NET 102 - Networking Essentials II

Chapter 4, Ethernet Basics; Chapter 5, Modern Ethernet

Objectives:

This lesson introduces the student to some basic knowledge and some history about Ethernet networks. Objectives important to this lesson:

1. Explain what an Ethernet is
2. Describe specifics about various versions of Ethernet networks

Concepts:

Chapter 4

The author begins chapter 4 with a history lesson that does not mean much without a context. This chapter is about Ethernet, a set of rules for running a network that was invented at Xerox PARC, their Palo Alto Research Center. Most histories give credit to Bob Metcalf as the inventor.

Back in my notes for NET 101, I gave the students a definition of networking. Networking can be defined with five features:

• users sharing resources (like printers or files)
• across a common medium (like copper wire or fiber optic cable)
• by way of specific rules (like TCP/IP or other network protocols)

Somewhere along the way, you are expected to learn that networks typically act like client-server or peer-to-peer networks. Let's review those ideas:

If entities on a network act as peers, sharing all their local resources with each other, then this is Peer-to-Peer Networking.

If entities act in strictly defined roles, as either servers or clients, but not as peers, then this is Server-Centric (Client-Server) Networking. Most PC networks are this type because peer-to-peer does not work well if you have more than ten hosts..

Most networks follow a client/server model. Clients typically perform some or most of the processing on the network, while servers provide services like data storage, instead of providing all the computing power. Client/server networks are typically easier to upgrade, both on the client side and on the server side.

The text move on to physical topologies. Topology is the study of shapes and configuration. Physical topology is the way a network is wired (or wireless-ed?) together. Logical topology is the way it works, regardless of the wiring. Network configurations typically fall into one of these types:

• Bus - essentially one continuous cable for each segment that all the devices share; in the early days, they ran one length of coaxial cable through the office and all the workstations used an adapter to tap into it; in the image above, a line is a special case of a bus
  
  • Easy to install
  • Moderately difficult to reconfigure. This is because it is difficult to add new devices or move existing ones without sufficient room to tap into the bus.
  • Difficult to troubleshoot
  • Units affected by media failure: All
• Star - starting at hubs or switches with individual cables radiating away from them for each node, like the picture at the top of page 68
  
  • Moderately easy to install
  • Easy to reconfigure
  • Easy to troubleshoot
  • Units affected by media failure: One, unless it is the hub/switch/router, in which case all nodes are affected.
• Ring - like a daisy chain, going from one station to the next and all the way back to the server or the first node in the ring
  
  • Moderately simple to install
  • More difficult to reconfigure as the number of stations increases.
  • Easy to troubleshoot
  • Units affected by media failure: All
• Mesh - redundant connections, to survive in case one cable link is broken; in the image above, the Fully Connected example is a fully redundant mesh
  
  • Difficult to install
  • Difficult to reconfigure
  • Easy to troubleshoot
  • Units affected by media failure: Few or none
• Tree - more easily understandable if you think about a collection of related networks, separated by geography, which pass signals from one region to another, then pass them for delivery within a region. How about islands, where you can't run cables from one to another whenever you want?
  
  
  
  In the image above, imagine a message originating on Niihau, passing to a network on Kauai, then to a network on Oahu, then Maui, then Hawaii, and to an observatory at the top of Mauna Kea. Each of the named nodes might have branches, but you would have to pass traffic along the main trunk to get to them, if the network was structured that way. (Don't know which island is which? Hover over the image. Or click here for more information.)
  
  
• Hybrid - combination of more than one type, such as a series of bus networks, connected together in a ring, or a combination of networks running different Network Operating Systems.
  
  • Can be Easy or Difficult to install
  • Easy to reconfigure
  • Easy to troubleshoot
  • Units affected by media failure: All, if the central hub fails. One, if a workstation fails.

The author has mentioned hubs several times, and uses the term generically most places in the text. Think of a hub as being like a power strip, but instead of devices pulling power from it they are sending network signals into it, so the other devices plugged into it can see those signals.

On page 68, the author starts describing what actually goes in an Ethernet frame. He tells us it has seven parts. We talked about four of them in the last lesson when he described a generic frame. Let's look at the seven pieces of an Ethernet frame:

• preamble - a 64 bit series of 1s and 0s, that alternate, and end in 11; this is like an alarm to a receiving device, like ringing a phone or a doorbell
• destination MAC address - where the frame is going; if if is not meant for me, I ignore it, unless I am a NIC running in promiscuous mode
• source MAC address - where the frame is from
• data type - very basic information; may be a label that says IPv4 or IPv6
• data - in addition to the data, you may find sequence numbers and other addresses in this part
• pad - some frames will not be full, so padding may be needed if the total payload is less than 64 bytes (minimum size for an Ethernet frame)
• Frame Check Sequence (FCS) - a calculated cyclic redundancy check number, which is matched against the same calculation on the receiving end to check for errors in the frame

On page 71, the author gets around to discussing a topic that belongs to the MAC sublayer of the Data-Link layer of the OSI model: media access. He has mentions a couple of times that if you were using a star-wired network, concentrating the connections with hubs, only one device at a time can transmit. This is because any signal going into a hub comes back out on all the ports, except the one it came in on. Sort of like turning on a water spigot that connects to three different sprinklers: they all spray water. That being the case in early Ethernet implementations, there had to be a way to make the devices wait until no one else was transmitting.

There are three classic methods. The author only discusses one, which is the only one used in most networks. To be complete, these are the methods:

Media Access - 3 methods:

• Contention - devices transmit when they need to, if the line is clear
  
• Token Passing - devices take turns transmitting
  
• Polling - devices are asked if they need to transmit

Many operational standards have been created by the Institute of Electrical and Electronic Engineers (IEEE) or the American National Standards Institute (ANSI). Standards are often referred to by a number. Some of the IEEE 802.x standards (there are about a dozen and a half, currently) are LAN protocols. 802.3 specifies the CSMA/CD access method, so this is often thought to be the Ethernet standard. 802.3 was based on Ethernet, but it is a more general standard. Ethernet can be thought of as one implementation of the 802.3 standard.

So , what's the deal about contention? Contention systems work by letting each device try to send a message on the net as needed, contending or competing with all the other devices for the bandwidth. Two examples of methods that support such systems are CSMA/CD (Carrier Sense, Multiple Access, with Collision Detection) and CSMA/CA (Carrier Sense, Multiple Access, with Collision Avoidance). A collision occurs when two signals collide (are sent at the same time) on the medium, causing signal loss. These protocols best support intermittent transmissions. Time sensitivity is good, as users do not often have to wait for media access.

In a CSMA/CD system (example: Ethernet), the collision is detected and the devices that caused it each wait a random number of nanoseconds before sending again. This usually results in one device going ahead of the other. In a CSMA/CA system, devices can be assigned time slices or can be required to ask permission to send, avoiding collisions. Apple LocalTalk is an example of this.

The author discusses collision domains on page 72. A collision domain is a circuit/channel on which a collision can occur. If you only had one device on a given circuit, you would think there would be no danger of a collision, but there still is because devices send and receive. This is usually not a problem, because transmissions are typically short. It becomes a problem when you have too many devices on a single collision domain, so the goal is often increase the number of collision domains, reducing the number of devices on each one, which improves throughput. In a star-wired network, each device wired to a switch is usually considered to be in its own collision domain.

On page 73, the text discusses 10BaseT networks, and it puzzles me that the author persists in not defining the term. He finally tells you that 10BaseT means 10 Mbps, baseband transmission, and UTP cable. Baseband means the cable carries one signal at a time, which is the cause of the collision domain problem.

Although many 10BaseT LANs were put together with Cat 3 (category 3) UTP cable, no one would do so now with cable less than Cat 5. A 10BaseT network uses UTP cable, and RJ-45 connectors. Only four of the eight wires in the cable are actually used by the network. In the wiring table below, it would be the orange and green circuits that are actually used.

In the chart below, the arrangement of wires for the standard known as TIA/EIA 568B is shown. In an alternate standard, the TIA/EIA 568A standard, orange/white is swapped with green/white, and orange is swapped with green. It does not really matter which standard you use, as long as both ends of the cable are connected in the same way. There are two exceptions to this: a crossover cable and a rollover cable. (See below.) A crossover cable is used to connect directly from one NIC to another, or from one networking device to another. A rollover cable is used to connect to a Cisco router's console port.

Most references forget to tell you the reason you do it this way instead of however you might like. Read the last two columns in the chart below. Pins 1 and 2 are used for the transmission circuit, which is why they need to be wired with two wires that are twisted around each other in the cable. Using a twisted pair of wires in a circuit reduces the amount of signal lost to other circuits (crosstalk). You need to use a real pair for each circuit that your network requires. Pins 3 and 6 are used for the reception circuit. The odd part is the 3-6 pairing, surrounding the 4-5 pairing. We wire a connector this way so that it follows a pattern of alternating stripes and solids, so a person can remember it, and because that's the way it works. Why did they decide to use the connectors and sockets this way? I don't know. Just know that this is how it works.

These two standards are illustrated in the text on page 76. These illustrations are kind of backwards. This is how you might see the wires if you were looking at the connector from the side with the clip on it, which no one would do, because it is much easier to see them from the other side.When you are inserting the wires into a connector, do it with the gold contacts up so you can see the wires enter each channel.

UTP cables are usually connected to devices with RJ-45 connectors. In the enlarged picture on the right, note the eight gold-colored connections for the eight wires usually found in UTP cables. Note also the clamp in the connector that grabs the cable where it is covered by its outer insulator.

If you insert the eight wires into the connector and the outer insulator does not extend past the clamp, pull the wires back out, trim them as needed and try inserting again. (I like the scissors on a Swiss army knife.) If everything is in the right spot, then you can carefully put the connector into a crimper and squeeze hard.

The insulation shown in the graphics above should NOT be stripped back on these wires.

Straight-through (standard) cable If you are making a straight-through cable (to run from a workstation to a hub or switch) connect both ends as listed above and shown on the right. Insert the wires into the RJ-45 connector, then crimp with the crimping tool. (There will be no spaces between the wires when they are inserted into the RJ-45 connector. Space is used here to make the color pattern more readable.)	End 1 568B
	End 2 568B
Crossover cable If making a crossover cable (to run directly from one NIC to another) swap the orange and green circuits on one end only: put orange/white on 3, orange on 6, green/white on 1, and green on 2. Insert the wires as shown on the right, then crimp. (This second configuration is actually EIA/TIA 568A.)	End 1 568B
	End 2 568A
Rollover cable Now, for something completely different, if you are making a rollover cable (to run from a workstation to an older Cisco router), prepare the cable like a standard cable, both ends in the same configuration. Before crimping the second end, roll the cable (or the RJ-45 connector) over, 180 degrees. That will make pin 1 on one end of the cable connect to pin 8 on the other end, pin 2 to pin 7, pin 3 to pin 6, and pin 4 to pin 5. If you don't want to think about rolling anything over, insert the wires as shown on the right, then crimp. This cable is used with an adapter to connect to a Cisco router's console port. NOTE: A rollover cable can also have an RJ-45 on one end and an RS-232 DB-9 connector on the other end. This is useful for connecting to a laptop/server/PC that has an open serial port but no open NIC port.	End 1 568B
	End 2 568B, RJ-45 turned over

The text describes two kinds of Ethernet, both of them old standards:

• 10BaseT
  
  • 10 Mbps speed
  • Baseband transmissions
  • maximum segment length is 100 meters (true for all UTP cabling)
  • No more than 1024 nodes per hub
  • Star-wired bus: physical star, logical bus
  • uses two pairs of Unshielded Twisted-Pair (UTP) cabling.
  • uses standard RJ-45 connectors
• 10BaseFL
  • 10 Mbps speed
  • Baseband transmissions
  • maximum segment length is 2000 meters
  • No more than 1024 nodes per hub
  • Star-wired bus: physical star, logical bus
  • uses multi-mode fiber optic media
  • uses ST (bayonet mount) or SC (snap) connectors.

The text continues with a section about networking equipment.

• Hub and switches - An advantage to UTP cable is that networks using it are usually wired as stars, which means that wire run from nodes to hubs or switches.
  The text continues to talk about hubs which is strange,since no one uses hubs any more. Why?
  A hub is a device that has several RJ-45 ports. You can plug in as many devices as you have ports, then every signal that is transmitted by any device that is plugged in to that hub will be passed on to the rest of the devices plugged in to that hub. Some hubs can retransmit (amplify) the signals, but none of them decide where to send a signal. Any incoming signal goes back out all ports except for the one the hub received the signal on. This means only one of those devices can transmit at any given time.
  A switch learns which devices are reachable on which ports by noticing the sender's MAC address on each incoming packet and making a list, called a Source Address Table (SAT). If a switch knows that a message is meant for a device connected to port 7, that's the only port that signal will be sent through. All other ports are available for other traffic. This is much more efficient. It should be clear that switches increase the bandwidth inside a network by connecting only the devices that need to be connected at any given moment.
  
  Note the illustration at the top of page 79. It shows one possible arrangement of lights on sockets on a hub/switch. Typically switches and NICs have two lights for each socket: one for power, the other for traffic. The one that shows that power is on is called a link light. It does not prove that communication is taking place, but if it is lit that typically means that you have a continuous electrical circuit. The traffic light should blink periodically to show frames being sent or received. Constant blinking would indicate either a lot of traffic or a malfunctioning NIC.
  
• Cable - straight-through cables, crossover cables, and rollover cables are made as shown above. As a matter of disclosure, it may not be necessary to make or have a crossover cable to connect two networking devices. Modern switches and routers often include internal software to recognize the kind of cable being used, and the kind of device it connects to. If you have connected the wrong kind of cable, the device may be able to compensate for that, changing what it does with the signals on the fly.
  
• Routers - pass signals from one network to another. Routers use software addresses instead of hardware addresses
• Uplink ports - Hubs, switchers, and routers all have uplink ports. An uplink port may be called that, called an MDI-X port, or may be called an OUT port. The uplink port is used to connect a hub or switch to a router, or to connect a hub to another hub. You connect from the Uplink/MDI-X/OUT port on one device to a Regular/MDI/In port on the next device.
• Bridges - Bridges connect network segments togetheer and act as filters, to minimize traffic. Without such filters, all traffic on the net would go to all stations on the net, on all segments. Bridges connect two LAN segments, and filter traffic so that not all signals have to appear on both segments. Since bridges use hardware addresses to make their decisions, bridges are considered Data-Link layer devices.

Now, a problem: a bridging loop. First, you need to know that redundant bridges can be put between segments, in case one breaks. A bridging loop can be created when packets pass endlessly from one segment to the next across the two bridges. It can also happen if the bridges generate a broadcast storm of new packets. An example: Assume two segments are connected by two bridges. A frame is generated on Segment A from workstation W1, and hits both bridges. Both bridges copy the frame, learn that W1 is on Segment A, and both forward the frame to Segment B. However, each bridge will receive the copy that the other bridge forwards to Segment B. This will cause the bridges to update their databases to show workstation W1 as being on Segment B, and they will forward each these frames back to Segment A. Then the process repeats, again and again and again. This is not good.

To avoid the bridging loop problem, IEEE (Institute of Electrical and Electronics Engineers) standard 802.1d gives us the spanning tree protocol. This says that in each redundant pair of bridges, one is the designated bridge, and the other is the backup bridge. The backup bridge in each pair does not forward any frames unless it determines that the designated bridge is down. Switches also use this protocol to avoid bridging loop errors.

Chapter 5

Chapter 5 continues the discussion of Ethernet networks with several standards that followed 10BaseT. The chapter mentions 100BaseT4, which used four circuits of a Cat 3 cable, as a contender that was not successful. These are the first two standards that it discusses more fully:

• 100BaseT (aka 100BaseTX)
  
  • 100 Mbps speed
  • Baseband transmissions
  • maximum segment length is 100 meters (true for all UTP cabling)
  • No more than 1024 nodes per hub
  • Star-wired bus: physical star, logical bus (or logical star, when using switches)
    
  • uses two pairs of Unshielded Twisted-Pair (UTP) Cat 5e (or better), or STP cabling.
  • uses standard RJ-45 connectors
• 100BaseFX
  • 100 Mbps speed
  • Baseband transmissions
  • maximum segment length is 2000 meters
  • No more than 1024 nodes per hub
  • Star-wired bus: physical star, logical bus
  • uses multi-mode fiber optic media
  • uses ST or SC connectors.

The text discusses some of the process to change from one network type to another. Changing your cable may not be necessary, if your existing cable is compatible. Changing all the NICs you own may be necessary, as well as changing all your switches, routers, and hubs that are not compatible with the new standard. Any time you send signals through a circuit that includes a slower device your throughput will downgrade to that lower bandwidth. This principle applies to any major change of network equipment.

In the discussion of 100BaseFX, the text brings up reasons to use fiber optic cable instead of UTP. It does not mention the most obvious reason: greater bandwidth. In this example, the bandwidth is the same as 100BaseT. There are, however, three differences mentioned. Fiber optic has a longer segment length (2000 meters in this case, fiber optic is immune to electromagnetic interference (UTP is not), and fiber optic is much more resistant to eavesdropping.

• Simplex - this is communication in one direction, more like a monolog than a dialog. It is like a public speech or a television transmission.
• Half-Duplex - this is a dialog that can flow both directions, but only one direction at a time. After one side transmits, the channel has to be "reversed" for the other side to transmit. It is like CB or Ham radio, using only one frequency at a time and taking turns.
• Full-Duplex - this is a dialog in which both sides can transmit and receive at the same time. It is like a telephone conversation, in which both sides have a live speaker and microphone.

Usually, NICs negotiate the level of service for this when a session is established. You can hard code a NIC to use only one type of conversation, but this is usually a way to make fail rather than a way to make it work better.

The next topic covers several types of Gigabit Ethernet, which is more fun to say than 1000-Base-Whatever. The most useful part of the discussion is the chart on page 96, which is more useful when paired with the comments on page 95.

• 1000BaseCX - uses an odd cable and connector (twinaxial) and has a short segment length (25 meters); twinax is a special case of STP (see the image on page 95)
  
• 1000BaseSX - uses multimode fiber optic cable, often uses LC connectors, segment length varies from 220 to 500 meters: popular
  
• 1000BaseLX - uses single mode fiber optic cable, may us LC or SC (snap) connectors, segment length 5 km: more likely to be used as a backbone
  
• 1000BaseT - not discussed in the chapter: Cat 5/6, uses four circuits. RJ-45 connectors, 100 meter segment length like all UTP
  

The text discusses the various connectors used with fiber optic cable. This site presents pictures and descriptions of many more types than the text.

• It tells us that the original ST connectors had problems with cable conductors breaking: ST bayonet connections required the technician to twist the cable, which is bad for glass or plastic fibers.
• SC connectors snap together, but they are somewhat large, and not popular since smaller connectors became available.
• LC connectors are about half the size of SC connectors, and seem to snap together more easily.
  

As you can see on the Fiber Optic Association guide (link above), there are adapters that can allow you to connect from one of these fiber technologies to another.

When connecting equipment with different kinds of jacks, it is important to keep track of which end of a cable goes where. For example, if you have an ST-LC cable, the LC end may be color coded yellow on one strand and white on the other. The yellow strand is referred to as A, the white strand as B. On the other end of that same cable, the colors may be red and black. Which is which? Red is the A strand, so it is the other end of the yellow strand. Black is the B strand, so it is the other end of the white strand. In the chart below, the Column A represents two ends of the same cable. Column B represents two ends of the other cable.

ST to LC cable

	A	B
ST	Red	Black
LC	Yellow	White

The text on that web site also lists several adapters that will allow you to connect fiber to other cable types: UTP, STP, and coaxial cable.

The next performance level described is 10 Gigabit Ethernet (10 GbE), which is obviously 10 gigabits per second throughput. The text tells us this is a newer technology. The wikipedia article behind the link in the last sentence shows that several years later, there are still many versions striving to be standards. The text tells us that there are two copper wire standards, the web article lists four. This area pretty volatile to use as exam material. Note the standards listed on page 98. Several end in R, while others end in W. The W is a WAN (Wide Area Network) indicator, the R is an indicator of a LAN (Local Area Network) standard. Standards that end in W are also for connecting to SONET equipment, which is an ANSI standard for optical systems.