Manfred Bogen <email@example.com>
Christian Bonkowski <firstname.lastname@example.org>
Richard Rodriguez-Val <email@example.com>
Clemens Wermelskirchen <firstname.lastname@example.org>
German National Research Center for Information Technology
As companies go international or decentralize, there is an urgent need to complement store-and-forward, multimedia communication via electronic mail or the World Wide Web with real-time collaborative tools that facilitate interaction and the sharing of concepts, resources, and information objects. Desktop video conferencing is supposed to be a step ahead: audio, video, whiteboards, and application sharing look promising because a big part of daily conversation, communication, and cooperation is person-to-person or within small groups with a fixed number of participants. Distinctive products are on the market and it seems to be an easy task to establish a regular, reliable service for everybody in a company who wishes to cooperate and communicate in real time.
It seems natural to use the already existing Internet or the company-internal Intranet as an underlying infrastructure for videoconferencing. The Internet today allows one to download a 30-minute video with full color, full motion, with a screen resolution about equal to NTSC, digitized and compressed into 300 megabytes of data, across a T3 link (45 megabits per second [Mbps]) in 53 seconds and very soon across an ATM infrastructure (up to 622 Mbps) in 15 seconds. It should be powerful enough to bear real-time desktop videoconferencing traffic.
Unfortunately, all this is fiction or plain advertisement. Most existing desktop videoconferencing products do not offer a quality that meets end users' expectations, nor can they be used on a global scale (Intranet/Internet) because they rely on ISDN (Integrated Services Digital Network). For the few products that support Internet technology (TCP/IP), the real network Internet is not performing as expected. Adequate quality and service will only come with bandwidth reservation, which will cost additional money at the very least. Unfortunately, this means there is still a long way to go.
Keywords: distributed applications, computer-supported cooperative, work, groupware, desktop videoconferencing, Internet, Intranet, service engineering, interoperability, quality of value-added services.
Real-time collaboration and synchronous communication are in the center of researchers' and technology developers' interest worldwide, now that electronic mail and the World Wide Web (WWW) have been completely deployed. Industry and economy need videoconferencing to save money. The idea is to take the huge number of meetings, with all the associated circumstances making them time- and money-intensive, and make them more efficient. If traveling is too cumbersome, dangerous (Gulf War), or expensive, and telephone talk is unsatisfactory, videoconferencing is an alternative.
Seeing the people you are talking to is likely to make communication more productive. Even though 90% of all communication is speech-based, images, pictures, and video play an important role in interpersonal relationships and complement and optimize speech by the real-time transmission of moving pictures. This statement is also applicable to the private sector, especially to old or disabled people. Tele-teaching and distance learning can be seen as one special form of videoconferencing.
Figure 1 shows that video conferencing will not entirely replace face-to-face meetings. Basically, a big part of communication among people can be done entirely by e-mail, fax, or computer conferencing. Especially if routine work has to be done, and only some decisions have to be communicated without discussion, multimedia is not needed. As soon as personal involvement is needed, when digital text-based information becomes unsatisfactory, and less formal behavior of communication partners is expected, videoconferencing is the last step before having a face-to-face meeting.
Figure 1: Communication matrix 
The experiences of the early eighties, when special rooms (studios) were used for videoconferences, showed that people do not like to make additional efforts to communicate or cooperate. A regular service, around the clock, as simple as a telephone is needed. Videoconferencing will only be accepted when "business quality," or near TV-quality audio and video, can be achieved.
Besides quality, what are the requirements? Worldwide interoperability ("I don't care about my communication partner's system and I don't know anything about it!"); extensive global and distributed directories with names, numbers, and connect information ("What is the IP address of the reflector?"), in contrast to local address or phone books; and archives ("What price did he offer in our last videoconference?"), to mention only a few.
The German National Research Center for Information Technology (GMD) is a distributed German company with 1,200 employees (Figure 2). For us to have a distributed desktop videoconferencing service is a must. This paper describes our efforts to implement this. Our work is embedded in the international TERENA project "DeViCe" (Desktop Video Conferencing in Europe) . Our emphasis is on multiplatform standard networks, and standard and off-the-shelf equipment for desktop videoconferencing between two people rather than multiperson conferences and broadcast systems.
Figure 2: GMD locations
This paper is structured as follows: Section II describes the most important videoconferencing standards and de facto standards. Section III shows our choice of desktop videoconferencing products and our promising candidates for a regular desktop videoconferencing service. Section IV shows how we finally rated the products in practice. The last section sums up the status of our work and our expectations for the near future.
In a worldwide network with a lot of different products and systems, it is important to have system standards, which describe the way to communicate between different products. Today there exists one recommendation from ITU for desktop videoconferencing over all kinds of networks, H.200 . This is a general framework that includes a lot of other standards. No product refers to H.200 itself. Most refer to one recommendation that works on the specialized network (Table 1).
|Standard||Network||Release date||Title of the standard|
|H.310||ATM||(11/96)||Broad-band and audiovisual communication systems and terminals|
|H.320||ISDN||(03/96)||Narrow-band visual telephone systems and terminal equipment|
|H.321||ATM||(03/96)||Adaptation of H.320 visual telephone terminals to B-ISDN environments|
|H.322||LAN||(03/96)||Visual telephone systems and terminal equipment for local area networks that provide a guaranteed quality of service|
|H.323||LAN||(11/96)||Visual telephone systems and equipment for local area networks that provide a non-guaranteed quality of service|
|H.324||Telephone||(03/96)||Terminal for low-bit-rate multimedia communication|
H.200 also includes a service definition of desktop videoconferencing with a reference to F.730 . F.730 classifies videoconferencing into two main categories: basic services and high-quality services. Most high-quality services are still under development, with the goal being broadcast-like quality. Quality measurements for basic videoconferencing systems are subject to quality and synchronization of video and audio signals and overall system delay.
Figure 3 shows an assessment of the different H.32x recommendations related to video and audio quality. Even though H.320, H.321, H.322, and H.324 can rely on a guaranteed bandwidth, the achievable quality is foreseeable. This is not at all clear for H.323-based conferencing, which has no bandwidth guarantees because it is subject to further investigation.
Figure 3: H.32x quality coverage
T.120 is also important in this context . Like H.200, T.120 refers only to a collection of subsidiary recommendations. T.120 focuses on interoperability on the data channel, thus offering additional features such as whiteboard, file transfer and, in the near future, other options such as application sharing.
Next to these fully defined recommendations exist standards and de facto standards for video codecs. Some products use well-known codecs such as M-JPEG, Indeo, QuickTime, and MBone . On the surface it seems possible to connect two products that use the same codec. But how do you want to connect them? How should a video stream come from one desktop to the other? It is not enough that only parts of a product comply with a standard. Fully implemented standards like H.200 are extremely important.
At GMD, we wanted to have a representative set of products for each quality level and platform, from low-end PCs/Macs to high-end workstations; from cheap black-and-white cameras to high-quality, studiolike equipment. Our quality characteristics were
In order to establish a companywide desktop videoconferencing service, costs could not be neglected. The result should form the basis for further deployment of desktop video in our company. Because the desktop video market had about 100 products available, it was not an easy approach. Sometimes the product specifications were inaccurate, often promising features to be included in future releases. Some products use special hardware to achieve superior results that cannot be compared with those from software-only products. Finally, product availability in Germany was often limited. Talking to vendors, obtaining information, and actually buying products were more difficult than expected.
Product interoperability is the key issue in our heterogeneous multiplatform environment. Unfortunately, the only widely deployed standard is H.320 for videoconferencing over digital isochronous networks (ISDN). Many products claim to be H.320 compliant, but in general they do not guarantee interoperability with particular products or they do not support ISDN at all. It is up to third-party institutions, such as the Telecom Test Center in Stuttgart, Germany, to perform independent H.320 interoperability tests.
In the LAN (local area network) area, releases of H.323-compliant products such as Microsoft NetMeeting, Intel ProShare, and PictureTel LiveLAN are expected in the very near future. However, because of the lack of standardized protocols for desktop videoconferencing on the LAN until now, all currently shipped LAN products use private protocols. Finally, we started with the products shown in Table 2.
|Supplier||Product||Standard||Expected quality-level||Platform||Available in Germany|
|Intel||ProShare Video 200||H.320||middle||Windows 95||yes|
|Intel||Internet Video Phone||H.323||middle||Windows||yes, as beta|
|Microsoft||NetMeeting||H.323||middle||Windows||yes, as beta|
should also use H.320
H.320 together with ERIS
H.320 with a special part
As there is such a variety of different standards and products involved, our results are given by categorizing the typical characteristics, i.e., low versus high network bandwidth needed, or guaranteed versus nonguaranteed bandwidth (QoS).
|Network bandwidth||Nonguaranteed QoS||Guaranteed QoS|
|High Bandwidth > 2 Mbps||Communique!
Intel ProShare Video System 200
PictureTel Live 50/100
|Medium Bandwidth >= 64 kbps||ISDN:
PictureTel Live 50/100
Intel ProShare Video System 200
PictureTel Live 200
|Low Bandwidth < 64 kbps||CU-SeeMe|
There are several products available in this low-level category. We chose Cornell University's well-known and widely used CU-SeeMe as an example of products that use IP as transport protocol and can be used over dial-in connections. We used connections with a minimum speed of 64 kbps (ISDN lines) because lower speeds result in unsatisfactory audio and video quality.
Enhanced CU-SeeMe is the commercial version of CU-SeeMe . Unlike the Cornell version, which is distributed free of charge and offers black-and-white video only, Enhanced CU-SeeMe includes color video, whiteboard, and IP multicast support. CU-SeeMe runs on almost any multimedia PC or Macintosh, there are low-cost video cameras - such as the Connectix QuickCam - available, and it works with TCP/IP across the Internet.
The products we have in house use vendor-specific hardware for audio, video, and connection to ISDN. Compared with low-end solutions that use off-the-shelf sound cards, audio quality greatly benefits from these dedicated solutions. In particular, problems in handling sound volume, managing audio controls, and solving the lip synchronization problem have been reduced. Full-duplex audio is standard. The disadvantage of this hardware is that loudspeakers will pile up on the desk because conferencing audio and standard sound cards are not compatible.
All of the products in this category have proven interoperability with other H.320 products. However, it is quite common to use other audio and video codecs when the same product is used on both sides. For example, Intel ProShare to ProShare uses the Indeo video codec to deliver improved video quality; PictureTel Live 200 to Live 200 uses a proprietary codec to improve the audio quality.
While the Intel product uses special PC boards (ISA and PCI are available) that are not "plug-and-play" and that result in typical PC configuration problems, the ERIS product is a simple-to-connect, external, self-contained black box that is connected through a standard SCSI interface.
We have tested the products using up to two 64 kbps (kilobytes per second) ISDN channels with no problems. When using the ERIS box connected to a Macintosh, conferences between ERIS/Macintosh and PictureTel or ProShare on a PC were no problem.
Early releases of H.323-based desktop video products are becoming available now. This includes Microsoft's NetMeeting Beta 2 and Intel VideoPhone Beta. Both products have some limitations right now, but our preliminary results are encouraging. In particular, Microsoft's NetMeeting, being a software-only, standards-based, full-featured product, will certainly become popular. We have successfully tested its currently supported features, including H.323-interoperability with Intel's VideoPhone and T.120-based collaboration features.
PictureTel's LiveLAN includes dedicated hardware support, and we expect it to become an interesting alternative to the above products. As has been the case throughout our tests, hardware-based solutions generally satisfy higher expectations but at increased costs.
Besides the PC-based products we also tested more traditional videoconferencing products that run on workstations. We expected them to be high-quality products, but our overall results were not satisfying.
We used Sun SPARCstations with Parallax PowerVideo boards (quite expensive) and Digital Alpha workstations with Digital's FullVideo Supreme video board. All machines were connected to Ethernet as well as to our ATM network. Each product uses its own proprietary video and audio protocol, which prevents any kind of interoperation. They all use IP as the underlying network layer. Insoft's (now Netscape) Communique! product and Paradise Simplicity could not be evaluated because of technical problems.
The well-known MBone tools "vic," "vat," and "sdr" have been developed to distribute audio and video on a large scale over the Internet using IP multicast. When used in a high-speed, low-delay network with zero packet loss, it is possible to conduct conferences with good quality. We demonstrated this with the University of Geneva using a 4 Mbps ATM virtual circuit over the European ATM pilot network for lecture transmission between Sankt Augustin and Geneva.
The MMC software was developed by a consortium of vendors under the auspices of the BERKOM project to provide a vendor-independent cross-platform desktop video solution . Unfortunately, MMC is not something to take out of the box, install, and put to work. It greatly depends on the different workstation environments.
MBone and MMC gave excellent results when we used manually configured ATM PVCs with guaranteed bandwidth between the participants.
We tested a version of the MMC desktop video software from Siemens that was especially adapted to ATM, running as a native ATM application without using the IP protocol stack. The result was a videoconference of excellent quality. We did not test the PictureTel Live 50/100 product running over native ATM ourselves. We have seen this product in a commercial environment with very good quality, using a 384 kbps dedicated ATM SVC.
Videoconferences are only acceptable when they guarantee a certain quality for video and audio. The video part should be close to a common TV standard (such as PAL), the audio part should be at least telephone quality. If such quality is not provided, the systems will not be used.
Desktop videoconferencing is real-time communication. A constant good quality (fixed frame rate, no audio gaps, and lip synchrony) can only be guaranteed on a network with a guaranteed QoS. This might be possible on dedicated LANs or idle high-speed links, but in general it requires ATM or ISDN connections. Traditional best-effort networks running IP cannot guarantee satisfying audio and video quality (they were not designed for this type of real-time application). This means we have to wait for data formats or protocols like RTP  or RSVP  together with IPv6  to be implemented in our communication infrastructure.
The current interoperability problem in TCP/IP-based desktop videoconferencing will very probably improve significantly with the introduction of H.323-compliant products. Major vendors such as Intel, PictureTel, and Microsoft have stated their intent to provide standards-based desktop videoconferencing products. Besides offering interoperability, the next-generation products will support multipoint conferencing and will integrate with existing H.320-based infrastructure, thus protecting our investments in this area.
Finally, we have found a clear need for high-end desktop videoconferencing based on native ATM's unique quality-of-service features. Therefore, we will consider H.321- and H.310-compliant products to satisfy high-end user requirements as soon as they become available.
Additionally, point-to-point desktop videoconferencing sessions on top of ISDN are not fulfilling users' needs. A worldwide desktop videoconferencing service is needed that has interoperable end devices, a common addressing scheme, global reachability through international service access points, standardized quality of service characteristics, application management, billing and accounting, and some answers to legal questions.
With this background, the questions to ask are: Do you really need interoperability between platforms and operating systems? Do you want to deploy wide-area conferencing using ISDN and H.320? Is a multiplatform MBone solution appropriate? Of course, you may also have a LAN environment with sufficient bandwidth, so your goal is to achieve the highest quality possible. As has been shown, the answer is not easy and has many facets. Even more complexity is added if you intend to set up a regular desktop videoconferencing service rather than buy such a service from a commercial provider of videoconferences.
Manfred Bogen has been active in group communication, X.400 development, and X.400 standardization since 1983. In 1987 he became head of the research group Value-Added Services (VaS), being responsible for the establishment and provision of value-added services for the scientific community in Germany on behalf of the DFN association. He studied computer science at the University of Bonn and is co-author of two books about X.400 and distributed group communication. At present, he is the convener of the TERENA working group on quality management for networking (WG-QMN), a member of the TERENA Technical Committee and the Internet Society, and is involved as program committee member in the organization of international networking conferences.
Christian Bonkowski finished his education as mathematic-technical assistant at GMD in 1992. He has been the postmaster of the DFN X400-BITNET-Gateway. As a member of the VaS research group, he is involved with the introduction of desktop videoconferencing on a company-internal and international level (TERENA).
Richard Rodriguez-Val has been working at GMD since 1987. He is responsible for GMD's VM systems and the technical support of all VM applications and networking software (TCP/IP, SNA, and OSI). He has been actively involved in supporting EARN/BITNET in Germany since the early days. He is currently the postmaster of the official DFN Interbit gateway at GMD. Additionally, he is involved in quality and performance measurement of PCs and high-speed networks such as ATM.
Clemens Wermelskirchen has worked at GMD since 1975. He received his master's degree in Physics in 1982 and his Ph.D. in 1988 from the University of Bonn. He worked in the development and operational area of electronic mail services using OSI and Internet protocols (X.400, SMTP, and Gateways). As part of the VaS research group he is responsible for the establishment and introduction of a desktop videoconferencing service at GMD. Since 1995 he has been chairman of the Network-SIG (special interest group) of the German chapter of DECUS (Digital Equipment Computer User Society).
The authors can be reached at GMD, German National Research Center for Information Technology, D-53754 Sankt Augustin.