About This Chapter…

This chapter of the Get Connected `96 focuses on background information all UAs should know about the FAS Network, the Internet in general, and what users need to do to get connected.


Keep in mind that you aren't expected to become an expert in networking minutiae and protocols - although that would certainly be nice! What we do expect is that you learn the fundamental concepts of what a network is and how it works, and what parts of a connection do what. We are more interested in your ability to apply broad concepts to challenging in-room situations. You've been hired because you're a problem-solver - we expect all UAs to be able to diagnose and solve a broad array of problems on-site by looking for the patterns described in this section.


Much of the material here has been simplified from Rick Osterberg's seminar presentation entitled "The Ins and Outs of the Harvard Network," which was presented at the April 17, 1995 General Meeting of the Harvard Computer Society.

Section I: An Overview Of Networking

1. What's A Network, and How Does It Work?
2. The Internet: A Network of Networks

Section II: The Four Keys to Successful Networking

1. The Ethernet Device
2. Device Drivers
3. Connecting to Local Services
4. TCP/IP: The Lingua Franca of the Internet

1. Running NetConnect
2. Turning on the Jack
3. Partitioning

Section I: An Overview of Networking

Section I.1: What's A Network, and How Does It Work?

The purpose of a network, plainly and simply, is to allow the sharing of data among computers. In reality, of course, things are much more complex. First, we must differentiate between the FAS Network and the Internet at large.


The FAS Network is a Wide Area Network, or WAN, which consists of many smaller Local Area Networks (LANs). In a rough sense, the purpose of the FAS Network is to connect every dormitory room and staff office to the central computing facilities located in the basement of the Science Center, and in turn to the Internet. Data is transmitted on networks in the form of packets - little chunks of information that are sped along the network to their destination.

The process begins at each datajack, the physical plug in a room which users connect their computers to. The datajacks connect our users via 10Base-T Ethernet, a common form of networking cable that uses RJ-45 plugs, to the FAS Network. Be careful: these look a lot like phone jacks, but the cables aren't interchangeable!


The wiring from each datajack in an entryway or hall connects to a hub, usually in the basement of the building. Usually, between 100 and 200 jacks are connected to any one hub. The hub is a device that takes all the packets headed to and from these jacks, "concentrates" it, and sends it out via the fiber optic backbone to the nearest router.


A router is a "hub for hubs," so to speak, taking all this data from dozens of hubs and sending it on to its next destination. There are six main routers on the FAS Network, servicing areas that include the Yard, the River Houses, the Quad. Other routers are used in the Science Center and in the northwest and northeast areas of the campus. These routers take all the packets they receive and send them further along towards their destination, either to another hub or to the main router in the Science Center, where the packets are passed along to their destination. Within the FAS Network, the destination might be another user's computer, or a connection to the mail server to bring new mail into Eudora, or perhaps a telnet session to one of our Unix machines.

Section I.2: The Internet: A Network of Networks

Of course, not all packets from a user's computer are headed to another location on our WAN; many are headed out into the Internet, a system which connects LANs and WANs around the world.


The FAS Network run by Harvard is just one of the many millions of networks that comprise the global Internet. The Internet is often, and perhaps best, described as a "network of networks," and this isn't too far from the truth. Initially conceived by the United States Military as ARPANET, the Internet was designed to allow data between computers and scientists at research universities, government labs, and military installations to continue to flow even in the event of nuclear war. (Despite this resiliency, however, technologies like RealAudio and Shockwave have shown a remarkable capacity to bog the Net down!) Later, the ARPANET/Internet passed into the hands of the National Science Foundation, and was finally "privatized" in the spring of 1995. Contrary to expectation, neither the privatization of the Net nor the entrance of 6 million AOL users onto it have caused any crashes or major disruptions of service.


The central and main features of the Internet are:


1. It is packet-switched, meaning that its central function is to transmit and route data via the packet;
2. It lacks centralization, in that there are many ways to transmit data from one site to another, since there is
3. No single dedicated wire from point A to point B. Keep in mind, however, that
4. Packets can be dropped or corrupted, making the Internet ultimately
5. A non-reliable system with no guaranteed delivery.


We receive our Internet feed from Harvard's University Information Services division (UIS, formerly called OIT). UIS manages Harvard's gateway onto NEARNET, a commercial Internet service provider. When a student tries to access the Internet from their room, packets follow the same path to the router we described for FAS Network traffic. The difference comes at the Science Center router. Instead of sending packets headed for the Internet back into the FAS Network, it sends these packets out to another gateway at William James Hall, where a T1 microwave link connects us to MIT. MIT transmits the packets via another microwave link to the top of the Prudential Center, which sits right on top of the NEARNET backbone. From there, the packet heads out onto the Internet, eventually (after passing through many other routers) reaching its destination. All packets headed out to the Internet use the TCP/IP protocol, described in depth later in this section.


As you go through training and this document, and your job, keep in mind the differences between the FAS Network and the Internet, and the different services available on each of them.

Section II: The Four Keys to Successful Networking

In the previous section, we described the FAS Network and the Internet and how they work in a rough sense. So the next question is, how does a user connect to these networks from their dormitory room?


To answer this, it is necessary that you keep in mind that there are four separate components to a connection to the network at Harvard. Assuming the user has activated their jack via NetConnect, every connection that they do, or that we do for them in an in-room assistance, requires all four of these elements to be working:

• Getting the Ethernet device physically installed or working;
• Installing or enabling the drivers that allow the computer and the Ethernet device to communicate;
• Enabling the appropriate networking protocols so that the user can access local software servers, networked printing, and the like;
• Configuring TCP/IP, which gives users access to the World Wide Web, telnet, ftp, and the Internet.


Our job in the field and in providing support is to make sure that all four elements are in place and working. If you understand what these elements are, and how they work, you will be able to diagnose and troubleshoot problems more efficiently!



One important thing to keep in mind: this four-part model of networking is a little simplified. In reality, there are ISO layers, virtual and physical links, medium-access protocols, and the like to worry about. This paradigm has been designed to assist you in understanding what needs to be done in order to get users' machines working on the network; for a more complete discussion of these issues, we recommend CS 143 as a good introduction.

Section II.1: The Ethernet Device

Users connect to a network using a Network Interface Card (NIC), also known as an Ethernet card or device. Think of it as a modem for high-speed networks - it is the bridge between the computer and the network. PC users will probably have to buy this device: a PCMCIA card for notebooks or an ISA or PCI card for desktops. Some PCs now come with Ethernet cards included internally, such as many HP and Dell machines. Most new Macintosh computers come with Ethernet capability built into the system, but may still require a transceiver to allow them to plug into our network.


Every Ethernet device possesses a unique address, which is usually in the form XX-XX-XX-XX-XX (12 hexadecimal digits, with hexadecimal being the numbers zero through nine and the letters A through F). The first six hex digits are known as the vendor prefix--this prefix denotes the manufacturer of the particular Ethernet device. The remaining six hex digits are unique to that device. If a card ever reports fewer than 12 digits, append zeros to the beginning of the address to find its "real" Ethernet address. (For instance, a card reporting 80AABBFF has an Ethernet address of 00-00-80-AA-BB-FF.) Be careful: zeroes, eights and the letter "B" often look alike.


Installation procedures vary from system to system, and you can find information on how they work for various Ethernet devices and systems in other chapters of this manual. In general, though, you should be very familiar with the inner workings of a computer: how to remove the cover to the computer, how to insert and remove expansion cards, how to ground yourself to avoid damaging the system, and so forth.


Another important aspect of installing the Ethernet device is checking the physical connection from the card to the datajack. Make sure the cable is 10Base-T Ethernet, and not a phone cable. Most 10Base-T cables are much thicker than phone cables, and with good reason: while telephone cabling uses 4 wires, Ethernet uses 8. Some cables can be tricky, however; cables for portable Ethernet devices are often designed to be phone-cable size for convenience of transportation. Have a look at the plugs on these jacks - no matter what the diameter of the cable, an Ethernet cable will always have a wider plug, and you'll see more wires inside it. Also be sure the datajack is free from dust or other nuisances that could prevent the connection from working!

Section II.2: Device Drivers

Software in the forms of drivers, control panels, and stacks allows the user's computer to communicate with this Ethernet card and the network. If you have ever installed a sound card, CD-ROM drive, or other peripheral inside your computer, you know that "sticking it in" isn't enough. A computer will only communicate with a peripheral, like an Ethernet card, if it knows that it's there.


Under Windows 95, this is accomplished through the Add New Hardware wizard, the System Control Panel, or simply inserting a Plug and Play device into the machine - point Win95 in the right direction, and it will install drivers from CD-ROM or floppy disks.


In DOS, an "ODI driver" is required for a computer to recognize the Ethernet device. For the 3Com cards, the driver is something like 3C5X9.COM, 3C589.COM, or a similar file. Usually, these programs will be between 15K and 60K, and will contain the name of the manufacturer in them, like MHZLAN.COM for Megahertz Ethernet cards or so forth.


Macintosh computers that come with built-in Ethernet devices have this software installed already. Older Macs and those which have PCMCIA Ethernet devices usually contain a diskette that installs all of this software for you.


Chapters 2-4 of this manual cover this information in more depth.

Section II.3: Connecting to Local Services

Beyond this, additional software is required to translate and communicate in one of the various "protocols," or languages, of a network. Think of the wiring of a network as the terminal of an international airport. Passengers are trying to get from place to place in the same building, but they speak many different languages that keep them from communicating with each other, even as they mingle and pass through the hallway. Similarly, a network doesn't care what protocols are being transmitted over it, but the receiving or transmitting computer needs to be able to understand.


Many of our network services - like access to General Software and Novell servers, networked laser printing, and so forth - use platform-specific protocols. For the PC, variations of Novell's IPX/SPX protocol are popular. The equivalent protocol on the Macintosh is known as AppleTalk/EtherTalk, which fulfills much the same functions as IPX/SPX. General Software, for example, works by using AppleTalk to connect to the appropriate server in the Science Center.


Both of these protocols work perfectly within the FAS Network, but their packets get blocked at the Science Center gateway before reaching routers outside Harvard.


Knowing the difference and functions of LAN protocols is critical for diagnosing networking problems. If a Macintosh is seeing the General Software server, for instance, but can't connect to the Web or run telnet, you know that the Ethernet device and its drivers are working fine, as is AppleTalk - but something is wrong with the TCP/IP settings on that machine. Same thing for Windows95 machine: if other computers show up in the Network Neighborhood window, but telnet doesn't work, then there is an error with the TCP/IP settings, and not with the Ethernet card itself.

Section II.4: TCP/IP: The Lingua Franca of the Internet

The main protocol on the FAS Network is the Transmission Control Protocol/Internet Protocol, or TCP/IP. TCP/IP is a universal language on the Net, spoken by PCs, Macintoshes, Unix systems, and pretty much everything else.


TCP/IP's historical use has tended towards applications involving the Internet; telnet, ftp, and Web applications have always used the flexibility of TCP/IP for their purposes. Whenever a user telnets to fas.harvard.edu or surfs the Web at yahoo.com, for instance, they're using TCP/IP. In addition, many newer applications that might have used IPX/SPX or AppleTalk in the past are now taking advantage of TCP/IP; the KeyServer software installed by FAS Computer Services on student machines, for example, uses TCP/IP to validate the user.


Without getting into too many technical details, keep the following general rule in mind when diagnosing a user's computer: IPX/SPX, AppleTalk, or TCP/IP can be used within a WAN like the FAS Network, but only TCP/IP allows a user to access the Internet.


Every computer hooked up to the Internet has its own IP address, a 32-bit numerical address, usually written in the form xxx.xxx.xxx.xxx where xxx is from 0-255, such as 140.247.30.30 (fas.harvard.edu). IP addresses can be broken down by network, subnet, and node. The first two set of numbers usually stand for the Class B network: for example, Harvard has two Class B networks (128.103 and 140.247). Within each of those larger networks, there are many smaller subnets, denoted by the third set of numbers: for example, many of the computers in Dunster House have IP addresses starting with 140.247.159, while computers in Eliot House have IP addresses starting with 140.247.171. The first two numbers are the same, because both subnets are on the same network. The third number is different, denoting that the two Houses reside on different subnets. The final set of numbers denotes the actual computer (node) within that subnet, hence the address 140.247.159.251 points to a specific computer within Dunster House, and 140.247.171.251 points to a specific computer within Eliot House.


Individual computers on the FAS Network find out what their IP address is without user or UA intervention from the network through dynamic addressing using one of two technologies: Bootp or DHCP. Users should never "hard-wire" their current IP address into their computer. On the FAS Network, the subnets may need to be reconfigured at any time, which can make users' IP addresses change. If they do, and a user has manually entered her IP address, they'll be in for a rude surprise: their IP address will be invalid, and without a valid IP address, no TCP/IP connections can be made. IP addresses also change when a student moves from one room to another. It's always best to let the Science Center servers grant IP addresses dynamically.


Bootp is a method of dynamically assigning IP addresses on the network. Currently computers running DOS/Win3.1, MacOS, and a few other operating systems use this method to obtain their IP address. Bootp uses a central database that matches Ethernet addresses to IP addresses. So, when a user's machine starts up, it sends out its Ethernet address (which never changes) onto the network, and the Bootp server returns the assigned IP address. This way, FAS Network Support personnel must only make sure the Bootp tables are up-to-date, and do not need to worry about notifying 3,000 different users individually if IP addresses change. Windows 95 uses DHCP instead of the Bootp protocol; it fulfills the same goal of dynamic IP addressing using slightly different and improved technology. From the perspective of an end user (and User Assistants), they are indistinguishable.


It would be cumbersome if you had to know the exact IP address of an ftp server or a Web site that you wanted to connect to - remembering 12-digits numbers isn't easy! But you don't have to - each machine on the Internet can have one or more symbolic names mapped to its IP address. Examples are fas.harvard.edu, which points to 140.247.30.30, www.das.harvard.edu, which points to 128.103.60.2, etc. Symbolic names are meant to make it easier, so you dont have to remember IP numbers. For students on the FAS Network, their symbolic name is in the form of username.student.harvard.edu. This remains their machine's symbolic name through all their years here; the IP address that the symbolic name points to, however, is modified whenever necessary, as determined by the machine's physical location on the FAS Network, by FAS Network Support.


This is accomplished by updating the nameserver on the FAS Network - the server which keeps a list of mappings between symbolic names and IP numbers. The nameserver's job is simply to take a symbolic name and return the IP address associated with that symbolic name back. It does this not just for symbolic names on the FAS Network, but also machines on the Internet. For example, when a user types in the URL "http://www.yahoo.com/" in Netscape, Netscape sends out a request to the nameserver asking for the IP address that corresponds to www.yahoo.com. The nameserver returns 205.216.146.105, which is what Netscape needs in order to connect to the Yahoo web server and start displaying the web pages located on that server.


---

Section III: Getting Connected!

Section III.1: Running NetConnect

The process of requesting a network connection begins with a program known, quite logically, as the NetConnect program. (accessible by logging in as "netconnect" on fas). The software will probably migrate to the Web later this year; for now, it is still accessed through telnet. There are three essential pieces of information a student needs to have when he/she runs NetConnect:
1. Student identification and housing location
2. Datajack number
3. Ethernet address of the network card, for network management purposes


Once this information is entered by the student, it is verified by FAS Network Support, usually automatically.

Section III.2: Turning on the Jack

The datajack is actually activated in the next step. If the datajack is already physically patched (ie. connected) on the hub ("quick connect"), then FAS Network Support need only turn on that port on the hub. Through automated procedures, this generally happens that evening, although it will likely take longer during busy periods like Get Connected. If the datajack is not patched, then a Network Support person must physically travel to the hub in the basement of the user's house/dorm and physically install the patch cord. Due to scheduling concerns, this can take a day or two. Jacks in Harvard Yard, the Quad, and the River Houses will definitely be pre-patched for quick-connect for the Fall of 1996. Keep in mind that if two computers register for the same jack using a splitter, a special patch has to be installed by FAS Network Support, which means that the connection will take longer to complete.


Once the jack is patched, more automated scripts add the student's entry to the nameservers (in order to allow translation of username.student.harvard.edu into 140.247.xxx.xxx) and add the appropriate entry to the bootp server (in order to facilitate translation of the student's Ethernet address to the corresponding IP Address). The connection is then ready to use, and the student is usually notified by e-mail.

Section III.3: Partitioning

So what happens if a user plugs into their jack before they get this E-Mail notification? Their jack gets partitioned. A partitioned jack is one whose network communications are being blocked by the hub it is connected to, and hence can no longer transmit information to any other computer on the network. The Ethernet hubs are intelligent and watch for activity on each connection that could potentially be detrimental to the network - as well as any packets coming from unregistered Ethernet addresses! Partitions may be caused by the following:
• Invalid Ethernet address: You may encounter this during an in-room assistance. If the user entered a Ethernet address into NetConnect that does not exactly match the Ethernet address of the user's Ethernet card, then the jack may become partitioned as soon as the card attempts to transmit packets on to the network. An improperly entered Ethernet address is one common reason why a network connection does not work properly.
  
• They haven't run (or finished running) NetConnect: This goes along with the problem of having an invalid Ethernet address, but is still a common source of frustration for users. Before helping someone out, always make sure the user has received E-Mail telling them their network connection is activated before trying to help them.
  
• "Trash" being sent through the Ethernet connection: This could mean that the user's Ethernet card is broken, the user's Ethernet cable is defective, or that the physical wiring of the datajack is faulty. The connection is turned off because forwarding this network trash onto the network ties up the network resources unnecessarily.
• They plugged into the wrong jack: Obviously, a user who plugs into a jack that's not registered to him or her will find their computer partitioned, since the hub won't recognize their Ethernet address as valid. Always make sure users are in the right jack!


No matter how hard you try, a computer will not work on the FAS Network if the jack is partitioned! When talking to or working with a user whose connection doesn't work, therefore, you should always take care to make sure that these common causes of partitions aren't at fault.


Partitions are cleared peridoically by FAS Network Support's computers; in most cases, a partition will be removed within a timespan of a few seconds to a few minutes. However, if a computer continues to partition many times in a row, the jack will stop unpartitioning, and FAS Network Support will be notified.