Site design and hosting by
OARnet Copyright 2001 MegaConference III.
All rights reserved.

Megaconference III was held on Oct 3, 2001. Over 200 sites from around the world on all continents participated. It had been planned that the Megaconference would be hosted at the Internet2 meeting in Austin, TX, but that meeting was cancelled due to world events. So instead, the moderator of the conference, Sandy Sprafka, ran the conference from her home location at North Dakota State University. The Megaconference was controlled from a technical perspective by Robert Dixon at Ohio State University.

About 25 H.323 multipoint control units around the world were cascaded together to make the Megaconference possible. Over half of these are located at Ohio State University/OARnet, where they are in the Internet2 Commons central bank of MCUs. Gabe Moulton and Megan Crabb of OSU, Arif Khan of OARnet, Andy Shapira of Polycom/Accord and Ben Pinkerton of Starbak ran the central MCUs.

The Megaconferences continue to approximately double in size every year. Megaconference I had about 50 sites, Megaconference II about 100 sites, and Megaconference III around 200 sites. They continue to raise the record for the largest ever Internet video conferences, and continue to push the state of the art of video conferencing and networking to new levels.

Megaconference III was organized just like any professional conference. There were invited speakers, who gave presentations on many aspects of how they are using internet video conferencing in their teaching, research and organizational activities. All of the speakers presented from their own locations, and there was no central site. All of the audience was distributed around the world, and all were fully interactive, able to ask questions from anywhere to any speaker. Mary Fran Yafchak of SURA arranged for the speakers, and Sandy Sprafka of NDSU and Megan Crabb of OSU completed all the final myriads of logistical details regarding the speakers presentations.

One of the new features this year by popular demand was the Megacon Cafe. These are the break and lunch periods between the presentation sessions. To make this as much like an in-person conference as possible, the MCUs were opened to everyone to talk about whatever they wished. In some cases people continued discussions with the speakers and in other cases some new collaborations were formed. These are always the unplanned great things that can happen in an open environment. During one of these breaks a roll call was taken of all participants, and 10 door prizes were awarded by a random drawing. The prizes were donated by George Wiley of Viewcom Corp.

In the six weeks preceeding the Megaconference, a series of load tests and rehearsals were conducted, which taught us many things. In the load tests, we asked all H.323 users in the world to connect to the central bank of MCUs at the same time, so we could find out what would break. About 100 people were very helpful in doing so. We found that we had local network problems that were never revealed before, requiring significant improvement in some campus fiber optic cables, and an upgrade to gigabit ethernet between the MCUs and the Internet. OARnet staff Gene Wallis,

Paul Schopis and Mark Fulmer were instrumental in diagnosing the fiber problems. We also found problems with the MCUs that were never known before, requiring considerable discussion with RADVision and Accord engineers. Even so, the load tests had only 100 users, and the Megaconference had 200 users, so we were still beyond the size of anything we could test.

All of the MCUs (RADVision ViaIP, Polycom/Accord MGC-100, RADVision MCU-323) had various kinds of failures during the tests and Megaconference.

But for the most part the conference ran smoothly, and the occasional failures were relatively easy to repair by remote control.

The most difficult problem to diagnose was occasional loss of communications between the RADVision ECS gatekeeper and the various MCUs.

After much investigation, we now know that the ECS gatekeeper has a hitherto-unknown problem that manifests itself only in large conferences. The ECS can support very large conferences with no problem, so long as not everyone tries call in at the same time. It is like the telephone; if everyone picks up the telephone at once (as in an emergency), not everyone can get dial tone. Despite our best efforts to get people to connect early, over an extended time period, most people waited until the last minute and then all called at once. This locked up the ECS gatekeeper. RADVision is working on this problem.

We did discover one difference between the Accord and RADVision MCUs which is very important when running large conferences. The Accord provides an indication of which user has captured the conference, and the RADVision does not. This is crucial when someone forgets to mute their microphone, and we need to mute them centrally. If we cannot tell who it is, we cannot mute them, and this causes significant disruption to the conference.

We have requested that RADVision add this feature. The rehearsals which followed the load tests helped to make all users more proficient in using their equipment, and better able to participate constructively in the Megaconference. All of the speakers were rehearsed, to make sure they could properly bring in their multiple locations and graphic materials and to decide what kind of multiple window display worked best for each one.

The video and audio quality of the Megaconference was in general good, and varied from location to location. In some cases it was not as good as we have in more typical H.323 video conferences, apparently due to the extreme traffic load on portions of the network. The archived video streams available on the web pages from Starbak in the USA and CINECA in Italy show this difference. The CINECA video is often better than the Starbak video, which we attribute to the Starbak streamer being located at the nexus of all the Megaconference MCU traffic (and not to any difference in the quality of the equipment used). Most people were willing to accept somewhat less than perfect video from an event that goes beyond what has ever been done before. The bottom line is: IT WORKED ! Despite all the challenges and problems.

An experimental T.120 server was run from Northwestern University by Larry Amiot, to allow everyone to see the speaker's slides with high quality. It was a RADVIsion DCS, located in the Via-IP MCU chassis there. Unfortunately we could not get more than about 20 users on it at any one time, for unknown reasons. RADVision is investigating that. Two other first-time events happened during Megaconference III.

The world's first Internet Picnic, complete with popcorn eating by the users around the world. And the first ever Megaconference prank, diabolically planned and executed by a gang led by Christopher Cook of OARNet. They played a spoof file about "Dr. Bob" during one of the Megacon Cafe breaks. It is viewable as a video stream on the web pages.

The Megaconference was also transmitted via streaming video from several sites. Ben Pinkerton of Starbak corp provided Torrent 100 and Torrent CE streaming equipment and operated it (Quicktime and Windows Media formats). Bill Miller of OARNet also streamed the event (Real and Windows Media format). The streamed archive files are available on the web pages. Note that you must use the Internet Explorer browser, and the latest version of Media Player, to view them.

An edited "Highlights of Megaconference III" video file has been made, and is available as a streaming video file on the web pages. The original of this is at high quality, and we can provide you with a videotape copy for the cost of duplication and shipping. If you would like a copy, please let me know, so we can plan about how many copies are needed.

More details about all this are available on the Megaconference III web site, located at www.mega-net.net/megaconference. For example, there is a world map showing the locations of all the participants, plus a diagram showing the MCU interconnections and locations, both made by Scott Weaver of OARNet. Jason Beekman of OARNet created and maintained the Megaconference web site, and did yoeman duty with the rapid updates required sometimes on an almost hourly basis. The conference agenda of all the speakers is there, as is the complete list of participants, and all the announcements that were made leading up to the big event. We received many comments from participants after the event, and they are available for you to read in a separate file on the web pages.