Top Banner
Scottsdale 2010 Broadcast-Quality Video over 802.11n Partha Narasimhan Aruba Networks [email protected]
9
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Airheads scottsdale 2010   broadcast quality video over 11n

Aruba Networks CONFIDENTIAL. © 2010 All Rights Reserved.

Scottsdale 2010!

Broadcast-Quality Video over 802.11n Partha Narasimhan Aruba Networks [email protected]

Page 2: Airheads scottsdale 2010   broadcast quality video over 11n

Aruba Networks CONFIDENTIAL. © 2010 All Rights Reserved.

•  Stored video content –  On-demand video, Youtube, etc. –  Typically Unicast to interested user –  UDP or TCP

•  Live Video Content –  IPTV, Video stream of live events (for example, company meeting) –  Typically Multicast/UDP

•  Video surveillance –  Live content, but typically headed to central monitoring station –  Mostly unicast, but multicast is sometimes used

•  Video coding means not all video packets are equal

Video Basics

Page 3: Airheads scottsdale 2010   broadcast quality video over 11n

Aruba Networks CONFIDENTIAL. © 2010 All Rights Reserved.

•  Reliability –  No feedback mechanism from stations to AP –  No ACKs, so no retransmissions –  Packet loss is unavoidable

•  Capacity –  Typically sent at lowest PHY rate –  PHY rate optimizations provide marginal gain –  802.11n PHY rate gains do not apply

•  QoS –  Multicast packets do not use WMM –  Sent in best-effort

•  Power-save –  Buffering of packets due to PS clients can cause latency and jitter –  PS clients may not be interested in any multicast streams

Multicast Video over 802.11

Page 4: Airheads scottsdale 2010   broadcast quality video over 11n

Aruba Networks CONFIDENTIAL. © 2010 All Rights Reserved.

•  Disable use of multicast for video delivery on 802.11 –  Causes video server scalability problems –  May require rewriting video applications –  Not very promising

•  Limit number of available multicast streams/channels –  Not desirable, which channels should be made available? –  Reliability problems still persisit

•  Use a wire –  Not an option

•  802.11 unicast delivery of multicast packets –  Solves 802.11 multicast delivery problems –  Buys reliability in exchange for capacity, but when do we choose

one over the other?

Possible solutions

Page 5: Airheads scottsdale 2010   broadcast quality video over 11n

Aruba Networks CONFIDENTIAL. © 2010 All Rights Reserved.

•  Unicast delivery at 802.11 MAC –  No changes at L3 and above required –  No modifications required at servers and applications

•  Unicast delivery only to users subscribed to that multicast stream –  Unicast delivery is derived from IGMP state at the controller

•  Dynamic switching between unicast and multicast delivery –  Unicast/multicast mode selection is independent for each multicast

stream on each VAP –  Switching between modes is determined by user-configurable

thresholds

Dynamic Multicast Optimization

Page 6: Airheads scottsdale 2010   broadcast quality video over 11n

Aruba Networks CONFIDENTIAL. © 2010 All Rights Reserved.

•  Additional load on the controller –  Currently DMO is supported only on the 3000 series and M3

controllers –  Network design should factor in multicast video usage

•  Additional load on the intermediate network –  May not be that much of a factor on most campus networks –  Presence of slow links between controller and AP impacts

performance

•  Additional load on wireless channel –  Different load factors for 11n vs 11abg clients

•  Video server tuning –  Video bit rate, IP fragmentation

Network Planning considerations Video Server

Controller

AP1

AP2

User1

Page 7: Airheads scottsdale 2010   broadcast quality video over 11n

Aruba Networks CONFIDENTIAL. © 2010 All Rights Reserved.

•  Video servers –  Video Furnace –  Microsoft Media Server –  VideoLAN server

•  Multiple video streams/channels –  Different video bit rates –  Most tests were with SD videos, not much HD testing yet

•  Mix of 11n and 11abg clients –  20 clients on a single AP tested with multiple 2 Mbps video streams

•  Packet error rate with multicast delivery is around 1-3% –  Significant impact on user experience

•  Unicast delivery can get PER very close to 0 –  User experience is superior at all video rates

Validation

Page 8: Airheads scottsdale 2010   broadcast quality video over 11n

Aruba Networks CONFIDENTIAL. © 2010 All Rights Reserved.

•  Current DMO thresholds are based on user count –  All users are not identical –  3:1 load ratio between 11abg and 11n is too simple

•  Migrate DMO thresholds to use channel time –  More accurate measure of channel load –  Load due to different PHY rates can be measured accurately –  Works across client types, do not need a ratio –  DMO threshold planning becomes simpler

•  Expand DMO decisions to both controllers and APs –  Controllers switch between delivery modes today –  DMO switching at AP may simplify network planning for DMO –  Provides finer coupling between airtime accounting and DMO

decisions

The Road Ahead

Page 9: Airheads scottsdale 2010   broadcast quality video over 11n

Aruba Networks CONFIDENTIAL. © 2010 All Rights Reserved.

•  Not all video frames are equal –  Video coding results in uneven distribution of information across

video frames –  Losing or delaying some video frames can cause a higher impact on

user experience –  Look deeper into IP packets in a video session to enable

differentiated handling of video frames

•  Load balancing and DMO –  Move clients between APs to increase the probability of unicast

delivery to more clients across more APs

•  Optimize multicast flows within wired network

The Road Ahead