Page 455
Some IP addresses are reserved or have special meanings. Some important ones include:
TIP |
The 10.x.x.x network has been set aside for sites wanting to use an unregistered IP address scheme. The routers within the Internet will ignore messages with a 10.x.x.x network address. It is highly recommended to use this scheme at most sites. |
IP Subnet Mask Another level of complexity added to IP network addressing is subnet masking. Subnet masking is used to expand the number of networks due to the 4-byte limitation of IP's addressing field. Bits can be robbed from the host number to allow additional subnets to be defined. For example, if you choose to use the 10.x.x.x Class A network number for your unregistered network, you can use a subnet mask to create additional networks using this address. The subnet mask specifies which bits of the IP address designate network addresses and which bits designate host numbers. The bit is given the value 1 if it designates network addresses and 0 for hosts. A Class A address, therefore, would have all 1s in the first 8 bits, since the first byte is the network address. The decimal value for a byte of all 1s is 255. Therefore a Class A address would have a standard mask of 255.0.0.0.
Now say you want to use the second byte of your 10.x.x.x address for subnet addresses. The first two bytes are being used for the network address, so the mask would be 255.255.0.0. This will give you 254 subnets with 65,534 hosts in each.
Default Router Another configuration needed to make the IP portion of TCP/IP route correctly over the network is to specify a default route. This is the IP address of a router in your network with connections to other network. If you are sending a message to an address on another network and your node doesn't know the address of the next "hop" on its way to the destination, IP will send your message to the default route host. It is assumed this router will know how to handle your message. If it fails, the message will come back with a "network unreachable" error.
Domain Name Server (DNS) Protocol Another important piece
of the TCP/IP protocol stack is DNS protocol. In very static networks, host names and their corresponding IP addresses
can be entered into a host table and stored on each client computer. As the network grows,
this
Page 456
becomes impossible to maintain. The DNS service, like the Oracle Names service, will deliver the correct IP address to a client from a centralized names database. DNS also has the capability to learn names by querying other DNS servers and caching the results. Implementing DNS is important if connection to the Internet is important to your organization. DNS is the basis of all Internet addressing and address resolution. DNS design and configuration is a complex topic outside the scope of this chapter.
Dynamic Host Configuration Protocol The DHCP protocol provides easy, automatic administration of IP addresses. Any Windows NT server, InterNetWare server, and most UNIX servers can be configured as DHCP servers. The DHCP server is configured with the network address of the network for which it will be granting IP addresses. This configuration is called a DHCP scope. The range of address that will be "leased" is also specified in the scope definition. DHCP can also supply your clients with the correct subnet mask, the default route, and DNS server address while supplying the IP address. DHCP is a very powerful protocol and should be implemented on any network with Windows 95 or Windows NT clients.
SQL*Net provides three general operations that are used with Oracle products and databases that use SQL*Net for distributed processing. These functions provided by SQL*Net operate in the background and are not visible to the user.
Connect Operations A client application initiates a request for a connection to a remote database by providing the database alias name for the desired destination database. This alias name, called a service name, is mapped to a network address contained in the connect descriptor stored in the TNSNAMES.ORA network configuration file or in a database serviced by Oracle Names.
Data Operations The client and server can send and receive data between themselves. Oracle supports both synchronous (single-threaded, ping/pong) requests and asynchronous (multi-threaded, windowed) requests.
Exception Operations SQL*Net supports several exception functions such as breaking, resynchronizing, and testing a SQL*Net connection.
All installation information applies to SQL*Net version 2.3.3. Refer to your operating system- specific Net8 Installation Manual for assistance in installing Net8.
Before installing SQL*Net, some planning needs to be done up front. Decisions to be made include:
Page 457
The most critical network layout decision is whether more than one LAN protocol will need to be used. Using multiple protocols will greatly increase the complexity of the Oracle network and will increase the number of points of failure. If possible, use only one LAN protocol.
The second issue to confront is the centralization of Oracle network management. As with most things, the more fingers in the pie, the more problems that will occur. Some sites are so big that decentralization is necessary. If possible, centralize network management.
Many things in an Oracle network require names. Each protocol type used in the Oracle network is called a community and is given a community name. Regions of a company can be broken up for decentralized management reasons. These regions are called domains. By default, Oracle Network Manager names the core domain .world. Additionally, all network listener nodes, database service names, MultiProtocol Interchanges, and Oracle Name servers need names. Document a scheme and stick with it. Consistency counts! See Chapter 20, "Advanced Oracle Networking," for more information on naming.
As you have probably figured out by now, the Oracle Network Manager is the tool for naming and configuring all these network pieces.
Client and server machines require configuration files for proper operation. Client machines typically have three configuration files that are created by the Oracle Network Manager program and one optional file that must be created manually. These files provide configuration information regarding network destinations, network navigation, and diagnostics such as tracing and logging.
TNSNAMES.ORA This file contains network destination information used by the client. Information contained in this file includes database "service names" and connect descriptors. With Net8, another section was added to this file to handle the failover mode of the Oracle Parallel Server Option where multiple instances can access the same database. Different parameters can be used in this section: BASIC, PRECONNECT, SELECT, SESSION, and NONE. A better understanding of these parameters can be found in Oracle's Parallel Server manual. Without Oracle Names, a client requires this file to expand short service names into fully qualified connections. This file can be generated by Oracle Network Manager or Oracle Net8 Assistant. Although not recommended by Oracle, experienced users who are familiar with the proper syntax can edit this file manually with success. It would be good practice to back up this file if you do intend to do this.
SQLNET.ORA This file contains optional diagnostic parameters and client information about Oracle Names. This file is created by Network Manager, but can also contain node-specific parameters that require manual editing.