The hardware required by our company to be an ISP will be quite costly.  Remembering how expensive the costs for leasing a T1 line and the other analog phone lines will cost, you will see that the hardware expenses will easily rival those costs.  We will need to make some decisions on whether to lease this equipment, or to buy it outright in hopes of soon offsetting the high costs of ownership with customer account revenue.  We will attempt to take the reader in a step-by-step process of the order in which a dial-up customer's request for service would be received, and my which piece of equipment.  It should be noted that our customers will have to have software running on their computer that implements TCP/IP and the Point-to-Point Protocol (PPP). 

Pictured here you will find a diagram of a typical ISP and its hardware.  The one difference between the diagram and our ISP configuration is the Firewall / Router that is pictured will just be a plain old router in our configuration.  The reason we will not use a firewall is because there is little differentiation from the pimple-faced geek hackers out on the internet and the pimple-faced geek hackers who happen to be ourcustomers.  If someone wants to come in and wreak havoc on our network, the most likely candidates would be our dial-in customers, so keeping the "bad guys" out of our network with a firewall does not make much sense. 

When our customers dial our ISP's phone number to gain access to the Internet, the first piece of equipment to receive the signal on our end will be a 32-port modem stack, capable of rolling over to an unoccupied modem port if the first port is already taken.  This stack of modems is directly connected to the next piece of equipment, the Remote Access Server, using standard serial cables.

We will need a Remote Access Server to handle the incoming customer requests for service.  This piece of hardware is sometimes called a "Dial-In Server," or a "Communications Server" because it handles password authentication (though we will use a RADIUS server for this task), shell access (if we give our customers this option), and even acts as a primitive IP router.  The RAS also dynamically assigns an IP address to the customer, so they can receive IP packets through our equipment.  It may be apparent to the reader that such a server has to have a large capacity for multitasking, as this server will have to be able to handle the request of dozens of customers simultaneously.  We need a piece of hardware that is durable and robust.  The Lantronix LRS-32F pictures above gives us all of the features we want, and even lets us interface with it using a browser from any one of our desks;  all you have to do is enter its IP address and you are inside the box.  The model show has 32 ports to match the 32 phone lines in our start-up phase.  As we grow, we can easily add components to this machine to add necessary ports to handle additional modems.  the RAS will be directly attached to our 10BaseT Ethernet, via an RJ-45 connector, so in effect our dial-in customers will be gaining access to our internal network before subsequent hardware routes them to the global Internet. 

We will need to offer out customers a number of other services besides just internet access.  To be competitive, we will want to offer them such services as the ability to use FTP software.  Also, our customers will want to be able to send and receive email, so we will have to employ SMTP (Simple Mail Transfer Protocol) and POP (Post Office Protocol).  Our customers may also want to take advantage of the plethora of Usenet groups around the world, so we will need to offer Network News Transfer Protocol (NNTP).  In addition, many ISPs are offering customers 5 megabytes of hard drive space on their servers for hosting Web sites for their customers.  Besides our customers' needs, we will need one of our core servers to act as a Remote Authentication Dail-In User Server (RADIUS). As the name implies, this RADIUS server will authenticate our dial-in customers passwords for our system. Also, a core server will need to act as our Domain Name Server. How do we do all of this?  Ideally, we would employ a server to handle each of the tasks.  But in the start-up phase of our company, we need to be the most cost-conscious and so would use only two servers to perform all of these functions.  We need two servers for two reasons.  First, we want to build a redundancy into our system.  In case one server were to break down, the other would be able to handle the load until we get the other one up and running again.  Second, we hope to have so many customers that we will need the extra processing power to handle all of the tasks demanded by them. 

We have chosen the Digital Alpha Generation 2100 as our server.  Each will have 64 bit RISC processing running at 533 Mhz, a SCSI connector for the hard drives and easy scalability, and 384 MB of main memory.  It may seem odd that we will employ a "workstation" as a "server." Actually, this is a common practice for ISPs of our size. If we put two 4GB hard drives into each of these servers, with a RAID setup, we can easily attain our goal of redundancy. Relatively speaking, these servers are not too expensive, and we can most certainly find some decent models on the resale market. Our hardware expert has worked with this model before and have found itto be topnotch.  These servers will have the necessary Network Interface Cards so they can be attached to the 10BaseT Ethernet. 

When a customer has dialed into our network, he will certainly want to do something useful with his time.  He could use a browser and explore the World Wide Web.  He could send email to his friends.  He could even download a new game using FTP.  How is this all made possible?  In two words:  the router.  The router is the piece of hardware that makes the internet work.  All across the globe, routers are presently routing packages of data, and we will need one for our ISP as well.  When a customer send email, for instance, that message is broken up into smaller pieces, called packets, by the TCP/IP software.  These packets are sent through our system to the router, which has lengthy routing tables telling the router which is the best path on which to send the packets out on the Internet.  The reverse is also true.  If a customer wants to access a web site, the packets that make up that web site will travel through our router on their way to the customer's machine.  Cisco is the world leader in router technology, and the model shown is similar to the one we will have to purchase. 
 

A Channel Service Unit / Digital Service Unit, or CSU/DSU, is a piece of equipment whose function is similar to that of a modem.  It can be thought of as a really BIG modem that connects our ISP to our telecommunications company, Sprint.  Just as when a customer uses a serial port connected to a modem connected to a phone wire to dial into our network, our Cisco router is connected via a V.35 cable (similar to, but more sophisticated than, a serial cable in a modem) to the CSU/DSU, which is connected to the T1 (a huge phone line) which connects to Sprint.  The Kentrox CSU/DSU pictured above gives us all of the multiplexing capacity we need to run our ISP. 

With all of this expensive, exotic hardware to buy, personal computers seem to almost be an afterthought.  Still, we will need at least two to start up our company:  there will be accounting, word processing, programming, billing, and network maintenance to be done.  There are man theories on what brands of PC to buy, and how powerful a PC should be.  Our hardware expert has had good experience with Gateway 2000, so we will go with that brand.  They are relatively inexpensive and have all the functionality we will need.  Being aware of Gateway 2000's poor reputation for technical support, we allay that concern with in-house techies.  TwoGateway 2000 E-3000's with 200 MHz processors should be more than adequate.  The $1199 price for each includes a monitor, a 4 GB hard drive, 6 expansion slots, and 32 MB of RAM.  These will be more than adequate for our business.