10 Virtual Collaborative Environments
The Virtual Collaborative Environments activity comprises a number of particular activities united by multimedia transmission over high-speed networks. The activities can be divided into two major areas: synchronous communication infrastructure designed for interactive communication (e.g., videoconferencing), and asynchronous tools where latency requirements are relaxed; these tools are often represented by unidirectional data transmission.
10.1 Synchronous Communication Infrastructure
Research in the area of synchronous communication infrastructure followed our previous results in research and development of user-empowered reflector - a tool which replicates data to a group of participants regardless of multicast support in the particular network [ICN04]. The design of the reflector stems from a router architecture for active (programmable) networks; its design is modular so that users can extend it using modules linked in during run-time, thus making it usable not only for simple data replication but also for data processing. Because the reflector works in user-space of the underlying operating system, it can be run by any user without a need for any administrative access to computer or network elements.
A single reflector suffers however from poor scalability - the limiting factor turns out to be number of outbound flows which grows quadratically with the number of connected clients. This problem emerges especially when using high-quality high-bandwidth video formats like Digital Video (DV) [ECUMN04]. Natural direction of further development are therefore reflector networks featuring better scalability and also robustness in terms of both network and reflector outages. We have developed and verified a number of possible data distribution models which provide different ratios between redundancy, robustness, and scalability [ICN05]. To achieve maximum overall robustness, we are prototyping a reflector network that uses JXTA-based peer-to-peer network for organizing the reflectors and for implementing out-of-band control channels.
10.1.2 Active Networks
As a theoretical base for developing reflector, we have further researched the concepts of programmable router for active networks. During this year we have designed and implemented the Active Router Transport Protocol (ARTP) [ARTP] for unreliable data transmission with guaranteed order of delivery and TCP-friendly flow control based on an AIMD algorithm. The implementation is considered to be a reference one without careful optimizations but still the performance of the protocol is interesting even for gigabit network links, though it does not saturate them totally now. We intend to use the ARTP for transmission of large multimedia streams.
In H.323 videoconferencing, we have focused especially on building sufficiently scalable infrastructure running in a production mode. A multipoint connection unit (MCU) serves as a basis of multipoint conferences; we have selected and purchased a MCU Polycom MGC-25 and put the device into semi-production mode. We plan to enhance it with SIP signalling support, integrate it with the rest of our H.323 infrastructure, and eventually put it into full production mode in the beginning of 2005 after the whole functionality is properly verified. After finalizing all planned changes in the H.323 and SIP infrastructures, we expect to hand over the whole infrastructure to be used by the Czech academic community in a production manner.
We have enhanced current videoconferencing testbed with powerful videoconferencing clients focusing especially on SIP signalling and collaboration over shared contents. We have further developed our cooperation with the IP telephony group; together we plan to deploy new services for IP clients with the H.323 and SIP signalling.
We have improved the quality of videoconferencing services provided at the CESNET premises. Two new videoconferencing workplaces are available using a fully modular technological solution, so that the workplaces are capable of working either independently or together and provide improved collaboration over shared contents even for larger groups. Thus we respond to an ever increasing demand for videoconferencing services from the academic communities located within the CESNET premises.
Based on requests for high-quality conferencing with strong security, we have prepared several new personal videoconferencing sets. In a first batch, we distributed a fully mobile personal set supporting the H.264 video format, AES encryption, stereo sound and interactivity over shared contents.
10.1.4 AccessGrid and Large Communication Complexes
During the last year, unique mobile videoconferencing node has been created to be used for AccessGrid-compatible conferencing. This activity was funded by the CESNET Development Fund and Masaryk University. Our activity is taking over further development of this node - e. g., by the end of 2004, all audio paths have been converted to balanced cabling to remove noise cumulated on audio paths due to its extremely compact solution and resulting high concentration of all cables including power cables.
We also started building a third AccessGrid (AG) node at the CESNET premises which should serve as another reference AG workplace in the Czech Republic. In the next year, we plan to finish this node and develop further tools for managing large communication complexes based on common client software (web, Java applets, etc.).
10.1.5 Transmission of DV over IP networks
Transmission of Digital Video (DV) over IP networks is standardized in RFC 3189 and RFC 3190; it uses the RTP/UDP network protocol stack. A purely software DV over IP implementation has been implemented by the DVTS project. Since we lacked a working display tool for UNIX platforms, we have completely revamped the xdvshow tool which allows displaying DV stream in X11 interfaces and which had not been maintained for quite a long time. Our new version supports displaying using X11 and SDL interfaces with the SDL version supporting also full-screen mode - either an up-sampling based scaled one or an unscaled one which switches the computer screen into the closest matching supported resolution and fills the rest of the screen with black background. The latter mode is suitable for slower computers. In cooperation with Internet2, the new xdvshow was distributed as a part of the official CD with tools for DV over IP during the Internet2 Member Meeting, Fall 2004. Development of DV over IP tools is further pursued by the DVTS consortium where CESNET is represented by the Masaryk University that can participate as an academic institution for free.
DVTS tools and xdvshow were also used for prototyping stereoscopic (3-D) video capturing, transmission, and display (Figure). Cameras positioned on a stereoscopic tripod head are used for producing separate DV streams for the left and right eyes. The DV is transmitted using an IEEE-1394 (FireWire) interface and sent over IP network using the DVTS tools. Both streams are received by a single dual-processor computer which displays them using a dual-headed graphics card. The display part uses two projectors with polarizing filters with perpendicular polarizing planes, a special non-depolarizing screen, and glasses again with perpendicularly oriented polarizing filters.
10.2 Asynchronous Infrastructure for Processing and Distribution of Multimedia
10.2.1 Distributed Encoding Environment
The Distributed Encoding Environment (DEE) [TNC04] was created in collaboration of the CESNET Development Fund and Masaryk University to allow efficient asynchronous processing of multimedia material. The DEE uses Grid computing facilities of the MetaCenter project (Chapter) for processing, and the storage capacity of Distributed Data Storage project for storing the interim data. The parallel transcoding process is based on splitting the input material into smaller chunks that are encoded in parallel and merged into a final file during the last phase. In this activity, we continue developing the DEE environment according to requirements from pilot user groups and we also provide a limited support for deploying DEE at various user sites.
We have enhanced the multimedia metadata search engine with support for new formats; as a result, the number of indexed files increased to more than 100,000 - and this number currently covers the Czech Internet only. We have developed a proprietary XML-based format [TR19/04] storing the metadata in the first phase of implementation for the sake of simplicity; we plan a transition to the MPEG-7 format during the next year. Despite the recently announced activities of global portals, our system remains unique because of the way it indexes and searches the data.
All our current streaming activities are based on a streaming farm we have built during last several years. This farm is modular, comprising a number of components that together provide all the services visible to our end users. During this year, we have extended the farm with servers supporting streaming of MPEG-1/2/4 and QuickTime formats. Together with existing RealMedia and Windows Media streaming servers, the academic community in the Czech Republic can use all the mainstream streaming formats. Some servers also support streaming using both the IPv4 and IPv6 protocols.
Within our development activities, we have also experimented with transitioning to distributed storage based in the iSCSI protocol in WAN environment. However, it turned out that the latency in current WAN environments hinders deployment of iSCSI-based services.
During December, we have participated in the MegaConference event which is a global videoconference organized by the Ohio State University with support of the Internet2 association. Our task consisted of streaming the conference using the Windows Media technology on both IPv4 and IPv6 protocols. In the last year, we have also built three Content Delivery Networks (CDNs) used for live broadcasting of the RIPE 47, RIPE 48, and RIPE 49 conferences.
10.2.4 Announcing Portal
We continued our participation in the TF-Netcast project organized by the TERENA association where we further developed the announcing portal (prenosy.cesnet.cz). TF-Netcast successfully ended up in the Spring 2004 and thus we are taking part in its follow-up project called TF-VVC (Task Force - Voice, video and collaboration) where we lead the activity E called an Academic Netcasting Channel (Live Streaming Infrastructure). We are also actively participating in other activities, especially in metadata handling. We have enhanced the functionality of the announcing portal (Figure) with RSS feeds, and started working on an Electronic Programming Guide (EPG) as well as on implementing a new architecture based on separating contents and view. The portal also supports additional languages so that their total number reaches ten. The portal was used to announce more than 150 streaming events.
10.2.5 High-Quality Audio Streaming
We have further developed an experimental system [TR16/04] relying on distributed audio signals in the Ogg and MPEG formats with bandwidths of 128, 192 and 224 kbps. During this year, we have also ceased supporting the MP3 format because of patent-related issues pertaining to this format, better quality achieved by the Ogg format and also sufficient support for the Ogg format in mainstream client software. Higher sound quality has been achieved by using new input signals: we have replaced analog tuners with new digital DVB-S cards and stations ČRo 1, 2, 3, and recently also ČRo 6 and Region are received directly from the Eurobird 1 satellite link. The resulting sound is clearer and more detailed with better separation of channels. Alltogether we are broadcasting the following radio stations: ČRo 1, 2, 3, 6, as well as the Region, BBC, ČRo Pilsen and ČRo Regina.
Since we need to process several DVB-S programs simultaneously, we are using the vls application server. This sever can receive all the streams at once, separate them and then pass them on in the original MPEG-TS format for further processing - e.g., by broadcasting them to specified multicast and unicast addresses. Each MPEG-TS stream is received and decoded into raw PCM format by the vlc client application and further encoded into the Ogg format using the Ices software. Primary encoding servers are located at the West Bohemian University in Pilsen and also at CESNET premises in Prague. All streams are available from the radio.cesnet.cz web pages and also from live streaming pages of the Czech Radio. We support both the IPv4 and IPv6 protocols.
In the future, we plan to concentrate on robustness of the whole system and its resiliency with respect to outages. We expect to deploy redundant configurations of encoding and broadcasting servers with support of automatic reconfiguration in case of network or node failure and also automatic client reconnection to different broadcasting server should some problem occur.
10.3 Direct Support and Consulting
We continue to collaborate with student groups that provide content for our streaming servers. We have evaluated several possible scenarios for acquisition and according to their results we have designed and implemented a mobile and fully digital workspace for contents acquisition. The workspace can capture small-to-medium events as well as support audio and video presentations. Verification of this solution is planned for beginning of 2005.
As a part of our consulting services, we have participated in designing and selecting audiovisual equipment for two reconstructed and one newly built lecturing halls at the Faculty of Informatics, Masaryk University [ITHET04] which are used for automatic lecture recording since Fall 2004. The captured video is further processed using the DEE deployed on dedicated processing infrastructure operated jointly by the FI MU and the MetaCenter (Chapter).
We have provided consulting services for reconstruction of two large-capacity lecturing halls at the FEL ČVUT in Prague-Dejvice and we succeeded in creating conditions appropriate for deploying our presentation and conferencing technologies. The whole design is in complete accordance with our experience.
Utilizing our videoconferencing rooms, we have facilitated organization of several tele-presence events for distant education, e. g., recurring remote lectures of Assoc. Prof. Kolář (Computer Department of FEL, ČVUT) for the Tecnológico de Monterrey, Universidad Virtual, Mexico virtual university.
In addition to the events mentioned above, we provide support and consulting to all members of the CESNET Association. We distribute videoconferencing sets, consult purchase of equipment, detect and solve problems even in local networks of our users and we provide support for regular international videoconferences of the CESNET members.