Many users of the Internet enjoy organisational access via local area networks, with Ethernet connections to personal computers on their own desks, accounts on hosts with a relatively high speed Internet gateway, and in house support from their own computing and network services staff. These users have continuous connections to the Internet, open or potentially open whenever their desktop computers are turned on. However, many other users or intending new users of Internet services are located in small organisations such as schools, small businesses and local libraries, without local area networks and hosts. Typically these users connect to a remote host via daily or less frequent calls via a modem and telephone line. Often, after gaining some Internet experience, these users will be expected by their organisation to tackle the daunting task of developing an upgrade path.
How does a school, small business or local library "scale up" from using a single modem and occasional telephone connection, to multiple users and continuous connections for a reasonable number of desktop computers? What do I need to learn, how much it will cost, what equipment, what software, who will provide an Internet gateway, who to turn to for advice, how to make it all work? In particular, these kinds of questions arise with special severity for schools, small businesses and local libraries located in rural and remote regions of Australia, isolated from metropolitan suppliers by apparently high costs for telecommunications. How can we minimise that disadvantage?
This paper offers some suggestions on ways to develop small scale local area networks with continuous Internet connections, and some comments on how some of the key obstacles may be overcome. It is not a complete coverage, because this is a relatively new and rapidly evolving topic in Australia, characterised by a steep learning curve for all parties.
In this section, we outline briefly a variety of solutions for a variety of scenarios. We will assume that cost is a major consideration and that the school, library or community group is not looking at a more expensive proprietary solution such as a Novell server and network. There are many solutions and scenarios and we cannot claim to cover them all. Let us try and outline solutions that gradate in complexity and ease of implementation.
On the Web, a novice's introduction to networking concepts for a LAN is available at the Web66 pages http://web66.coled.umn.edu/Construction/Construction.html
There are several types of Ethernet connectivity and new standards are emerging, such as 100Base-TX for fast Ethernet. However, the most common standards are 10Base-2 (thin coaxial cable, which connects with BNC video-like connectors), and 10Base-T (unshielded twisted pair, UTP, which plugs in with telephone-like RJ-45 jacks). In practice, it is probably easier to connect up a LAN using 10Base-2 type Ethernet with the BNC type connectors. In that case, the cards will need to have 10Base-2 BNC connectors on them, although many Ethernet cards include both the BNC and RJ-45 connection. These may be purchased for about A$100.
Figure 1: Example of small Ethernet LAN and Internet connection
An additional point to consider if choosing to use 10Base-T (UTP), rather than 10Base-2, is that UTP will require the addition of an Ethernet hub, with a number of ports for the RJ-45 modular jacks to plug into. The number of ports can vary (8, 12 and 16 being common) and hubs may be stacked ("chained") to support a LAN with more computers than the number of ports on an individual hub. However, 10Base-T hubs are not cheap and a cost of $1000 - $2000 dollars may be incurred, though the discerning shopper might pick up a bargain by looking for second-hand hubs; this avenue applies to all that we discuss. On the other hand, 10Base-2 simply requires that the machines be daisy-chained off each BNC T-connector, with proper termination at each end of the chain. The 10Base-2 cable segments can be up to about 185 metres without the need for a "repeater" to amplify the signal, whereas 10Base-T is restricted to about 100 meters. Thus, it would be possible to have a LAN in a school's computer lab, and another machine at some distance away in the library (or indeed, another LAN in the library!).
Ethernet cabling distances are larger than those available for plain old serial lines (like your printer cable), which are restricted to about 15 metres. However, serial connectivity is cheap and we will discuss this below.
An Ethernet may be applied for PCs, Macintoshs and others such as Acorn. It is the defacto networking standard and provides very fast connectivity (10 Mb/s) between the machines on the LAN and is desirable for the use of a local server or "host".
Figure 2: Example of serial connected LAN and Internet connection
Such really cheap serial solutions can be set up in several ways, which require varying degrees of technical proficiency. "Low budget" planners could give serious consideration to using this solution, as it is possible to get a LAN connected to the net in this fashion with minimal outlay, for serial cabling is cheap and often public domain freeware can be used.
One proprietary solution implemented under the Windows environment is the LANIS (Local Area Network Internet Server) system, which is provided with both software to setup the server box and a serial card (8 or 16 ports). Only the server is required to have the Windows operating system, for example a 486/66 computer running Windows 3.1; both Windows or Macintoshs can hang off this server, using a single dialup account.
Interested readers are referred to the WWW site http://www.v-net.com/directory/lanis.html
Of interest also is the "Internet filter" that is available which can be configured to filter-out "objectionable" material. The cost of such a system is not cheap, being $US1950 for the 8 station LANIS system and $US2750 for the 16 station LANIS system.
We have not tested the above implementation but it appears to be promising, if expensive; offering simple implementation.
You can check out Vicom on the Web at http://www.vicomtech.com/vig.main.txt.html Again, costs are not cheap, with the starter pack (3 simultaneous users) priced at about $US700, and each additional user about $US50.
If you wish to use simple serial connections using Vicom's gateway product, you will need to purchase a serial card, such as the 4 port Hurdler at around $499 or the cheaper 4 port Port Juggler at $180. A 68020, or better, Macintosh is required. One down side is that this "gateway/router" Macintosh is "dedicated", ie. the server Mac cannot itself be used as an Internet-access machine - it simply routes the TCP/IP packets to the remote Internet host or to the local LAN.
One attractive feature of the Vicom product is its ease of setup and ability to route Ethernet packets. Thus, it could be used as a gateway/router for an Ethernet LAN with Macs or PCs, or both. The dedicated Macintosh should be a higher end Mac. Vicom's gateway can allow a "firewall" to be set up to protect the LAN from unwanted intrusion from the Internet, as well as acting as a dialin server allowing remote users to dial in to the gateway and access both the local LAN and the Internet.
An additional attractive feature is the ability of Vicom's gateway to allow "inbound mapping". That is, any machines or hosts on the LAN can be Internet servers, providing such services as FTP, mail and httpd for the World Wide Web. Hence, an outside call from the Internet to the gateway host will be redirected to the particular machine on the LAN that provides the appropriate service. It is virtually essential in any LAN configuration that the ability to be Internet "publishers", and not just consumers, is supported by the system.
An additional complexity may arise if the ISP account (ie. the "real" connection to the Internet) is itself an emulated one via SLIP/PPP emulators such as SLiRP or TIA, rather than a "real" account whereby your connected dialup machine has a "real" IP and hence could act as an Internet server. We are investigating further to confirm that a single account which uses slip or ppp emulation can be readily used to allow the connection of a LAN consisting of computers whose IP numbers are also dummies. This does work with the Linux set up described below, with Vicom's TCP/IP Gateway on the Macintosh, and quite likely with the MicroRouter 900i discussed below. This particular unit will be evaluated shortly and results will be reported to the conference.
For a particular LAN, a suite of IP numbers is required and these are routed by the router to the Ethernet-connected machines on the LAN. Hence, the ISP must provide not just one IP number but a set of them, and this might involve greater cost. An ISP may be unwilling to provide a block of IP numbers for a non 24 hour semi-permanent type connection. However, you never know and investigation is warranted. As discussed above, the use of PPP emulation can obviate the need for a set of "real" IP addresses and just a single user account on the Internet that uses PPP emulation such as that provided by SLiRP or TIA will suffice. Some ISPs do not provide nor permit SLiRP or TIA. However, SLiRP is public domain and it is possible for a user to install and execute it from their own single user Unix account. It is yet to be determined whether or not the MicroRouter 900i will work with PPP emulation.
In the scenario whereby a real IP number is possible for each machine on the LAN, each machine can have its own host name, such
as:
samplehost.internetprovider.com.au
or, for an additional small fee, a special unique domain name can be officially registered on your behalf, such as mydomain.edu.au. It may not be obvious but domain names cannot be assigned willy nilly by the end user but need to be officially allocated, by the ISP if their domain name is to be used, or by the authorising body if one's own domain name is to registered.
Here, any computer on the LAN can be an Internet "server" providing Web, email and ftp services to the outside Net, since each machine has its own real IP number. In addition, you may consider setting up a machine as a "firewall" to prevent unwarranted outside access to the LAN.
Another routing solution that we will look at further below is a computer running Linux, a public domain Unix, or a proprietary Unix such as a Sun or Solaris, to perform the task of routing and act as an Internet server.
However, at present installation and rental for a permanent connection is relatively expensive. One provider that is directly connected to the Internet backbone gives pricing of $5,000-6,000 annually for a 28.8 kb/s dialup connection, plus set up costs of about $1,500, plus the costs of a modem at both ends and a small monthly modem "housing" cost. This would provide a dedicated line always available when needed. If you intend to set up an Internet host, such 24 hour connectivity is necessary. In addition, included in this setup cost is the allocation of a set of permanent IP numbers and domain name registration.
An ISDN semi-permanent is of course even more costly, with similar setup costs, but an annual cost of around $10,000, in addition to the incurred Telstra costs of connection to the ISDN terminal server, which will vary according to geographical distance. If say the above MicroRouter 900i is used, an additional terminal adaptor (TA) will need to be purchased. If say it is desired to connect a Unix box to ISDN, again a terminal adaptor and ISDN card will be required.
Linux is one public domain freeware implementation of Unix suitable for 386 or 486 PCs, and NetBSD is another. A "port" of Linux to the Macintosh is in progress. Linux permits full TCP/IP networking, including routing and gateway functions. Getting Linux up and running and networked is not for the faint hearted, but it is immensely satisfying to set up, configure and get networked.
Once installed, the use of DIP scripts enable the automatic connection of the Linux/Unix box to the ISP account. A LAN is easily connected using the above 10Base-2 BNC-T type connections and it can include Windows PCs and Macs running the usual public domain TCP/IP network information retrieval tools; or 10Base-T can be used with the addition of an Ethernet hub.
There is ready information of the Net on setting up Linux and networking over TCP/IP. In addition, many good books are available.
One beautiful implementation is connecting a Linux box to an ISP that provides SLIP/PPP emulation, such as the freeware SLiRP or proprietary TIA ("The Internet Adaptor"). In this case, a single account can be used to provide access to the machines on the LAN, which are set up using "dummy" IP numbers. For further information, read http://www.cnrcoll.nf.ca/~andy/slirp/slirp-LAN.html or the documentation for SLiRP. Without SLiRP or TIA, the remote LAN will need to use "real" IP numbers for each of its computers and these will need to be provided by the ISP to you and some routing tables to be set up at the ISP's end as well as the Linux/Unix end, which may or may not cost extra.
Alternatively, if one's single ISP account is the usual "real" SLiRP or PPP (although an increasing number of ISP's are opting for the use of SLiRP themselves as their default SLIP/PPP service), one can set up two Linux boxes - the first connects to the ISP and has a "real" IP but in addition is running the public domain SLiRP on it, such that the second Linux box can connect to it (via say SLIP) and then the LAN (on Ethernet) can connect to the second Linux box. Again, the user is referred to the above Web reference.
If the budget is not a limiting factor, an "off-the-shelf" pre-configured Unix box can be purchased for A$6-8000. The hardware provider may pre-configure the necessary TCP/IP networking parameters for your LAN as well as establish dialup connectivity to your ISP.
These Unix boxes can be used to provide the standard Internet services, WWW, email, news and FTP. However, such services are provided usually only where the host is permanently connected to the Net.
A Linux/Unix computer can be used to provide connectivity to a serial-based LAN using SLIP or PPP to connect Windows PCs or Macs to it, via a multi-port serial card on the Unix host, instead of using an Ethernet LAN. Serial cards for PCs (running Linux) are much cheaper than those for a Mac, an 8 port card costing under $200, and a 4 port card being about $70. The Linux box can be configured to provide SLIP or PPP service to each of these ports. If the SLIP/PPP emulator SLiRP is running on this Unix host, network PCs can connect using TCP/IP software such as Trumpet Winsock and "proxy" IP numbers. This solution is very similar to the commercial LANIS system which also enables uses a single dialup account to connect a LAN to the Internet.
However, there is a fundamental key to enabling lower cost Internet access for rural and remote regions. This is "IP traffic aggregation", the loading of a "leased line" network as heavily as possible with the largest possible number of users and transporting their traffic as Internet Protocol data traffic, not as telephone traffic. Within a given leased line capacity, the larger the number of users, the better the cost per end user can be minimised.
The number of local users "loaded" upon a single leased line and the total user base which may be supported cannot be specified precisely. The estimate of 15 to 30 concurrent users on one leased line may relate to 500 to 1000 as the total user base. The main factor is users' and management's perceptions of the optimum. With more users, the cost per user is lower, but there is a degradation of service as more users are added, arising from heavier competition for a limited number of modem connections, and slower response times when a leased line is too heavily loaded with too many concurrent users.
To extend the example, consider a regional centre which is between 800 and 2000 km distant from a capital city Internet connection point. Suppose that this centre has a demand for 120 hours of user connection per day (eg, 120 users averaging 1 hour per day, or 360 users averaging 20 minutes per day, etc) The comparison between data transport as 1. telephone traffic, or 2. Internet Protocol traffic via a local host or terminal server, is as follows:
Telstra has no good reason to engage in the highly competitive "retail market" aspect of Internet access. Telstra is unlikely to become a provider of local modem pools, because the work involved tends to be closely related to user support and services areas in which it has little expertise deployed at rural or remote centres.
A "local terminal server" and attached modem pool provides an efficient way to aggregate traffic. The terminal server converts modem traffic into Internet packets for digital transmission on a leased line to an Internet gateway, thereby enabling sharing of leased line capacity by concurrent users in number up to the size of the modem pool. Some of the larger private retailers of Internet access adopt this method, which enables them to work with a central host providing its services nationwide. The ADEnet Project also provides a terminal server approach, using existing Internet connections and permitting traffic to any of the participating universities (Atkinson, 1995).
A local terminal server may cater for several types of connection, including users wishing to have their individual modem traffic transported to hosts elsewhere on the Internet as permitted by the terminal server (individual users hang up upon completing their work), and users of a local host or LAN which connects via modem, either by a leased line connection or by an intermittent dialup connection kept open as long as its users are active. For example a school or business or government agency may have its own local host, which may be part of a local area network (as described in the first section of this paper) and may even have its own local modem pool.
A local host provides local users with a login name and password, with access to a range of Internet services. Thus a user may have a local email address and then will not require a login name and password for a host elsewhere. For example, a university distance education student with an email address on a local host does not need an email address on one of the university's hosts.
The relative merits of "local terminal server, remote host" and "local host" are difficult to assess on technical grounds. Either approach will fulfil the primary goal, aggregation of IP traffic to minimise the costs of long distance carriage, and the two approaches may co-exist as indicated above. However, support for a local host and staff known to the regional community, businesses and schools may be more readily mobilised than support for use of a remote host. That factor may be vital in recruiting a sufficient number of users to reduce the cost per user down to attractive levels for rural and remote regions.
Local terminal servers or hosts in rural and remote regions may face high costs for a long distance leased line to a Telstra Internet gateway in a capital city. Thus there is scope for regional gateways, located in major non-metropolitan centres to aggregate traffic from smaller centres in adjacent regions. For example, consider 20 local terminal servers or hosts each having its own leased line all the way between its region and a capital city. If these can be replaced by shorter distance, larger capacity leased lines to a major centre, connecting to a capital city via a single shared leased line, costs will be reduced in most cases.
In many cases, Australia's rural and remote regions may have computer network connections operated by state government departments and by corporations. These may provide a basis for regional aggregation of Internet traffic, but typically these capabilities are exclusive to each department or corporation and are not made available to others. For example, in Western Australia the major corporations in Western Australia's mining centres have private networks, some extending world wide, but with no regional aggregation of traffic except for their own cost saving purposes. The major departments in the WA State Government at present each have their own leased line networks extending to major regional centres, with virtually no regional aggregation of their traffic. However, WA State Government steps towards improved management of telecommunications, rapidly increasing awareness of the advantages of Internet communications and the lack of Telstra gateways outside the capital city of Perth may initiate a State role in regional aggregation of traffic. This would reduce Internet connection costs for WA's rural and remote regions, at least via educational and community agencies which can relate readily to the university sector, including schools, TAFE colleges, public libraries and local government.
Unfortunately, there are considerable political and social obstacles to local action prevailing over capital city inertia in relation to the Internet for rural and remote communities. For example, at present it is unlikely that any single State government agency will take an initiative these cases and become a gateway reseller to the rest of the community, although in Western Australia's case, Bureau Services is one possibility. Nevertheless, united community pressure eventually will bring low cost Internet access, comparable to capital city prices, to rural and remote users. Sustained, widespread local interest in the Internet, even if that requires several years of expensive practice using STD calls to a metropolitan provider, will help considerably to hasten the arrival of low cost access for rural and remote users.
Atkinson, R. (1995). Internet access for schools via cleo.murdoch.edu.au. In R. Oliver and M. Wild (eds), Learning without limits. Proceedings of the Australian Computers in Education Conference 1995, Perth, 10-13 July, Vol.1, 99-109. Perth: ECAWA. http://www.users.bigpond.net.au/atkinson-mcbeath/roger/pubs/confs/atkinson-acec95.html
Atkinson, R. (1995). ADEnet Project. http://cleo.murdoch.edu.au/asu/edtech/net_access/adenet.html
Telstra (formerly Telecom Australia) (1993). Prices guide. Melbourne: Telecom Australia. (Note: At the time of writing, details of 1995 decreases in ISDN pricing were not available to the authors).
Le Roux, Graeme (1995). AusComms rates remote access solutions. Australian Communications, Nov 1995, 63-76.
Rehn, G. (1995a). The Macintosh and Dialup Access to the Internet: Implementing the Graphical User Interface. In Technology 95: Proceedings of the 1995 Apple University Consortium Academic Conference, Perth, 3-6 July, 245-255. Perth: Apple Australia. http://cleo.murdoch.edu.au/asu/edtech/pubs/rehn/auc95/auc95.html
Rehn, G. (1995b). Playing with TCP/IP: having fun on the Internet. In R. Oliver and M. Wild (eds), Learning without limits: Proceedings of the Australian Computers in Education Conference 1995, Perth, 10-13 July, Vol.1, 237-246. Perth: ECAWA. http://cleo.murdoch.edu.au/asu/edtech/pubs/rehn/acec95t/acec95t.html
Rehn, G. (1994). Software tools for dialup Internet access. In J. Steele and J. G. Hedberg (eds), Learning Environment Technology: Selected papers from LETA 94, 259-269. Canberra: AJET Publications. http://www.aset.org.au/confs/edtech94/rw/rehn.html
Authors: Geoff Rehn, Lecturer in Educational Technology rehn@cleo.murdoch.edu.au Roger Atkinson, Senior Lecturer in Educational Technology atkinson@cleo.murdoch.edu.au Voice: +61 8 9360 6840 Fax: + 61 8 9310 4929 Academic Services Unit, Murdoch University, Murdoch WA 6150. Australia Please cite as: Rehn, G. and Atkinson, R. (1996). Remote access to the Internet. In C. McBeath and R. Atkinson (Eds), Proceedings of the Third International Interactive Multimedia Symposium, 345-354. Perth, Western Australia, 21-25 January. Promaco Conventions. http://www.aset.org.au/confs/iims/1996/ry/rehn.html This article (including an addendum) is also available at: http://cleo.murdoch.edu.au/asu/edtech/pubs/rehn/imms96/imms96.html |