The con ference qua l i t y achieved when running the SPIN app l ication depends on many factors. The avai l able network band width, the processor speed, the desired fra me- rate specifica tion, the compres sion setti ng, the pictu re size, and how the p ictures are rendered a l l a ffect the qual ity of the confe rence. Ta ble l contains performance data for DECspin ver sion 1 .0 at various combinat ions of settings fo r these factors.
IJECspin: A Networked Desktop Videoconferencing Application
(a) LAN Usage of SPIN
ETH ERNET LAN
ETHERNET LAN
DECNIS 600 ETH E RNET LAN
( b) WAN Usage of SPIN Figure 6 LAN and WAN Usage of SPIN
FDDI
(54 M I LES, SINGLE-MODE F I BER) LITILETON, MA
D
hhhhhh
Figure 7 Digital's MAN Test Bed for SPIN
Multimedia
Table 1 SPIN Performance on a DECstation 5000 Model 200 with DECvideo and DECaudio Ha rdware
Width X Height Frames/s (Bit
( Pixels) Render/Compression Network rate in Mb/s)
256 X 192 Black and White/ No FDDI 10 (4.0)
256 X 192 Black and White/ Yes FDDI 16 (1 .5)
256 X 192 Color/ No FDDI 3 (5.0)
256 X 192 Color/ Yes FDDI 1 0 (4.0)
160 X 1 20 Black and White/ No FDDI 19 (3.0)
160 X 1 20 Black and Wh ite/ Yes FDDI 25 (0.9)
160 X 1 20 Color/ No FDDI 12 (6.0)
160 X 1 20 Color/ Yes FDDI 19 (3.0)
256 X 192 Black and White/ No Ethernet 9 (3.8)
256 X 192 Black and White/ Yes Ethernet 16 (1 .5)
256 X 192 Color/ No Ethernet 2 (3.0)
256 X 192 Color/ Yes Ethernet 8 (3.0)
160 X 1 20 Black and White/ No Ethernet 19 (3.0)
160 X 1 20 Black and White/ Yes Ethernet 25 (0.9)
160 X 1 20 Color/ No Ethernet 8 (5.0)
160 X 1 20 Color/ Yes Ethernet 19 (3.0)
Using a DECNIS Router (Ethernet-to-Router-to-T1-to-Router-to-Ethernet)
256 X 192 Black and White/ No
256 X 192 Black and White/ Yes
256 X 192 Color/ No
256 X 192 Color/ Yes
160 X 1 20 Black and Wh ite/ No
160 X 1 20 Black and White/ Yes
160 X 1 20 Color/ No
160 X 1 20 Color/ Yes
As shown in Table 1 , we tested SPIN performance
using two basic picture sizes: 256 by 192 pixels and 160 by 120 pixels. The tests were performed over both Ethernet and FDDI networks for black-and white and color cases. Also noted in the table is whether or not software compression was enabled for a specific test case. The far right column shows the frame rate achieved for the different combina tions and also summarizes the network bandwidth consumed in each test. The table is presented pri marily to give a sampling of the frame rate and, hence, the level of visual quality achieved for a spe cific combination of parameters. Frame rates affect
an observer's ability to detect change within a sequence of frames. With a slow frame rate, the resulting video sequence may appear choppy and incomplete, whereas a normal frame rate (24 to 30 frames/s) leads to a smoothly varying video sequence with even continuity from one sequence to another. The frame rates in Table 1 below about 6 to 7 frames/s are considered low quality. Those in the 8-to-19-frames/s range are considered good quality, and those in the 20-to-30-frames/s range
74 T1 4 (1.4) T1 1 5 (1.4) T1 1 (1 .4) T1 4 (1 .4) T1 1 0 (1 .4) T1 25 (0.9) T1 3 (1 .4) T1 10 (1 .4)
are high-quality video. The best cases in Table I are
those that used software compression to del iver a pleasing frame rate with the least amount of net work bandwidth consumed together with some degradation of individual frame qual ity. The soft ware compression was tuned to provide nearly the
same frame quality as the uncompressed case.
Table 1 also shows performance data measured
using a DECNIS router. As noted earlier, wide area
usage of SPI N depends on a router with correct algo
rithms for hand ling of bidirectional continuous
stream traffic. The DECNIS family of routers can
supply the ful l Tl bandwidth when presented with
bidirectional SPIN traffic. Other routers on which
SPIN was tested typically del ivered only 25 to 50 percent of the Tl bandwidth. Note that this was only true on the particular routers we tested and that routers other than DECNIS routers may also be able to del iver ful l Tl bandwidth for this particu lar traffic pattern.
Hardware compression technology mentioned in the section Overview of Underlying Hardware and Software reduces the bandwidth requirements for
DECspin: A Networked Desktop Videoconferencing Application
conferencing. Experimentation with motion JPEG
compression (using the Xv extension with com pression fu nctions on an Xvideo frame bu ffer board) has shown that at a resolution of 320 by 240 pixels, true-color frames can be used at 15 to 20 frames/s at a bit rate of j ust u nder 1 .0 Mb/s. This bit rate produces a good- to h igh-quality conference with very low latency. H .261 and MPEG technology resul t in similar frame rates and picture size a t about one-ha l f t h e bandwidth b u t higher overal l l atency. Using motion J PEG as the exa mple, high qual ity conferences require about 1 Mb/s per connection . If a l l conferences are to be high qual ity, this bit rate al lows 1 two-party conference on a T l connection, 5 two-party conferences on an Ethernet segment, and 50 two-party conferences on an FDDI network. Using GIGASWJTCH FDDI
switches, more than 500 two-party conferences can take place simu ltaneously on a network. More users cou ld be supported on T l , E thernet, or
GIGASWITCH networks, if lower-quality confer ences are acceptable.
Conclusion
It became clear d u ring the development and deployment of SPIN that high cost per user l imits the widespread use of the application. The cost of video for DECspin version 1 .0 adds about $8,000 to the price of a workstation. Audio for version 1 .0 adds about $2 ,000 per workstation . These costs, which are prohibitive to most potential users of the technology, do not include the network cost impact.
D igital's Alpha AXP family of computers come with audio input a nd output hardware as part of the base workstation. In spring 1993, Digital released to the Internet com munity a version of DEC:spin that uses this hardware to carry on audio-only confer ences and shows the user a voice waveform instead of a video image. This version eliminates the add-on hardware cost for audioconferencing. A new low cost video option would go far to reduce the add-on cost for video and facil itate a wider use of the SPIN appl ication .
The SPI N appl ication and i ts associated protocol have been demonstrated on D igital and non-Digital compu ters, operating systems, and networks. In particular, SPIN has been shown on SPARC worksta tions run ning Solaris software. Add itionally, SPIN
has been demonstrated on a personal computer using the Microsoft Multimedia Extensions (MM E)
to Windows software. This platform provides a
Digital TecbTtical journal Vol. 5 No. 2 Sjn-ing 199.)
very large user com mun ity of potential SPIN u sers and dramatica l ly drops the price per user compared with the original product. l nreroperabil ity among platforms and a com mon user i nterface give Digital a leadership position in this fast-forming market. Today, h igh-quality conferencing can scale to h undreds of seats on a LAN with lower- quality con ferencing scal ing to larger, more geographical ly dis persed networks. Several factors wi l l lead to the w idespread use of this technology: better and less expensive hardware, program mable codecs, and h igher-speed and Jess-costly cross-country net works. Less-expensive video hardware al lows many users to upgrade their systems to include video, while program m able compn.'ssion technology al lows users to achieve i mprovements in picture quality, compression transcoding, and lower net work needs. Higher-capacity and less-cost ly cross country networks al low more users to access conferencing services. U l timately, even homes wi l l have better computer con nectivity and bandwidth. As these changes occur, and we bel ieve they will, desktop conferencing can become the interactive telephone of the twenty-first century.