SPIN uses standard data networks to transport the
information that composes a conference. Data net works are usually private networks that a user com munity maintains. Such networks often i nclude a number of individual networks joined together by bridges and routers. Unlike publi c telephone net works, wh ich are most frequ ently used for phone calls, private networks are used for a variety of computer data needs, incl ud ing file transfers, remote log.ins, and remote file systems. However,
70
telephone networks often provide the long distance li nes used to make up private wide area data networks.
The use of data networks al lows conferencing data to be treated as woul.d any other type of data.
SPI N requires no special low-level networking pro tocols to transmit its data and uses the transmission control protocol/internet protocol (TCP/IP) or the DECoct protocol . Also, SPIN requires no changes to existing operating systems. When performing the p rototype work t()r the SPI N application, we were not certain whether the real-time nature of confer encing could be accompl ished on inherently non-real-time networks and opera ting systems. Consequently, we developed a special. high- layer synchronization conferencing protocol, cal led the
SPIN protocol, that uses existing data networks.
DECspin: A Networked Desktop Videoconjerencing Application EXIT PICTURE SIZE IN PIXELS AVERAGE FRAME RATE IN FRAMES PER SECOND TOTAL N U M B E R OF CALLS TOTAL NUMBER OF ACTIVE AUDIO CHANNELS TOTAL NUMBER OF ACTIVE VIDEO CHANNELS AVERAGE NETWORK BANDWIDTH USAGE I N MEGABITS PER SECOND
Figure 4 SPIN Status Pop-up Window
This protocol is responsible for the synchronization of audio and video information. The SPIN protocol monitors the flow of data to the network in order to alleviate network congestion when detected. As the network becomes congested , the protocol makes the decision to withhold further video data, since video is the largest consumer of network bandwid th. This withholding of video data is a key feature of the SPIN protocol, because it al lows a conference to vary the video frame rate on a user by-user basis. Thus, video bandwidth can scale to the lesser of either the bandwidth available or the number of frames/s of video bandwidth that a given platform can sustain.
If the withholding of video corrects the network congestion, video d:ua is once again al lowed in the conference. If not, the SPIN protocol delays audio data and stores it in a buffer until the network is able to handle this data. Jf the network outage lasts approximately 10 seconds, audio data is lost. Periods of audio silence are used as a means of recovery from periods of network congestion.
Digital Tee/mica/ Journal Vol. 5 No. 2 Spring 1993
Thus, variable video frame rates along with this treatment of audio data allow for the gracefu l degra dation of a conference as the network becomes busy.
SPIN has been demonstrated over a variety of public and private data networks including
Ethernet (10 Mb/s) ,
FOOl
(100 Mb/s) , Tl ( 1 .5 Mb/s),T3 (45 M b/s) , cable television (10 Mb/s, more cor
rectly, Ethernet running over two 6-megahertz cable television channels), switched multimegabit
data service (SMDS) (1.5 or 45 Mb/s), asynchronous
transfer mode
(ATM)
(150 Mb/s), and frame relay( 1 .5 or 45 Mb/s). Some of these networks are local or metropolitan area technologies, i . e . , local area networks (LANs), whereas others are wide area technologies, i . e . , wide area networks (WANs), as illustrated in Figure 6.
Each type of network provides SPIN with differ ent latency and bandwidth characteristics. SPIN makes corresponding adjustments to a conference to account for these differences and does not require a dedicated bandwidth allocation to carry
Multimedia
SELECT OVERVIEW POP-UP WINDOW
O v e rv i e w
SELECT VI DEO
v Start a C o n fere n c e D e m o n strati on
v DOC UMENTATION
U s e a Feature v Introducti o n
v
Network
vSELECT TEXT
v Solve a
Problem
v DOCUMENTATIONGloss:ary
vB ro w s e
F<�ure 5 SPiN
lnformutiun
Pop-upWindows
on a conference. If a given net work su pports band width a l location, this feature o n l y enhances SPI N 's abi l i t y to del iver video and audio inhnmat ion.
WANs may use a router to i n terco nnect two or more LANs. SPI N has been tested on a n umber of rou ters with mixed resu lts, i.e., some rou ters cor rectly hand le SPI N 's bic.lirec rional traffic pattern whereas others do not. Since some n>u ters clo not correctly hand le bid irectional data traffic w i thout packet loss, wide area routers must be i n d ivid ual ly tested with SPIN to verify proper opera r i u n . Some router problems were traced ru tilt" use of old firmware or software. Consequent] }', SI'I N acted l ike a d iagnosti c tool in poimiug out these prob lems. For example, r u n n i ng t he SPIN appl ication with audio on ly, across Digita l 's private II' net work, y ields varied results. Digi tal's IP network is an exam ple of an open network, with routers from most rou ter vendors. We traced must i nstances of poor SPI N performance to old or obsolete routers (some in serv i ce fo r the last six years without upgrades). These routers usua l ly dropped packets wht:n rou t ing between adjacent Ethernet neL wurks that were only 10 percent buS}'· After these ro urers were
72
upgraded to rhe DECNIS family of routers, the SPIN appl ication fun ctioned correctly, even on con gested networks.
To demonstrate dai ly use of SPIN, we created a metropol itan area network (MAN). Figure 7 shows the network topology, which spanned the states of
New Hampshire and Massachuset ts. The test bed
a l lowed us to demonstrate our FOOl prod ucts, includ ing end-station FDDI adapter cards, multi mode FDDI wiring concentrators, and single-mode FUD! wiring concentrators. SPI N was used in :)0 workstat ions, two of which were attached to large
sc reen projection u n i ts in con.ference rooms.