A look at the exam's format and content
Exam 70-059: Internetworking with Microsoft TCP/IP on Microsoft Windows NT 4.0 measures your ability to implement and administer TCP/IP on Windows NT Server 4.0. The TCP/IP exam is important for three certifications. First, you can become a Microsoft Certified Professional (MCP) with an Internet specialty if you pass the TCP/IP exam and two other exams. Second, the TCP/IP exam counts as an elective toward becoming an MCSE, which requires passing six exams. Third, passing the TCP/IP exam is mandatory for becoming an MCSE with an Internet specialty, which requires passing seven exams. (For the specific exams that you must pass for each of these certifications, go to Microsoft's Web site at http://www.microsoft.com/train_cert/cert/cert.htm.)
A candidate taking the TCP/IP exam has 90 minutes to answer 58 questions. Out of a possible 1000 points, you must get a score of at least 750. To obtain this score, you must answer 44 questions correctly.
The multiple-choice questions in this computer-administered exam are extremely verbose and include a large number of exhibits. Typically, you choose among four responses, but some questions include a possible fifth response. Of the questions with five possible responses, the majority contains a scenario, a solution, several required results, and several optional desired results. You then must choose how well the solution met the desired results. As Figure 1 shows, these questions are often complicated and difficult to answer. My advice is to mark such questions the first time through. Then when you have finished the exam, you can check to make certain you read and answered them correctly.
The answer to the sample question in Figure 1 is E (the required result was not met). To route TCP/IP packets on an NT server functioning as a multihomed host (i.e., a host that has more than one networking adapter installed--in this case, three network cards), you must assign an IP address to each card. As for the optional results, the Simple Network Management Protocol (SNMP) component is correct, but for routing tables to automatically update, you must install Routing Information Protocol (RIP) for IP.
At this point, you might be thinking that RIP is no longer needed because of the new Routing and Remote Access Service (RRAS) in Service Pack 3 (SP3). But Microsoft's exams test you on the core products only; Microsoft doesn't include service packs and add-ons in its scenarios.
The TCP/IP exam contains five sections: Planning, Installation and Configuration, Connectivity, Monitoring and Optimization, and Troubleshooting. If you understand the concepts covered in these sections, you will have little difficulty passing the TCP/IP exam. Here is a summary of the primary concepts in each section. If you need more information, see the sidebar "Suggested Study Aids," page 147, for additional resources.
Planning
To plan a TCP/IP network, you need to know about TCP/IP and Point-to-Point Tunneling Protocol (PPTP). Microsoft has built support and implementation of TCP/IP into NT Server 4.0 and NT Workstation 4.0. To use TCP/IP, every host computer must have a unique IP address and share a subnet mask with other hosts on the same network. With PPTP, you can securely transmit Point-to-Point Protocol (PPP) packets over a TCP/IP network. PPTP is part of an emerging technology called Virtual Private Networks (VPNs). PPTP can use X.25 technology, ISDN, or a Public Switched Telephone Network. NT enhances TCP/IP by including services for Windows Internet Naming Service (WINS), Domain Name System (DNS), and Dynamic Host Configuration Protocol (DHCP).
Installation and Configuration
More than half of the exam questions address TCP/IP installation and configuration. You need to understand a variety of concepts associated with these topics, including subnetting, DHCP, HOSTS and LMHOSTS files, DNS and WINS servers, TCP/IP printing, and SNMP.
Subnetting. Many of the questions in the installation and configuration section deal with subnetting. Although you can use the onscreen scientific calculator during this exam, you can save yourself a lot of time if you memorize binary-to-decimal conversion and subnetting information.
IP addresses are 32-bit binary numbers, most often written as four octets (separated by periods), converted to decimal values. For example, the decimal IP address of 192.14.200.2 in binary form is 11000000.00001110.11001000.00000010. To simplify the conversion, always break the address (whether it is decimal or binary) into the four parts and then convert each part. Memorizing the binary place values (128, 64, 32, 16, 8, 4, 2, and 1) will help you easily convert decimal and binary values.
For example, the decimal value 131 equals 128+2+1, so the binary representation is 10000011. Likewise, 191 is 128+32+16+8+4+2+1 or 10111111. The decimal value 255 is 11111111, and 0 is 00000000. Conversely, the binary value 01010101 is 64+16+4+1 or 85.
The left most part of the IP address identifies the class of network, and the right most part identifies the host. The middle forms the network ID.
There are three classes of networks: A, B, and C. Table 1, page 146, lists these classes and their addresses, the number of available hosts, and default subnet masks. Notice that address 127 is not used. This value is reserved as a loopback address. The number of available hosts remains the same when you leave the subnet mask at the default. However, you can use the subnet mask to decrease the number of hosts and increase the number of subnets. Table 2, page 146, shows the tradeoffs between the number of hosts and the number of subnets.
DHCP. You can manually assign IP addresses to every machine on your network, or you can let a DHCP server do this task for you. DHCP is an extension of the boot protocol for diskless workstations. It issues an IP address from a pool of addresses (i.e., a scope) and assigns the subnet mask, default gateway address, and other parameters. As a result, DHCP reduces configuration problems and simplifies IP administration.
You must have an IP address and subnet mask for any TCP/IP client to properly operate. If your machines are going to interact with computers outside your local network, you must also have a default gateway. This address lets your network route messages outside a subnet. The default gateway must be on the same logical network as the host IP address.
The DHCP server issues, or leases, a DHCP address to a client for a given period, called a lease period. About halfway through that period, the client will try to renegotiate the lease to keep the address longer. If the client doesn't get an acknowledgment, it will try again at regular intervals. If the lease expires, the client will try to get another lease from that or another DHCP server.
A DHCP server cannot be a DHCP client because the DHCP server must have a unique static IP address, subnet mask, and default gateway address. The DHCP server must have at least one scope. Multiple DHCP servers need to have non-overlapping portions of each other's scopes so they can serve the clients of a failed server.
The DHCP server will always assign a DHCP client the same IP address because the server reserves that address (i.e., client reservation). The DHCP server can also reserve addresses within a scope for hosts that use static addresses. For example, the DHCP server can reserve addresses for hosts that don't use DHCP.
HOSTS and LMHOSTS files. You can use an ASCII-readable file, called the HOSTS file, to map host names (common system names) to IP addresses for address resolution. Similarly, you can use the LMHOSTS file to map NetBIOS names to IP addresses. You can even load portions of the LMHOSTS file in the DHCP server's memory so the server doesn't have to read the LMHOSTS file each time name resolution takes place. The LMHOSTS file must at least contain each domain controller on the local domain and the Primary Domain Controller (PDC) of every remote domain. Both HOSTS and LMHOSTS files are static.
DNS and WINS servers. You can use DNS and WINS servers to dynamically answer queries to name resolution requests. DNS servers eliminate the need for HOSTS files because DNS servers map host names given in the format of fully qualified domain names (FQDNs--such as iquest.net) to IP addresses to answer queries. You need one primary DNS server and one secondary DNS server in each zone.
WINS servers eliminate the need for LMHOSTS files because WINS servers map NetBIOS names to IP addresses to answer queries. WINS servers work with Microsoft clients but not with Macintosh clients. (You can, however, use DNS servers to work with Macintosh clients.) Unique entries are those that manually map a computer name to an IP address. Four other static entry types are Domain Name, Group, Internet Group, and Multihomed.
TCP/IP printing. TCP/IP printing can take place between UNIX and NT hosts. The NT server can send print jobs to the UNIX host. To enable this option, you must run a Line Print Daemon (LPD) service on the UNIX host, install a TCP/IP printing service, and give UNIX clients the IP address of the LPD print server. The UNIX host can send print jobs to the NT server. To enable this capability, you must use the Line Print Remote (LPR) command to submit the job.
SNMP. This protocol captures, or traps, configuration and status information on the various machines in a network running an SNMP agent. (An SNMP agent is the SNMP client software.) SNMP then sends this trapped information to a designated machine for network management. For SNMP to work, you must install the SNMP service on the server and specify which machine will receive the trapped messages. This designated machine must be running SNMP Manager.