Dedicated ATM video area network spans several Ohio high schools - Technology Information

Communications News, July, 2000 by Steven Alan Levine

Collaboration-management software and interoperable products aid in learning process.

The proliferation of broadband networks has created the need for higher quality video collaboration and telepresence in the corporate, industrial and public sectors. Live business television, distance learning and video-area networks require multimegabit bandwidth to produce a visually transparent experience.

Existing data networks were not designed to accommodate real-time traffic such as video. Several standards have emerged that are enabling the reservation of network resources, optimized routing and negotiated transport--crucial to obtaining quality of service (QoS), that makes the real-time video experience transparent and reliable enough to satisfy all session participants.

Many school districts receive grants and apportioned funds to incorporate video collaboration capabilities into their data infrastructure. Medical centers vie for remote consultation, telemedicine and patient video archives that may be recalled for future review by medical students, physicians and legal personnel.

Governments seek to overlay video collaboration on their high-capacity network infrastructure. Desktop client viewers need to tap into the large available video archives, which may not reside locally but can be securely accessed over the network. Live test footage, produced by various military agencies, needs to be shared securely by remote groups and often published on removable mass storage media.

Network managers are clamoring for vendor solutions that are affordable, scalable and manageable.

NETWORK ISSUES

Traditional data networks are built around IP protocol with bandwidth allocation based simply on the number of terminals fanning out from the backbone. Only recently have network managers been capable of reserving resources throughout the network and limiting some nodes while boosting others--not ideal in networks that carry a scalable video user base. Video tends to saturate network routers, switches and buffers. Due to the continuous streaming nature of compressed video, the network designed for asynchronous data traffic is often unable to bear the load of multiple time-critical payloads consisting of compressed video. Compressed video clients, usually H.261, H.263 or MPEG (moving pictures experts group) -I, -II and -IV, are sensitive to network load fluctuations, dropped packets and delays. These anomalies normally don't affect traditional data but are immediately noticeable in video transport manifested by visual and aural jitter. Conventional IP networks that are, by nature, "connectionless" do not solve this on a global standards-based level. There is no widespread methodology established for QoS over packet networks with connectionless service. In the future, virtual private network (VPN) technology based on multiprotocol label switching (MPLS) will appear in disparate network products that should solve the resource reservation issue.

ATM

Currently, ATM forms the backbone of many public networks. Carriers offering VPN and remote enterprise connectivity usually employ synchronous optical network (SONET) infrastructure with ATM topologies. At the carrier's local point of presence, an IP-packet network is emerging from the ATM backbone, which travels the last mile to the enterprise. Even xDSL in the home is an ATM connection with packet service on top of it. ATM was devised to allow guaranteed connections between ingress and egress points on a network. ATM provides a connection-based path with QoS between two or more nodes in a network. Since paths are established a-priori, there is no on-the-fly routing made with each datagram as in the packet-based network.

ATM networks use "cells"--cells are 53 bytes, five of which are header and 48 are payload--instead of packets to create virtual circuits (VC). These circuits may be permanent or switched, depending on the policies of the network manager. A permanent virtual circuit (PVC) is "nailed up" and left up for the time of provisioning. A switched virtual circuit (SVC) may be setup and torn down many times so that the service may be charged as used. SVCs also offer more sophisticated operation than PVCs, at a lower setup complexity.

VIDEO COLLABORATION ISSUES

With a lack of QoS and the inherent packet network delays, it is important that the system be capable of low-latency transmission. This means moving out data units almost as soon as they are produced using user datagram protocol instead of buffering them on the sending and receiving sides, delivering a more transparent experience during interactive sessions. In a point-to-point full-duplex video-teleconference, the latency from end to end should be less than 300 ms in each direction; 150 to 200 ms is more typical with 4-6 mbits of MPEG-II video compression.

Beside latency, lost packets and priority delays may cause video jitter at the receiving end. However, in a network with QoS, such as ATM, it is less likely that packets will be lost or that video will jitter, although latency is still an issue.