The Internet is not growing only in the numbers, the Internet has also expanded to new areas. During the last few years, business has adopted the Internet. E-mail is extensively used in sharing the data between company employees, the different branch offices and even different companies [1]. Nowadays, almost every company has its own WWW home page, it is considered old-fashioned not to have a show window on the net. The Internet is even used as an extension to company's TV commercials: more and more often there is an http address to the company's home page at the end of a TV commercial. Several companies have transferred their product development to the net: databases are read from remote locations from other side of the country and instructions to subordinates are sent by e-mail. Definitely, one can say business has come to the net. But how can the strict requirements of business users be fulfilled on the crowded Internet?
This text discusses the user expectations and requirements of using
the Internet and how they could affect the Internet and create a need for
broadband networking. Also, some future trends and their effects are inspected.
As well, there is a description of some solutions for transmission and
router technology and how they could respond to requirements of increasing
number of Internet users. This document concentrates mainly on wide band
backbone technologies, the access technologies are not addressed. The Internet
and broadband networking is discussed overall and the technical details
are not inspected intensively. This text does not try to define the future
of broadband networking, this is merely a thought provoking text about
what could happen in broadband networking.
Even today, the Internet suffers from severe congestion problems. For example in USA, the national Internet is very crowded during the working day. As well, it takes longer to connect to an American host from Finland during the American business hours than, for example, during the early Finnish morning hours. Also, some special occasions, such as the collision of a comet with Jupiter and the death of a Princess, can bring the network to a temporary halt, when everyone is trying to receive and send more information about what is happening. [1]
Especially demanding Internet users are the business users. Company confidential data cannot be transferred through insecure Internet. Also, the business users require high quality networking: a company cannot use the Internet for critical tasks before it is absolutely sure that there is always certain level of service for them in the Internet. For example, it would be a totally unacceptable situation for a company, when they are trying to transfer multimegabit technical CAD drawings to a branch office, that the transmission paths they use get congested and almost nothing could be sent.
Although there are some problems with the security and the guaranteed level of service in the Internet, it is reasonable to assume that there will be some solutions. And when there is a proven and credible way to guarantee security and the level of service, the business users will exploit the possibilities of the Internet more and more extensively. Already now, e-mail and WWW services are very common in business, but when the companies are assured about the real possibilities of the Internet, the majority of the available bandwidth will be used by business users.
Figure 1. Applications driving the requirement for broadband. [1]
Client-server architecture, increased use of PC and file transfer are all factors that create demand for broadband in terms of high traffic volume. Some client-server applications, like web browsing, need broadband transmission path also for lower latency time. The first inherent application, video conferencing, is only the fourth important driver for broadband. So, the sheer number of users is creating more need for high bandwidth than the inherent broadband applications.
E-mail, Worldwide Web and file transfer (ftp) are the single applications that create the greatest growth in traffic in the Internet [1]. Increased PC use has brought the WWW available to almost everyone. As well as the number of WWW users grow, also the Web pages are coming more and more complex: in addition to good quality graphics, Web pages include audio and video clips. In spite of greater amount of data to be transferred, users still want to download Web pages as quickly as possible. This need for low latency combined with huge number of user creates constantly growing need for broadband networking.
Another application with enormous amount of traffic, e-mail, requires broadband to cope with the volume of traffic and to reduce the time required to download mails from a service provider. Both the number and the size of e-mails have been increasing steadily. A few years ago, e-mails were usually ASCII text and their size was seldom bigger than a few kilobytes. Nowadays, e-mails are likely to include word processor documents, presentations, graphics and even moving pictures as attachments. The size of the e-mails has increased from kilobytes to megabytes. E-mail is also becoming a more and more important business tool. As business becomes increasingly dependant on e-mail, the need for service level guarantees will grow: excessive delays in sending and receiving an e-mail are not acceptable.
As well as the user requirements will increase, also applications' demand of resources will get higher. Many applications are being designed to expect high bandwidth availability. For example, the same text file saved in Word 2.1 takes much less storage space than saved in Office 97 format. In the same manner, the new versions of the old program loaded with graphics, sounds and moving images require more bandwidth than the ancient ASCII-based versions did. It is unreasonable to think that the applications would solve the problem of increasing need of transmission capacity, the applications are not getting less bandwidth thirsty.
One interesting new evolving application developed for TCP/IP is voice over IP (VoIP). Packet voice has a number of advantages compared to circuit switched voice. For example, VoIP uses transmission capacity more efficiently because capacity is used only when there is something to send and the data is compressed (Internet telephony typically uses 9,6 kbit/s - compared 64 kbit/s in ordinary PSTN network - and sometimes even as little as 2,4 kbit/s). In addition to that, VoIP has lower billing costs and the ability to pick the cheapest route for the call. For the caller, this means much lower minute rates - especially in long distance and international calls - than in ordinary PSTN calls.
The first commercial Internet telephony program was published in January 1995 by an Israelian company VocalTec. Since then, the needed software and hardware have advanced rapidly, and the focus of the developing efforts has changed from VoIP software to gateway techniques. Last year, first operational gateways were opened to the public use. Gateway offers a possibility to call from your PC to ordinary PSTN phone or even from PSTN phone to another PSTN phone. The only difference from conventional PSTN call is that the voice traffic is routed via Internet.
If the available technology is good enough, the Internet telephony could
be regarded as an attractive option for conventional PSTN. Widespread adoption
of the VoIP would create an enormous need for bandwidth, but with the present
backbone capacities, this need cannot be fulfilled. However, Internet Service
Providers (ISP) are working with the problem. For example, the world's
biggest ISP, UUNet, is participating in a planned transatlantic cable,
in order to ensure the transmission capacity for their clients [1]. This
kind of development could create interesting scenarios: with own transmission
capacity, ISPs could transfer from the telecom companies' clients to their
competitors. Especially this is true, if ISPs start to widely offer telephony
services. The situation is also noticed in the industry: for example, both
Mr. Jorma Ollila (President and CEO of Nokia) and Mr. Pekka Vennamo (President
and CEO of PT Finland
Ltd) estimate that VoIP is the next "killer" application in the Internet
after WWW.
Another interesting trend affecting the need for broadband is the increasing
significance of TCP/IP. Traditionally, the amount of transferred data has
been much less than voice traffic. However, data - and especially IP traffic
- has been growing faster than voice for many years. The Yankee Group estimates
that "IP traffic may overtake voice in four years, when measured as total
volume over the backbone network" [2]. The development is depicted in the
figure 2.
Figure 2. The development of the total voice and data traffic. [3]
More optimistic view is presented by Ovum: during this year, the total amount of voice traffic is exceeded by non-voice traffic and within five years, large corporate customers will treat voice as just one application on their broadband networks, although important one [1]. These estimates show that the future is for TCP/IP and as the total amount of traffic grows, so should also the capacity.
The new access technologies create also an interesting situation. xSDL and cable modems are coming to market and they offer Interenet access at about 2 Mbit/s speed. With the new access technologies, Internet users can fill the transmission capacity of Internet backbone easier than before. If the backbone capacity is not increased, there is no much use of greater access speed. Consumers will not buy the technology, if there is no significant improvement in the access speed, so the manufacturers and retailers of the new access technology should also concentrate and invest on increasing the capacity of Internet backbones.
In the following chapters, a few transmission technologies are discussed as a possible option for future Internet backbone transmission media. Also, some router manufacturers and their products are inspected. The technical details are discussed in other Tik-110.551 Internetworking seminar papers, only the main principles are highlighted here.
Figure 3. The old vision of the broadband ATM network. [1]
For many years, the view of the development of broadband networking has been that it would use ATM transmission and switching, but then the rise of TCP/IP has challenged the old vision. Especially the manufacturers and retailers that rely completely their business on the Internet, have introduced some new alternative approaches. Those approaches are collected to figure 4.
Figure 4. Alternative approach to broadband networking. [1]
The need for ATM in broadband networking is not evident anymore. Almost all the router manufacturers offer solutions which use IP over SDH, without ATM. In addition to SDH, there is even a faster transmission technology, WDM (Wavelength Division Multiplexing) that offers much needed speed up in backbone transmission media technologies. Ciena, a router manufacturer, has concentrated successfully on developing WDM router technologies.
There are few features in ATM that do not favour ATM. ATM is not optimised for any application. The choice of the 5 + 48 byte cell structure was a compromise. It is not ideal for any of the main applications - voice, data or video [1]. If ATM is used to carry TCP/IP traffic, there is a duplication: both ATM and TCP/IP have addressing, routing and flow control. To transfer TCP/IP data, it would be more economical to use a transmission technology that does not provide the same features as TCP/IP. In that case, SDH could be one solution.
However, ATM has also several strengths. Through standardisation, ATM has achieved a position that other, un-standardised technologies do not have. Also, ATM can guarantee certain level of service. That is something TCP/IP cannot provide, at least not yet. One important benefit ATM has is that ATM can be used as both LAN and WAN technologies, which allows improved efficiency by eliminating the need for conversion between protocols.
Although ATM can be used in LAN networks, ATM is not likely to be the access method at the desktops. Ethernet technologies are advancing so fast that ATM cannot provide much advantage, even though ATM could connect each user to the net at the speed of 25 Mbit/s. Also, the use of ATM in the building or campus backbone will be reduced by Gigabit Ethernet: ATM with a speed of 625 Mbit/s is not regarded as a better solution to Gigabit Ethernet, especially in the situation when Ethernet technology is already familiar from LAN networks and ATM requires a switch to a completely new and unknown technology.
One of the advantages SDH has is that it is designed to carry a wide range of signals, including PDH signals, frame relay, FDDI, TCP/IP, ATM cells and any foreseeable future data streams [4]. In other words, SDH is backwards compatible with existing PDH systems. The SDH frame has been designed to carry management information, allowing the network to be fully software controllable. SDH can operate at different rates and at the moment, the fastest line rate is 10 Gbit/s, compared to 625 Mbit/s of ATM. Even faster SDH systems are under development, several vendors are working on 40 Gbit/s SDH systems.
Because SDH is becoming more and more widely used digital transmission technology, there is now an increasing amount of interest in carrying IP traffic directly over SDH, by-passing the ATM layer. Especially ISPs are interested in adopting SDH as their transmission media in order to avoid the duplication ATM creates with TCP/IP traffic. Many router manufacturers have already published router models that are able to use directly SDH without ATM. The new gigabit routers will provide sufficient speed so that ATM switching will not be required and TCP/IP can operate directly on SDH using routers with optical interfaces.
The fastest existing WDM systems are sixteen channel systems which give a capacity of 40 Gbit/s, but 32 and 40 channel systems are being tested giving a capacity of 80 Gbit/s and 100 Gbit/s respectively. Even greater speeds are achieved in laboratories: in January 1997, Nec announced that it has achieved a transmission speed of 3 Tbit/s using WDM [4]. WDM significantly increases the capacity of the link without the need to install more fibre or the need to use transmission equipment running at higher bit rates. For example, if four 2,5 Gbit/s SDH channels at different wavelengths are transmitted in a single fibre, the total capacity of the link is increased to 10 Gbit/s.
However, there are a few drawbacks in WDM: WDM is expensive and there is not much competition among the few manufacturers to lower the prices. Also, WDM systems are at the moment proprietary, so interworking between different manufacturers' systems is difficult. However, the standardisation work is progressing in ITU.
Several router manufacturers have announced that their future routers, that will be launched on the later half of 1998, will support WDM tehcnology. At present, there are not many available routers that could operate directly on WDM. Instead, SDH is used as an interface to WDM systems.
IP9000 router is based on ATM and Ethernet technology. The router can be connected to Ethernet, Fast Ethernet, Gigabit Ethernet and ATM at a speed of 10 Mbit/s, 100 Mbit/s, 1 Gbit/s and 155 Mbp/s respectively. Depending on the configuration, IP9000 could route over 20 million packets per second. The eight-slot router with 64 Ethernet / Fast Ethernet ports costs $55 000. [5]
Avici's Terabit Switch/Router (TSR) technology is developed for ISPs and other carriers, but the technology is scalable and can be used also in environment requiring less bandwidth. TSR technology exploits the benefits of Sonet technology, which is North American counterpart for SDH. Avici's TSR is designed to scale from 600 Mbit/s to multiple terabits. TSR will use hardware-based routing, forwarding, multicasting and Quality of Service to maintain gigabit line rates. Each TSR line card sports a Direct Connect Fabric (DCF) that serves as a 70 Gbit/s router. Twenty of these line cards will fit into a dual-shelf chassis. With DCF, each line card adds 60 Gbit/s of switching capacity to TSR. In a fully loaded dual, TSR will have up to 1.4 Tbit/s of switching capacity. ISPs can even scale beyond that by adding more TSR shelves and line cards to their equipment. [6]
Because the product is far from ready to be published, no cost estimates were given. Obviously, a router capable of terabit routing/switching speed is not very cheap.
Ciena has introduced its newest product, Multiwave 4000 System, which is capable of combining 40 to 96 2,5 Gbit/s Sonet/SDH channels into one single fibre. Also, the system's open architecture enables carriers to mix Sonet/SDH, ATM and IP traffic on a common optical network. [7]
Ciena's situation resembles Cisco's about 20 years ago: Ciena is a market leader in a very new area of tehcnology and the company is very profitable. Is Ciena the next Cisco?
On the one extreme, there is a complete convergence to one network: there is no anymore need for different voice and data networks, everything is handled in the same network which can guarantee the level of service and high security for every single user. Could TCP/IP provide this?
It is probable that different technologies are used for different applications. For example, telecom companies could use ATM in long distance transmission paths and ISPs could rely on IP over SDH in their lines. The winner technology, if there is one, could be the technology which is the most flexible and provide truly interoperability with other technologies. Or maybe the winning technology is still on the drawing board of some talented researcher.