Alternative WAN Solutions

Japan is connected with the rest of the communications world by undersea optical-fiber cable from seven points. This communications capacity not only makes Japan a major cable junction point it gives those of us living and working in Japan the capabili ty to easily connect to the rest of the world for high quality communications. Although there are many methods to combine voice, video, and data communications on a single platform, this article concentrates on presenting some options for wide-area data n etworking.

The LAN environment Most multinational offices in the Tokyo area have installed (or are now in the process of deploying) LANs (local area networks to improve intraoffice communications They have found that the ability to share files and electronic mail -- and, in many cases, to take advantage of the power futures of groupware -- is a major factor in the success of corporate communications and organizational re-engineering. Wiring every desktop in a company's "campus" can greatly enable the work and information flow, both by increasing productivity and decreasing lead times when compared with the inefficient use of paper messages and the difficulty in scheduling face-to-face meetings. (This enablement, of course, assumes that employees are information-literate and that manage ment has taken care to ensure thorough training in the use and functions of the LAN.)

One limitation of the corporate LAN, though, is connectivity to other sites. For most large companies working in the Tokyo area, the local office is either a branch (or the main) office of a larger multinational company. Business efficiency necessitate s routine (and frequent) communications and information transfer among all company locations, something that is beyond the scope of a LAN. To make the most effective use of its LAN environment, an office must be able to connect its internal network to the internal networks at the rest of their company's locations, or even outside of the network to other customer or partner corporate LANs and networks.

Such connectivity is known as a WAN (wide area network). Fortunately, there are now a variety of wide area solutions available for nearly all network applications: solutions that allow easy and inexpensive communications to nearly any point on the glob e.

The WAN environment

Wide area networks commonly come in one of three flavors:

• Leased line networks (private)
• Packet switching networks (public)
• Broadband networks (frame relay)

Each of these solutions has unique characteristics that make it attractive for different types and levels of network design.

Leased line networks

When you need a lot of bandwidth (the capacity to transmit large amounts of data in a short period of time), and you want 100% of your bandwidth to be available to network applications, the best choice for connecting your LAN to a distant site is via a leased line network. A leased line is basically an agreement between you and a communications vendor (or group of vendors) to make a permanent (24-hour) physical connection available between locations.

If you have an office in Yokohama and another office in Tokyo, for example, your company could purchase a leased line connection between the two sites through an agreement with NTT. This connection would be from your Yokohama office via local fiber or copper wire, much like that used in your telephone lines, to a local Yokohama telephone switching office. From the telephone office, the connection would follow an NTT physical path to a larger area office. This would subsequently attach through a larger trunk system to an area office near your Tokyo site, then to a local telephone office, and finally to the your Tokyo office. In this way, the LAN in your Tokyo office and the LAN in your Yokohama office would be continuously linked. If your company has mu ltiple locations, each location could be similarly connected to one or more sites via such a physical wiring system.

If you are considering an international leased line, though, you should be aware that Japanese law currently prohibits foreign communications carriers from owning physical cable within Japan. Your international leased line will require purchase of a "half circuit" from one of the three licensed international carriers (IDC, ITJ, or KDD). That carrier will then make arrangements with the distant end-country for interconnection to the desired destination location through an agreement with one of that co untry's carriers, such as British Telecom, AT&T, Sprint, or MCI. Depending on the communications providers involved, there are occasions when the customer will be responsible for arranging both sides of the international half circuits. Be warned that this can result in a very complex and expensive management and billing process.

As you might suspect, while a leased line connection offers the best performance, it also carries a hefty price tag. For many companies, the nature of their data communications network iustifies the expense of leased line networks. These networks gen erally meet the profile of very high capacity private corporate networks, running sophisticated switching and routing backbones.

But what can medium or small sized sites, whose volume and urgency do not require leased lines, do to ensure reliable connections at lower cost' The answer in this case may be connection via a switched public data network (PDN).

Starting simple, the packet switching solution

Packet switching first came upon the communications scene during the 1960s. It was originally designed as a transport mechanism to allow voice communications in a secure environment for the United States military. Packet switching was soon adopted by the US Advanced Research Projects Agency as a transport medium for its new project, ARPANet (the predecessor of the current, immensely popular Internet). Recommendation X.25 (CCITT/ITU) defines the standard for packet switching protocols, and it remains the most widely used networking protocol worldwide.

The idea behind packet switching is to segment data into packets of a fixed length (between 128 and 1,024 bytes). These packets are then transported over a "switched network" to a distant destination location. The standardized packet format allows ea ch switching "hop" in the network to check each packet for data integrity, and to pass it along to the next switch so that the packets arrive error-free at their destination.

The error-correction capability was (and is) very important due to the sometimes marginal quality of telephone lines, and the associated potential of data loss. Each packet contains such information as originating and destination addresses, a cyclic redundancy check (CRC) formula for maintaining data integrity, sequence information for the packet, and the actual data. The packets are created by a device called a Packet Assembler and Disassembler (PAD), which takes a native network protocol (such as A ppleTalk, Ethernet, or TCP/IP (transmission control protocol/internet protocol)) and encapsulates the data into a packet suitable for transmission over the packet switching network.

Packet switching networks were originally used to connect distant terminals, or clusters of terminals, to a single host. For the user, each terminal had the "feel" of being directly attached to the host, regardless of the terminal's physical location. Connections to a packet network are through either a software or a hardware PAD for each terminal, or by multiplexing a group of terminals onto a single line using a network concentrator. The concentrator combines both PAD software and multiplexing featur es into one device.

Recently, packet switching has become popular as an efficient way to connect distributed LANs in remote locations, or LANs that require limited bandwidth (from 9.6 to 64 kilobits per second), both for domestic sites (within Japan) or to international l ocations. Connection to the packet switching network can be through a router, or by using an on-site minipacket switch that allows "all-to-all" switching of traffic between LANs on the network.

Packet switching networks offer companies the option of giving their users, such as salespersons who travel frequently, the ability to connect to the company network via dial-up ports. Several network providers, such as Sprint, British Telecom, and Inf onet, provide extensive global coverage, with dial-up ports in nearly every major city throughout the world. This capability allows users to access network facilities or send and receive electronic mail from virtually any location in the world. The newer capabilities of packet networks -- using either routers or X.25 encapsulation -- are particularly attractive in countries such as Japan, which use special character encoding for kanji. The actual applications data can be run independent of the network, be cause the PAD or router will accept the native LAN protocol and encapsulate it into an acceptable network protocol (X.25) for wide-area transmission.

The WAN described above is commonly referred to as an "enterprise network." This type of network is actually an independent network, one that uses the public data network merely as a transport facility. Network control is normally done within the enter prise internal network (EINet) information management group. The packet network provider works closely with the EINet staff to ensure that wide area performance and integrity is maintained. Most public data networks (PDNs) now also offer a managed network service, allowing the EINet to "outsource" router configuration and management to the provider. The PDN will maintain 24hour surveillance on the network, as well as provide "end-to-end" configuration and troubleshooting support within the network. As com panies continue to adopt distributed networks, and juggle the personnel and training costs associated with the new technologies, the outsourcing of enterprise network management and control will become increasingly popular.

When 64 kbps isn't enough: the frame relay solution

Frame relay has been around for several years, balancing a need for high bandwidth, low-cost communications and the lag time associated with future high-speed transport technologies, such as ATM (asynchronous transfer mode). Frame relay offers near-lea sed line performance at a significantly reduced price. This is made possible by the multiplexing capabilities of frame relay switches, which efficiently allocate the bandwidth to users as required. Any unused bandwidth is returned to the network and made available for other users.

Most user and network applications are "bursty" in nature; that is, file transfer is normally not a constant, electronic mail is transferred intermittently as required, typing is not bandwidth intensive, etc. In other words, no one user will require al l the system bandwidth resources all the time. Multiplexing has reached a level where efficient switches can effectively connect many users to one leased line, with no degradation in transmission throughput (the amount of data that can be sent over a netw ork during one period of time). This enables the cost associated with the leased line to be spread among the users, significantly reducing the cost of the network connection per user.

Frame relay allows access line bandwidth from 64 kilobits per second to T1 (1.54 megabits per second). This permits users with high data transmission requirements to build powerful global networks much more economically than via international leased li nes. Japan now has several public data networks that offer both domestic frame relay service as well as international connections.

The basic concept of frame relay is a bit different than that of X.25 packet switching. Where packet switching requires error correction at each switch "hop" in a network, frame relay relies on the "pure" performance of terrestrial fiber optics. On a f iber optic network, it is assumed that there will be no corruption of the data transmitted between distant locations. Therefore, the data "frames" are checked for integrity only at the end points of the network connections (normally within a router). If a bad frame is detected, the router requests retransmission of that frame from the source router, and the data stream continues. This all occurs at a very high speed; even when retransmission occurs within a network, the users rarely notice any delay in pr ocessing.

Frame relay networks can process almost any type of data, from IBM/SNA to AppleTalk. The most common type of application seen, though, is the EINet connecting widely dispersed, distributed LANs. The function of the frame relay network in these cases is to bring users within the EINet together, giving them the ability to share resources, send/receive mail, and participate in collaborative projects using a high-speed interconnection. Most frame relay vendors offer the option of external mail connections -- for example, X.400 or SMTP, which can be transmitted to nearly any Internet or private mail system (such as ATTMail or CompuServe). This further empowers organizations to communicate both within and outside of their company data network. digital cross connects, this is not the case in many other Asian and Third World countries. If your company has a factory or branch office in Thailand or Indonesia, requiring routine data transfer back to the home office in Tokyo, a single solution such a s those described above may not be possible. While your connection from Tokyo to I-long Kong may ride on a high bandwidth, high-quality frame relay network, the connection to Bangkok will most likely be on a marginal quality, low-speed analog circuit. Thi s drives the requirement to integrate all the different types of protocols and solutions (frame relay, packet switching, leased lines) into your overall EINet design.

In addition, if your company intends to do a significant amount of bandwidth-intensive traffic to a single site, or to a location that directly supports a closely located group of bandwidth-intensive sites, high-speed leased lines can be used interconn ect major hubs. Protocol translation between different network types in this case is accomplished either within the router, or within the network switch.

The EINet and the Internet

While the main focus of this article is to present possible options for connecting your Japan office to other geographically dispersed branches (primarily within a single organization or company), it might be valuable to compare the EINet concept with other host or regional networks. The best known (and largest) amalgamation of individual networks is, of course, the IP Internet.

As the Internet continues to play a growing role in bringing all sectors of the business, academic, and research communities together, it may soon be a business necessity to have at least one gateway to the broader IP Internet. Most important will be the ability to transfer electronic mail between Internet-connected companies or correspondents. Internet access can also allow companies to freely communicate via electronic mail, remote login, and file transfer, thus extending the reach of the company t o the rest of the connected world. By having a gateway to the Internet available to the corporate EINet, even the most remote company locations can enjoy an equal amount of connected communications power, both within and outside of the organization.

An internet working glossary

Enterprise Internet (EINet). An Enterprise Internet (or Enterprise Internal Network) is a local or wide area network of internetworks, one that covers the broader requirements and wider geographic distance of a company or organization.

Internet. An internet (inter-network) is two or more networks bridged (connected) in such a way that intercommunication and resource sharing is possible among and between the networks.

IP Internet. The IP Internet (or just Internet with a capital "I") is an internetwork of internets. The Internet covers the traditional academic and research communities, commercial internets, commercial hosts (such as CompuServe, Nifty Serve, Delphi , and America Online), and EINets.

Network. For the purposes of this article, a network is defined as a group of computers or workstations that have a common server or host. Communication and resource sharing is generally confined to users or workstations within the group, or between the server and host.