Date post: | 03-Jan-2016 |
Category: |
Documents |
Upload: | evan-griffith |
View: | 213 times |
Download: | 0 times |
Global Multimedia Collaboration System
Wenjun Wu
Indiana University Bloomington IN 47401
http://www.globalmmcs.org
Outline Service-Oriented Collaboration Current main stream real-time collaboration
technologies• Videoconferencing: H.323, SIP, Access Grid
• Instant messaging & VoIP: MSN/Aol/Yahoo, Jabber, Skype Global-MMCS introduction
Service Oriented Collaboration Collaboration has
a) Mechanism to set up members (people, devices) of a “collaborative sessions”
b) Shared generic tools such as text chat, white boards, audio-video conferencing
c) Shared applications such as Web Pages, PowerPoint, Visualization, maps, (medical) instruments ….
b) and c) are “just shared objects” where objects could be regarded as Web Services
We can port objects to Web Services and build a general approach for making Web services collaborative
a) is a “Session Service” which is set up in many different ways
Shared Event in Collaboration All collaboration is about sharing events defining state changes
• Audio/Video conferencing shares events specifying in compressed form audio or video
• Shared display shares events corresponding to change in pixels of a frame buffer
• Instant Messengers share updates to text message streams
• Microsoft events for shared PowerPoint (file replicated between clients) as in Access Grid
Using Web services allows one to expose update events of all kinds as message streams
Need publish/subscribe approach to share messages
Session Service in Collaboration Membership: Participant list Role & floor assignment
• Management policy based on shared objects ( audio/video, text, whiteboard, game
Ad-hoc or formally schedule Session is also a “shared meta object” associated with
shared objects • It needs eventing to keep consistent state as well
Multimedia Streaming Service
Streaming-In Stream-Out filter process media “events” between a stream
source and stream sink; and can be shared composite media service is a DAG ( directed
acyclic graph) common case ~ a filter chain QoS(in) QoS(out) ( bandwidth, delay,
jitter,loss) for example: video/audio mixing, transcoding,
video-audio synchronization
H.323 Introduction Major audio-video standard but broader “Binary” format for both “data” and “control” Supported by many commercial vendors and used
throughout the world in commercial and educational markets
Supports small-scale multipoint conferences Has conference management functionality and the call
signaling functionality H.225 ~ call set-up H.245 ~ call control H.243 ~ Audio/Video multipoint control T.120 ~ Data Collaboration
H.323 Protocols H.323 is a “framework” document that describes
how the various pieces fit together H.225.0 defines the call signaling and
communication between endpoints (Call Signaling) and the Gatekeeper (RAS)
Annex G/H.225.0 defines communication between Border Elements
H.245/H.243 is the conference control protocol
T.120 is the data conference protocol
Typical H.323 StackH.323
IP
UDP
RTP
RTCP
TCP/UDP TCP UDPUDP TCP
Audio
Codecs
G.711
G.723.1
G.729
..
Video
Codecs
H.261
H.263
H.264
..V.150 T.120
TCP/UDP
T.38
H.225.0
Call
Signaling
H.245H.225.0
RAS
Terminal Control and ManagementData
ApplicationsMedia Control
Multimedia Applications, User Interface
http://www.packetizer.com
H.323 Architecture
Gatekeeper(security, QoS, routing etc.)
MC MPMCU
H.323 Terminal 1
H.323 Terminal 2
H.323 Terminal N
....
Packet Switch Network
SIP Initially SIP was designed to solve problems for IP
telephony. SIP basic functions
• user location resolution, • capability negotiation• call management. equivalent to the service H.225 and point to point part of H.245
The major difference from H.323• SIP was designed in a text format and took request-response
protocol style like HTTP• SIP doesn’t define the conference control procedures like
multipoint parts of H.245 and T.120.
rtspd
Quick-time
GatekeeperSIPUA
SIP
H.323
RTSP
sipd
sipconfsipconf
sipumsipum
sip323sip323SIP-H.323
signaling gateway
Conferencing
Programmable SIP servers
Unified messaging
Streaming media
Hardware SIP phone
Desktop SIP clients
sipgwsipgw
PSTN
MGCPSIP-MGCP gateway
SIP-PSTN gateway
Regular telephones
A Integrated SIP Service System: CINEMAFrom Columbia University
Sipconf : SIP based Centralized Sipconf : SIP based Centralized conferencingconferencing
sipcsipc
http://www.cs.columbia.edu/~kns10/software/sipconfhttp://www.cs.columbia.edu/~kns10/software/sipconf
SIP323SIP323
SIP/PSTNSIP/PSTN
SIP based conferencing server SIP/SDP and RTP/RTCP Audio mixing Play-out delay algorithm Web based conference setup G.711 A and Mu law, G.721, DVI
ADPCM Multiple simultaneous
conferences
Summary of H.323/SIP Conferencing Systems
Most products are Centralized conferencing system
MCU integrates the service of media processing service and session management
Call-based
A conference call represents control connections between clients and MCUs.
MCU is just a endpoint connected to VoIP softswitch cloud Most vendors offer hardware solutions Thought as services and controllers but specialized protocols and
implementations; NOT Service-Oriented Architectures! Only support small size or medium size meetings ( < 20 sites )
Access Grid Access Grid : a large scale audio/videoconference based on
a multicast network provides the group-to-group collaborations among 150
nodes connected to Internet 2 world wide. Use improved MBONE audiovisual tools VIC and RAT Depends upon high-speed network ( each node needs
20Mbps ) Peer to peer architecture for distribution with centralized
non standard session control (venue server) Did not develop many new capabilities but made existing
public domain software better packaged and easier to use
Proprietary IM MSN/Yahoo/AOL
• Ah-hoc small-group collaboration
• Text, audio-video, gaming and others….
• Remote Presence Service
( typical publish/subscribe messaging application )
• Massive “chat servers” running behind to support millions of users across the world
• Limited size of buddy list and multi-party meeting
• Poor/fair quality for audio/video communication
• Close protocol which is unfriendly to third-party developers
Skype Skype: p2p IM&VoIP solution
gained a big success. • improving sound quality ( use new iLBC audio codec ) from
Global Sound
• Uses a variant of IETF Stun to identify NAT and firewall
• using p2p overlay (Kazaa) rather than expensive, centralized infrastructure.
• provided supplemental features like instant messaging service.
• Free on-net VoIP service and a fee-based off-net SkypeOut service that allows calling to PSTN and cellular phones
• Millions of download and on-line users in the world
Why is Skype so successful? Better voice quality excellent audio codec, fancy echo cancellation algorithm Global IP Sound ( iLBC audio codec ) Ability to work behind firewalls and NAT Ease of use ( quite simple UI ) based on IM metaphor P2P style without centralized MCU any peer that has enough resource can be selected to host
the mixing service limited the number of participants in a conference ( at most
4 which is common for private social meetings ) use p2p overlay to discover resources and route packets
But they are simply not good enough!
Although all of these systems have advantages, they are not sufficient for building more advanced and integrated collaboration systems:
SIP :
had a huge development recently, especially in wireless world very limited supported for conference control H.323 :
AV collaboration and T.120 are not well integrated. the AV communication services and T.120 overlay networks don’t have verygood scalability.H.323 and T.120 are designed in a relative complicated OSI model. It is not easy to
understand and develop in their APIs
Most H.323 and SIP conferencing products are based on centralized MCU And no way to take full use of private MCU resources ( Imagine how to use these private MCUs to create a meeting that have
thousands of participants )
But they are simply not good enough!• Access Grid heavily depends on multicast service and limited number of uni-cast bridge servers in the Internet 2
No way to be deployed in current Internet
• Skype : Most promisinguse its own propriety protocols and can’t interoperate with other legacy VoIP clients such as H.323 and SIPonly support small-scale audio conferencing ( at most 5-party ) and have no video serviceSkype-2 is said to be able to support 10-party in dual-core Intel machines
Above all, no system can support medium / large size meetings in current Internet and adapt different client devices
What’s the ideal videoconferencing system I A unified, scalable, robust “overlay” network is needed to
support AV and data group communication over heterogeneous networking environments• go through firewall and NAT • provide group communication service in whatever unicast and multicast
networks • offer reliable data delivery in whatever loss network• to be configured as P2P or distributed server-based overlay to provide
differential services for VIP and regular users A service-oriented architecture for hosting media processing
service and session control service• More scalable than centralized MCU• Support various style of conferencing ( massive scale of broadcasting as well
as medium size of private social meetings )• Service providers can be highly distributed and p2p ~ Skype p2p audio mixing• Scalable service discovery based on p2p search • Customized media filters for different clients ( PC, PDA, … )
What’s the ideal videoconferencing system II
A core conference control mechanism is required for establishing and managing the multi-point conference• Complete conference control service like T.124 (Generic
Conference Control) in T.120 framework
• more flexible facilities to describe application sessions and entities ( role-based, XML ) for all kinds of collaboration:
audio/video, game, whiteboard
• Session border management
Integrate different AV sessions ( H.323 , SIP, Access Grid, RealStreaming … )
Simply regard these bridging gateways as “add-on services”
Global-MMCS Service Architecture
Media Delivery, Storage Services( QoS : Reliable and Secure Delivery,Transport Mechanism,
Massive Dependable Storage )
Media Processing Service( Adaptation, Mixing,
Transcoding ... )
Session Management Service( Membership, Role
Management, Floor control)
AudiovisualCollaboration
Shared DataApplication
InstantMessaging
XGSP Web Service MCU Architecture
SIP H323 Access Grid Native XGSPAdmire
Gateways convert to uniform XGSP Messaging
High Performance (RTP)and XML/SOAP and ..
Media ServersFilters
Session ServerXGSP-based Control
NaradaBrokeringAll Messaging
Use Multiple Media servers to scale to many codecs and manyversions of audio/video mixing
NB Scales asdistributed
WebServices
NaradaBrokering
Break up into “Services” Monolithic MCU becomes many different “Simple Services”
• Session Control• Thumbnail “image” grabber• Audio Mixer• Video Mixer• Codec Conversion• Helix Real Streaming• PDA Conversion• H323/SIP Session/Signaling Gateways
As independent can replicate particular services as needed• Codec conversion might require 20 services for 20 streams
spread over 5 machines 1000 simultaneous users could require:
• 1 session controller, 1 audio mixer, 10 video mixers, 20 codec converters, 2 PDA converters and 20 NaradaBrokers
Support with a stream optimized Grid Farm in the sky• Future billion way “Video over IP” serving 3G Phones and home media
centers/TV’s could require a lot of computing
GlobalMMCS and NaradaBrokering All communication – both control and “binary” codecs are
handled by NaradaBrokering Control uses SOAP and codecs use RTP transport Each stream is regarded as a “topic” for NB Each RTP packet from this stream is regarded as an “event” for
this topic Can use replay and persistency support in NB to support
archiving and late clients Can build customized stream management to administer replay,
and who gets what stream in what codec NaradaBrokering supports unicast and multicast Use firewall penetration and network monitoring services in NB
to improve Q0S
NaradaBrokering
Stream
NB supports messagesand streams
NB role for Grid isSimilar toMPI role for MPP
Queues
Incorporating Support for Audio/Video Delivery into
NaradaBrokering
Added support for an unreliable transport protocol, UDP
Implemented a fixed size (fast) topic (8 bytes). Designed a new compact event with minimum headers. Added support for legacy RTP clients (both unicast
clients and multicast groups) Improved the routing algorithm to handle real-time
audio and video stream delivery.
XGSP Conference Control Architecture
Conference Manager
NaradaBrokering
User 5
User 3User 2
NodeManager
ApplicationInstance 0
ApplicationInstance 1
User 1
App Sessions
ConferenceCalendar
ApplicationRegistry
UserAccounts
XML based General Session ProtocolThe XGSP conference control includes three services: Conference management supports user sign-in, user create/terminate/join/leave/invite-
into XGSP conferences conference calendar service Application session management provides users with the service for creating/terminating
application sessions, managing session related services such as audio/video mixing
Floor control manages the access to shared collaboration resources in
different application sessions for example, in a large scale of meetings having thousands of people, only
limited people are allowed to become presenters so that they can send audio/video
Global-MMCS Community Grid This includes an open source protocol independent Web Service
“MCU” which will scale to an arbitrary number of users and provides support for thousands of simultaneous users of collaboration services.
The function of A/V media server is distributed using NaradaBrokering architecture.• Media Servers mix and convert A/V streams
Open XGSP MCU based on the following open source projects• openh323 is basis of H323 Gateway• NIST SIP stack is basis of SIP Gateway• NaradaBrokering is open source messaging• Java Media Framework basis of Media Servers• Helix Community http://www.helixcommunity.org for Real
Media http://www.globalmmcs.org open source release
Audio/Video Meeting Tests for single broker
0
5
10
15
20
25
30
35
40
0 500 1000 1500 2000
number of participants
Ave
rag
e la
ten
cy i
n m
s
first userlast user
Average
0
20
40
60
80
100
120
0 500 1000 1500
number of participant
Ave
rag
e L
aten
cy i
n m
s
f irst
last
average
Audio Meeting Tests Video Meeting Tests
Distributed Brokers Tests
Machine 1
Broker 1Broker 2Broker 3Broker 4
VideoTransmitter
VideoReceivers
VideoReceivers
VideoReceivers
VideoReceivers
MeasuringReceivers
Linux Cluster 1Linux Cluster 2
0
30
60
90
120
150
180
0 200 400 600 800 1000
Number of receivers
Avr
g. L
aten
cy in
ms
broker1
broker2
broker3
broker4
• Going through multiple brokers does not introduce considerable overhead. • Scalability of the system can be increased almost linearly by adding new brokers
1
10
100
0 20 40 60 80 100 120 140 160
La
ten
cy (
Mill
ise
con
ds)
Number of Users per Site
Average Latencies for Video Conferencing Clients at different locations. Sites in Indiana, Florida, New York and Cardiff
IndianaNew York
FloridaCardiff UK
Analysis of the broker network’s performance
Test results showed that the broker network can scale well for both single large size meetings and multiple smaller size meetings.
In large size meetings, the capacity of the broker network is increased with respect to the capacity of the added brokers.
In multiple smaller size meetings, the distribution of users among brokers are important. Inter-broker stream exchange can reduce the scalability. Few users should not be scattered around the broker network.
In wide area networks, this videoconferencing system provides many benefits with distributed broker architecture: bandwidth savings, latency savings, and better quality services.
In summary, thousands of concurrent users can easily be supported in distributed broker settings.
media services computation overhead
Media Services Computation Overhead
Audio Mixing 46% while 20 audio mixers ( six active speakers ) are running
Video Mixing 94% while 4 video mixers ( 4-way mixing ) are running
Image Grabber 70% while 50 image grabbers are running
Real Streaming Producer
90% while 4, 23fps stream producers are running
Improved JMF Performance VIC Old JMF Client Fast JMF Client
1 8% - 9% 15% - 16 % 6% - 7%
2 13% - 14% 24% - 25 % 9% - 10%
3 17% - 18% 33% - 34 % 15% - 16%
4 23% - 24% 40% - 41% 17% - 18%
5 26% - 27% 46% - 47% 23% -24%
6 32% - 33% 51% - 52% 27% - 28%
7 35% - 36% 58% - 59% 31% - 32%
8 40% - 41% 62% - 63% 34% - 35%
Fraction of CPU used versus number of received streamsThe CIF-size video sequence from a 30-second movie with a lot of motions is streamed to the clients. Each stream is encoded in H.261, and has average bandwidth of 400Kbps and 20 fps.
MPEG-4 vs. H.261
We added MPEG4 video to Java Media Framework Higher quality and flexible video sizes including
distributed pixels
Coupled Diverse Streams GlobalMMCS supports many diverse streams managed by “video
system”• Different audio and video codecs• Shared display using video codecs (MPEG4 or H261)• Motion JPEG – stream of images to and from PDA
NaradaBrokering represents these and other collaborative streams just a “topics”; collaboration from multiple clients subscribing to a topic• Text Chat• Traditional lossless codec based shared display• White boards• Control streams
Streams can be linked to provide composite topics• eSports project linking video streams and real time annotation
of any frame• Can rewind and choose any frame of a real-time stream
eSports SnapshotMaster Video AnnotationWhiteboard
Collaborative Video AnnotationWhiteboard
Synchronized Replay of archived video and annotation
GlobalMMCS Status/Futures I 1. New Collaboration tools
• Shared IDL (Visualization), PowerPoint, OpenOffice (Applications need a month or so more)
• SVG game ( stable )
• Whiteboard ( stable )
• e-Sport ( prototype)
• Jabber IM client ( prototype)
• XGSP needs extension to support 2. JMF Audio/Video client ( stable)
• performance enhancement finished
• new codec ( MPEG4-DivX finished; try MPEG4-Xvid and H.264)
• support different platform ( Linux, Mac – Mac well developed but need to chase bug(s) )
• support NAT/firewall transparently like Skype
• Google Desktop PlugIn (under development )
GlobalMMCS Status/Futures II 3. Replay & Archive (prototype)
• Replay Engine based on NaradaBroker Storage Service
• XGSP-RTSP gateway
• Extend RTSP and NaradaBrokering for Instant Replay 4. Web Server Portal ( stable)
• Standard calendar service ( iCalendar, vCalendar)
• Flexible conference management
• Need to package UI’s as portlets 5. Conferencing Media Processing Service ( Stable)
• Support new codec (H.264 )
• Add DCT domain MPEG4-H.261 transcoder 6. H.323 Gateway ( Stable)
• Import it to Linux platform
GlobalMMCS Status/Futures III 7. RealStreaming Gateway ( Stable )
• Import it to Linux• Support Mobile clients
8. Global-MMCS deployment & test• Core performance measurements complete• Test under the setting of multiple NaradaBroker and
NAT/Firewall• support deployment for AFRL, NASA, DOE portals• test with remote sites
9. SmartPhone Clients (prototype) 10. Improved video codec-based shared display 11. Scheduler of dynamic services sensitive to streaming
bandwidth requirement as well as CPU use of codec conversion