Asynchronous transfer mode (ATM) is a great technology. It promises responsive, high-quality voice, video, and data services over a single network. ATM will enable new applications and reduce bandwidth costs. So why hasn't ATM become popular yet?
Simply put, today's applications cannot take advantage of ATM's capabilities. ATM switches have problems shunting bursty 100-Mbps local traffic onto 34-Mbps (E3) WAN links. Also, current LAN protocols like TCP/IP and IPX cannot take advantage of the bandwidth offered by ATM. The ATM standards bodies are starting to remodel legacy protocols, such as TCP/IP and IPX, to bring them up to scratch for high-speed cell switching, but this will take time to implement.
So if you want to be an early ATM pioneer, you currently have only one option: Install single-vendor solutions based on native ATM links between workstations and switches. And you must rely on proprietary software for it to work.
A key hurdle in the ATM interoperability race is the need for the LAN and WAN industry, plus the ATM Forum and the International Telecommunications Union (ITU), to come to a real agreement over the details of the standard. The ATM Forum is not a standards body, but it makes recommendations and sets guidelines. There are over 650 members, with that number expected to top 1000 next year. Clout, yes. But universal agreement? No. Yet it has to be acknowledged that the forum has made enormous strides in getting standards ratified.
Ask major European buyers of network products what they want and the usual reply is frame relay and ATM. But when it comes to relating the technology to the aims of their businesses, it's a different story. To the guy who signs the check, the business needs do not relate to the hype about frame relay and ATM.
Misconceptions abound over ATM. There has been a lot of debate over whether you should run 155-Mbps ATM over Category 3 untwisted pair (UTP) wiring, the most common kind of cabling in corporate networks. You can bring synchronous-digital-hierarchy (SDH) technology to the desktop without having to pull new cables. However, certain technical difficulties argue against bringing 155-Mbps ATM to the desktop over Category 3 cabling. There could be problems with transmission-signal strength, crosstalk, and radio-frequency interference. Also, ordinary PCs would have trouble coping with 155-Mbps traffic, and the necessary adapters would need extensive and expensive buffers.
For the backbone (Category 5 and fiber cabling), 155-Mbps ATM is the logical choice. For the desktop, 25-Mbps ATM -- which runs over Category 3 UTP without special technology -- is the most technically sound and cost-efficient way to deliver ATM to your site.
Another concern about ATM is that there may be some new technology waiting in the wings to take its place. The fact is, all the PTOs and carriers are busy setting up pilot ATM networks and building cross-connect overlays. The commitment is there, and ATM handles both WANs and LANs.
Though some analysts say that ATM switches will replace routers, it is not true. Hub vendors relish the thought that the inherently higher performance of emerging switching architectures tolls the death knell for routers. But this is correct only if a company can afford to completely upgrade its entire network and applications in one go to achieve a clean ATM base. Switches and routers will coexist for a long time. Businesses will not dump their legacy systems all at once.
Another rumor says, "ATM is not the only real choice for switching." This is not true, since ATM specifies how cells are to be assembled but does not define a product's internal switching architecture. So, anyone looking at switching technologies needs to consider throughput rates, congestion control, standardized interfaces, and network management. Furthermore, it is not true that the bigger a switch, the better its performance. The best metrics for a proper assessment are LAN adaptation software, virtual LAN support, video compression, bandwidth reservation, and priority queuing.
On the other hand, some network experts predict that Switched Multimegabit Data Service (SMDS) will knock ATM out of serious contention. (SMDS is called Connectionless Broadband Data Service (CBDS) in the European definition made by the ETSI.) The fact is, however, that SMDS has been around a long enough time to acquire the unflattering moniker of "somebody might deploy someday" technology. SMDS was defined by BellCore before the arrival of ATM and was developed for data transmission. Its commercial success is poor. According to OST's Sylvie Ritzenthaler, a member of the ATM Forum, SMDS has no chance in a face-off against ATM.
Much of the work now under way in the standards camp will extend ATM to narrowband speeds, enhancing LAN-WAN internetworking, providing high-speed LAN connectivity over ATM, and expanding voice-compression options. Cell-based ATM standards apply now to broadband infrastructures from speeds of 34 Mbps (E3) upward. For narrowband speeds, only the frame-based data exchange interface (DXI) standard applies. But DXI lacks the key flexibility and performance benefits of cell-based ATM.Standards are progressing for ATM at narrowband speeds using the fixed-length 53-byte cell format. These standards, which analysts expect will be ratified soon, open the door for more broadly applicable narrowband ATM services from carriers. Standards for ATM at T1/E1 speeds (2 Mbps) and below will accelerate the migration from private networking based on time division multiplexing (TDM) to ATM-based hybrid networking, as users look to increase the performance and control the costs of their communications.
Work is also under way on the Private Network to Network Interface (PNNI) for LAN-to-WAN internetworking, multiprotocol transport over ATM (MPOA), and LAN emulation for high-speed LAN connectivity over ATM.
Voice Compression Standards
Having approved the G.728 standard for low-delay Codebook Excited Linear Prediction (LD-CELP) compression at 16 Kbps, the ITU is now working to ratify a standard for CELP-based compression at 8 Kbps. Results are expected early next year. Meanwhile, the ITU has standards for Adaptive Differential Pulse Code Modulation (ADPCM) voice compression at 32, 24, and 16 Kbps.
Voice transport over frame relay and ATM has been the topic of much discussion. It appears that no standard for voice over frame relay will be ratified, since most frame relay products have delay and discard characteristics that are inappropriate for reliable transport of voice samples. StrataCom's FastPad, however, is a backbone access device that effectively uses the frame relay protocol to carry voice. This is made possible by the low-delay, low-discard cell-based characteristics of the company's implementation of frame relay technology.
Voice over ATM is currently defined using uncompressed PCM at 64 Kbps per channel, streaming into cells using the simplest ATM adaptation layer, AAL1 (see the figure). However, there's a lot of interest in compressed voice over ATM using ADPCM and LD-CELP, together with adaptation that does not generate cells during silence. While no standards bodies are doing specific work on this issue, expect it to be raised once the more fundamental PNNI and LAN emulation issues have been addressed.
According to Sylvie Ritzenthaler, the ATM Forum is studying "legacy voice at a native ATM terminal." One of the aspects of this new service will be an internetworking link between ATM networks and public or private ISDN.
The European WAN market is driven by value-added-service providers and PNOs. Pricing is a major issue, since E1 leased lines cost four times as much as T1 lines. Separation of LAN and WAN environments is expected to continue in Europe. For this reason, data and voice services are usually offered on different networks in Europe, which is likely to remain so until providers of alternative infrastructures emerge.
ATM Must Deliver the Goods
Users are not going to shoulder the expense and trouble of installing ATM merely to run low-end voice and video applications. ATM is going to have to deliver higher-quality voice and video service than standard multimedia applications offer today. For example, videoconferencing, even at 128 Kbps, is not brilliant, particularly when one looks at video and audio synchronization and even when the latency is within a tolerable limit.
Although its availability is scarce today, ATM has a lot going for it. It's more scalable in terms of speed and network size than competing technologies (switched Ethernet and Fast Ethernet). Most important of all, it can bridge the LAN and WAN worlds, which really does set it apart from its LAN-based rivals.
WHERE TO FIND
General DataComm Wokingham, U.K. phone: + 44 1734 774868 fax: + 44 1734 771505 Cabletron Systems Newbury, U.K. phone: + 44 1635 580000 fax: + 44 1635 44578 Fore Systems Contact U.K. business partner, K-NET Yateley, U.K. phone: + 44 1252 877443 fax: + 44 1252 872890 StrataCom Fleet, U.K. phone: + 44 1252 815554 fax: + 44 1252 815428
illustration_link (6 Kbytes)(/art/9512/img/507005i2.htm)
In Europe there are a few sites where potential ATM users can
test equipment for conformance to specifications, signaling, and
interoperability. Projects have to be approved by a panel of members
from a research organization, the European Commission, or an outside
agency. Some costs may be involved, too, although prices will be low
compared to what you would pay to use a private laboratory.
The UK Education and Research Networking Association (UKERNA) site (Didcot, U.K), has a 34-Mbps set-up, with links to other ATM test-beds available.
Phone: +44 1235 822 200
Fax: +44 1235 822 399
The Institute of Computer Science (Kista, Sweden) offers testing at 155 and 622 Mbps.
Phone: +46 8 752 1527
Fax: +46 8 752 7230
Deutsche Telekom's Deteberkom site (Berlin) offers 155-Mbps testing.
Phone: +49 30 46701 213
Fax: +49 30 46701 444
illustration_link (15 Kbytes)(/art/9512/img/507005j2.htm)
ATM has been chosen as the underlying transport technology within the broadband ISDN protocol stacks. The BISDN reference model is divided into multiple planes and layers. The User plane provides for the transfer of application information. The Control plane protocols deal with call establishments. The Management plane defines a platform to exchange information between User and Control planes.