Previous | Table of Contents | Next

Page 452

FIG. 19.2
The "simplified" 3-Layer
Reference Model
compared to the OSI
Model.

The IEEE 802 committees set standards for the electronics that push bits over the cabling. It is the Media Access Control (MAC) protocol that defines how bits are pushed onto the network. This software talks directly to the network hardware. For performance and technical reasons, this is a complicated and cryptic interface. The MAC protocol calls for the assignment of unique addresses to individual network adapters so that each computer's network adapter is uniquely addressable, much like a telephone number. This address, called a MAC address, is a 12 digit, hex serial number "burned" into a chip on every network adapter. The first six digits of the MAC address are assigned to each hardware vendor by the IEEE. The last six digits of the MAC address are the serial number set by the vendor.

IEEE standards exist for most of the popular network technologies, including EtherNet, Token Ring, FDDI, 100BaseT, ATM, and Gigabyte Ethernet.

Why should you care about this layer? There are two very good reasons. First, the MAC address (sometimes called the DLC address) is a critical addressing component for almost every protocol stack. Individual workstations can be identified and tracked based on this number. The other reason is that this layer is the most unreliable layer in the network. It is often said that 80 percent of network faults occur in the wiring. These faults are not only due to bad or loose wiring connections. Wiring faults can also be due to wrongly configured electronics.

TIP
With EtherNet, one of the main sources of failure comes from excessive packet collisions. The causes for this can be attributed to an excessive number of repeaters between stations (more than three), EtherNet transceivers connected to repeaters with SQE turned on (SQE should never be turned on in this instance), and failing EtherNet adapters not listening before sending.

The Interface

On top of the foundation is an interface. The main interface for LAN protocols is the IEEE 802.2 Logical Link Control protocol. It creates a simple and standard means, called Service

Page 453

Access Points (SAPs), for accessing the wire. This is not a very important layer for troubleshooting, but you probably should remember that SAPs are the methods used to access the foundation, since this term comes up occasionally.

The Protocol Stack

Above the interface is the protocol stack. The protocol stack encompasses the network and transport layer of the OSI model. Multiple protocol stacks can share the same wire, while the same protocol stack can use different wires. The network layer of the stack handles the routing of messages through the network or through the multiple connected wires. The transport layer of the stack provides error-free transmission of data, data flow control, error recovery, and receipt acknowledgment. TCP/IP has established itself as the standard protocol stack. Other protocol stacks include IPX/SPX, NetBIOS, DECNet, LU6.2, AppleTalk, Named Pipes, Banyan Vines, and so on.

The protocol stack is very complicated. A lot goes on in the stack, and each stack is very different from another. Protocol stacks often have another addressing scheme different from the MAC address. These addressing schemes often break a network up into sub-networks or subnets, also called zones or areas in some stacks. Much like an area code is needed when dialing a long-distance telephone number, special subnet addressing is used to communicate with another subnet. Specialized devices, called routers, are placed at the intersections of adjacent subnets to inspect messages and deliver them to the correct subnet.

The TCP/IP Protocol Stack

TCP/IP is the most popular protocol stack used for SQL*Net and client/server connections. Developed in the 1960s by the U.S. military and a group of west coast American universities, TCP/IP was developed for the original Internet, called the ARPANET. Core to the design of TCP/IP was the ability to "self-heal" in case of the loss of a portion of the network. This was important to a Cold War military concerned about command-and-control during a nuclear attack. It is also a good design for a huge network without any central control—the Internet.

The TCP/IP stack can run over almost every type of foundation, including EtherNet, Token Ring, Frame Relay, and ATM. This flexibility accounts for much of TCP/IP's success. The TCP/IP protocol implements two protocols, IP and ICMP, to handle the routing of messages through the network.

Internet Control Message Protocol (ICMP) ICMP is a very important protocol for troubleshooting, because it provides error and other control information for testing of routes through a network. Some of the most useful ICMP messages include:

Page 454

Another indispensable tool for TCP/IP uses this feature to trace the route a message takes to get to its destination. The TRACEROUTE tool begins by sending a message with a TTL of one. This first router will expire its counter and send back the message. The address of this router is then recorded by TRACEROUTE. A message with a TTL of 2 is now sent to discover the second router, and so on until the destination is found. In Windows 95, this standard Microsoft utility is called TRACERT.EXE.

TIP
A super, free combination PING and TRACEROUTE tool called Sub-O-Tronic is available for Windows 95 from Virgin Interactive at http://subspace.vie.com on their download page. It is a diagnostic tool for use with their SubSpace Internet game to identify network bottlenecks hurting the performance of their client/server game. It works just as well looking for network problems and bottlenecks in your network. A must have!

Another area of concern for most administrators is correct IP addressing and configuration. For the IP protocol to correctly route your messages across the network, several parameters must be set up correctly.

IP Addressing The IP address is expressed in 4-byte, dotted-decimal notation. In decimal notation, a byte can have a value between 0 and 255, so a 4-byte, dotted-decimal notation looks like an address between 0.0.0.0 and 255.255.255.255. As mentioned earlier, the IP address refers to both a specific network or subnet and a specific host. This address is divided into classes.

Previous | Table of Contents | Next