:

•  Switching is more scalable, but how to adjust the sending rates?

•  Several “helpers”, not only one server. Interesting problem: how are packets relayed

•  SVC relevant for switching. Metadata can save effort at measuring. PfR could help, e.g., to relay traffic among helpers

•  Can helpers do network coding?

•  How can network feedback (Perfmon, MT) be used here?

Conference Server

<active speaker>

Conference Server

<active speaker>

Transcoding Switching


Sending Rate

Congestion Feedback






•  Prototype


•  Conclusion


•  Android

•  ABR streaming (http streaming)

•  WebRTC


What are your ideas?


•  Medianet: www.cisco.com/go/medianet

•  IoE: http://www.cisco.com/web/tomorrow-starts-here/index.html

•  Videos: http://youtu.be/nWqzn2Eri7U http://youtu.be/TcDkDoQSrVA http://youtu.be/5FSmkKXNxq8 http://youtu.be/BJSjbttGaVM






•  Prototype


•  Conclusion


COLLABORATION

•  Connect people not endpoints

•  Conversations not calls

•  Familiar, intuitive, natural

•  Seamless software & hardware integration

•  Any media, any device to anyone at any time


Wave 2: COLLABORATION

Wave 3: BUSINESS INTEGRATION

Wave 1: COMMUNICATION

VIDEO INTEGRATED INTO THE WAY WE WORK

Video A Business Critical Application

Video Over IP Networks

Moving Video to Mainstream


Mobile

Executive

Contact Center

Desk-less InfoWorker

Cisco Unified Communications

Manager COLLABORATION


SCHEDULE

Voice, Video and Content

From within Email

Click to Join, One Button to Push

LAUNCH

MEET

End-to-End Security

1

2

3

Ease of Scheduling

Launch from Anywhere

Two-way video and content


•  Automatically identifies, prioritizes and protects

•  Vastly expanded portfolio of supported endpoints

•  New metadata and media monitoring capabilities

•  Now supported in flagship switching and routing platforms

•  Using network intelligence to put a “crystal ball” for video in the hands of IT managers globally

Available Q1 CY13


Medianet

Environments End User Business and Customer IT Administrator

Platform Services

Activity Stream and Social Graph Content Management

IM and Presence Directory and Identity Scheduling and Calendaring

Voice and Video

Delivery Model Cisco Hosted On Premises Partner Hosted

Experience






•  Prototype


•  Conclusion


Thank you.

C97-666590-00 © 2011 Cisco and/or its affiliates. All rights reserved. Cisco Confidential 46 46

C97-666590-00 © 2011 Cisco and/or its affiliates. All rights reserved. Cisco Confidential 47

1995 First Office in Switzerland

2009 First Server Shipped

1993 First Acquisition Crescendo

2006 TelePresence Launched

1999 First IP Phone

150+

Since 2010 ‘Cool IT’ ranking Top 3

2011/12 Cisco Switzerland Ranked #1

2007 WebEx Acquisition

2002 Andiamo Acquisition – developer of intelligent storage, switching products

2003 Acquisition 2012 Swisscom sustainability award

to Cisco

Acquisitions over 19 years

1984 Company Founded

1984 1993 1995 1999 2002 2003 2006 2007 2009 2010 2012


170 Years Ago: Invention of the Telegraph

100 Years Ago: Invention of the Radio

70 Years Ago: First general purpose electronic computer

40 Years Ago: First Internet connection

20 Years Ago: World Wide Web

Today Intelligently Connecting People, Process, Data, and Things

2003: 0.5B Connected Devices

2005: IP Traffic: 29 exabytes

2005: First Smartphone

2008: Video Traffic: 21 exabytes

2012: 1B Smartphones

2010: 7B Connected Devices 2017:

300M Connected Cars

2010: 0.5B Smartphones

2016: IP Traffic: 1.3 zettabytes

2013: 10B Connected Devices

2012: 50M Connected Cars

2011: 90M Smartmeters

2020: 1B Smartmeters

2020: Global Cloud Market: 500% Growth

2015: 2/3rds Video Traffic

2/3

500%

2020: 4.5B New People, 37B New Things


Internet of Everything

Digitize the World Connecting:

•  People •  Process •  Data •  Things

Intelligent Connections

Connectivity Digitize Access to Information •  Email •  Web Browser •  Search

Immersive Experiences

Digitize Interactions (Business & Social) •  Social •  Mobility •  Cloud •  Video

Networked Economy

Digitize Business Process •  E-commerce •  Digital Supply Chain •  Collaboration

Business and Societal Impact


Our Focus on Integration Provides our Customers with a Competitive Advantage

Our Diverse Product Portfolio Means We’re Here for the Long-Term

Breadth and Depth of Technology Portfolio

62%

#1 Routing:

Edge/Core/Access

50%

#1 TelePresence

55%

#1 Wireless LAN

68%

#1 Switching:

Modular/Fixed

40%

#1 Voice

41%

#1 Web

Conferencing

17%

#3 x86

Blade Servers

44%

#2 Storage:

Area Networks

32%

#1 Security


Mobile | Social | Visual | Virtual

Video Collaboration Data Center/

Virtualization/ Cloud

Architectures for Business

Transformation

Intelligent Network

•  Routing •  Switching •  Services

Leadership in the Core…

Strategic Building Blocks Mobility | Security | Any to Any


Find

Filter

Business Unit Formed

Graduate

Eliminate

Accelerate

Ideas

Incubate Initiate

Customers New Customers Employees Corporate All Business Functions

Fixed time Startup-like team Adoption focus

Market focus Structured BU

Idea è Prototype Market trial Incubate/Initiate Accelerated path


REV

ENU

E G

RO

WTH

TIME

INCUBATE/INITIATE GROWTH MATURITY Patents, Pre-standard

Standardization groups Round A/B funding

Standardization Competitive position

Best practice Testimonials

Customers or Industry requirements

Demo, Prototype

Market share Vertical market solutions

Cust Sat (quality)

Market share Churn/Roadmap Cust Sat (quality)

Service attach rate

Differentiated margins Growth (units & revenue)

Standard margins Top/Bottom line

Standard margins Top/Bottom line

Cost optimization

Next business?


Encoder Packetiser

Encoder Packetiser

Video

Audio

M U L T I P LEXER

Transport Stream

PES

PES

ES

MPEG Compression Layer MPEG Systems Layer+ DVB/ATSC

Programme Specific Information(PSI) or Service Information (SI)

M O D U L A T O R

DVB ATSC

Data

ES

Timing DTS/PTS

System Time Clock (STC)

Packetiser Data PES

Packetiser PES ?

An example for TS streaming


*Note: Latency Here Is Referring to Network Latency


•  Latency ≤ 150 ms •  Jitter ≤ 30 ms •  Loss ≤ 1% One-Way Requirements

Voice

Bandwidth per Call Depends on Codec, Sampling-Rate, and Layer 2 Media

Video-Conf

§  Latency ≤ 150 ms §  Jitter ≤ 30 ms §  Loss ≤ 1% One-Way Requirements

IP/VC Has the Same Requirements as VoIP, but Has Radically Different Traffic Patterns (BW Varies Greatly)

Data

Data Classes: Mission-Critical Apps Transactional/Interactive Apps Bulk Data Apps Best Effort Apps (Default)

Traffic Patterns for Data Vary Among Applications

TelePresence

Telepresence bandwidth varies based on the resolutions and has Radically Different Traffic Patterns

§  Latency* ≤ 200ms §  Jitter ≤ 10 ms §  Loss ≤ 0.05% §  Bandwidth (5 Mbps

per screen @1080P) One-Way Requirements

§  Smooth §  Benign §  Drop sensitive §  Delay sensitive §  UDP priority

§  Bursty §  Greedy §  Drop sensitive §  Delay sensitive §  UDP priority

§  Smooth/bursty §  Benign/greedy §  Drop insensitive §  Delay insensitive §  TCP retransmits

§  Bursty §  Drop sensitive §  Delay sensitive §  Jitter sensitive §  UDP priority



•  Streaming with some additional needs

•  Camera feed is approx 1 – 3Mbps (UDP)

•  Uni-‐cast or, optionally, multi-‐cast

•  Storage uses iSCSI (TCP)

•  Total Bandwidth depends on simultaneous viewings

•  Latency* ≤ 150ms; Jitter ≤ 10 ms; Loss ≤ 0.05%

•  Bandwidth (200K bps to 3.5M bps)

One-‐Way Requirements

•  TrafHic patterns could be engineered

•  Uni-‐cast or, optionally, multi-‐cast

•  Could leverage content networking

•  Total Bandwidth depends on simultaneous viewings

•  Latency. No real constraints could impact experience (i.e. channel change)

•  Jitter ≤ 1s (player will leverage buffering); Loss ≤ 0.05% (for HD); Bandwidth (250Kbps to 4M bps)

One-‐Way Requirements

Surveillance

§  Bursty §  Drop sensitive §  Delay sensitive §  Jitter sensitive

Media Streaming

§  Bursty §  Drop sensitive §  Delay and Jitter

insensitive

Cisco Confidential © 2012 Cisco and/or its affiliates. All rights reserved. 57

Performance Routing

High Packet Loss over this link, let’s see if I can find a better path

Cisco Confidential © 2012 Cisco and/or its affiliates. All rights reserved. 58

Performance Routing

Let’s move Telepresence Traffic to other link


Date post:	15-May-2015
Category:	Technology
Upload:	amine-choukir
View:	531 times
Download:	3 times