As your business grows, it may become necessary to divide a network or connect two separate networks. When a network is split or when two networks with different addresses are connected, an internetwork is created. An internetwork has subnetworks, or network segments, that have different network addresses. Even a modest-sized business often has several subnetworks operating, each serving a specific portion of the organization.
The most common reason for segmenting a network is to enhance network performance. If a network has too many users or devices that need access to resources or services, the transmission media can become so busy that devices have to wait for an unacceptable period of time to transmit. When this happens, you begin to notice delays when you try to save or open files or perform other operations.
When you segment a network, you give each subnetwork its own network address. This results in two separate transmission media segments, which can be used simultaneously. Both segments will have only half the users of the original network. Thus, you double network performance. Moreover, on some networks performance can be more than doubled, because on an overloaded network the resources required to manage transmission collisions use more network "bandwidth" (the amount of data that can be transmitted in a fixed amount of time) than those on a modestly busy network. Networks are also segmented to enhance data security and to minimize the effect of equipment failure on any part of the network.
There are several devices and protocols used to internetwork subnetworks. The following sections discuss these items and how they apply to an internetworked environment.
The devices used to interconnect network segments are divided into three classifications: bridges, routers, and gateways. Each of these devices plays a very specific role in internetworking. Bridges and routers are generally used to connect networks that use similar protocols, while gateways are used to connect networks that use dissimilar protocols.
Bridges and routers are usually separate hardware components that are connected directly to the transmission media at the intersection point of the two separate networks. There are also bridges and routers that are software-based and function as part of a server's NOS or run in conjunction with the NOS. Software-based bridges and routers can also be installed on standard computers to create dedicated, standalone devices.
Gateways, on the other hand, are usually a combination of both hardware and software, and they perform much more advanced functions than either bridges or routers. The following sections explain the differences between these internetworking devices.
A bridge operates at the data-link layer (Layer 2) of the OSI model. A bridge acts as an address filter: based on information contained at the MAC level, it relays data between subnetworks.
Simple bridges are used to connect networks that use the same physical-layer protocol and the same MAC and logical link protocols (OSI Layers 1 and 2). Simple bridges are not capable of translating between different protocols.
Other types of bridges, such as translational bridges, can connect networks that use different Layer 1 and MAC-level protocols; they are capable of translating, then relaying frames.
After a physical connection is made (at OSI Layer 1), a bridge receives all frames from each connected subnetwork and checks the network address of each received frame. The network address is contained in the MAC header. When a bridge receives a frame from one subnetwork that is addressed to a workstation on another subnetwork, it passes the frame to the intended subnetwork. Figure 28 provides a simple illustration of how a bridge relays frames between subnetworks.
Figure 28: Internetworking through a bridge
A bridge assumes that all communication protocols used above the data-link layer at which it operates (OSI Layers 3 through 7) are the same on both sides of the communication link. If not, translation between unlike protocols at Layers 3 through 7 will need to be performed by something other than the bridge.
Spanning Trees and Source-Route Bridging
There are two terms connected with bridging that are useful to understand: Spanning Tree Protocol and source-route bridging.
Spanning Tree Protocol prevents problems resulting from the interconnection of multiple networks by means of parallel transmission paths. In various bridging circumstances, it is possible to have multiple transmission routes between workstations on different networks. If multiple transmission routes exist, it is also possible to have an endless duplication and expansion of routing errors that will saturate the network with useless transmissions, quickly disabling it. Spanning Tree Protocol is used to specify one, and only one, transmission route. When bridges use this protocol, they send out special frames to each other so that they can be "aware" of the network's topology; they then disable all redundant pathways.
Source-route bridging is a means of determining the path used to transfer data from one workstation to another. Workstations that use source routing participate in route discovery and specify the route to be used for each transmitted packet. Source-route bridges carry out the routing instructions that are placed into each data packet when the packet is assembled by the sending workstation—hence the name "source routing". Source-route bridging is used on IBM Token-Ring networks.
Although it includes the term "routing", source-routing is a part of bridging technology. Bridging technologies and routing methods can be combined in various ways. For example, there is an IEEE specification for a source-route transparent bridge, a bridging scheme that merges source-route bridging and transparent bridging in one device. When choosing internetworking products, it is important to select those that support multiple bridging methods.
Routers function at the network layer of the OSI model (one layer above bridges). To communicate with each other, routers must use the same network-layer protocol. The sending and receiving workstations on different networks must either share identical protocols at all OSI layers above Layer 3 or something must perform the protocol translation at these layers.
Like some bridges, routers can allow the transfer of data between networks that use different protocols at OSI Layers 1 and 2 (the physical layer and the data-link layer). Routers can receive, reformat, and retransmit data packets assembled by different Layer 1 and Layer 2 protocols. Different routers are built to manage different protocol sets. Figure 29 illustrates how a router transfers data packets.
Figure 29: Internetworking through a router
Wireless routers can be used to join two remote LANs or to connect a LAN with the Internet instead of using expensive WAN technology such as a leased line. Wireless routers can have a transmission range of up to 30 miles and a transmission rate of up to 11 Mbps.
In contrast to bridges and routers, which function at only one layer of the OSI model, a gateway translates protocols at more than one OSI layer. Therefore, a gateway is used to interconnect computer systems that have different architectures and that therefore use different communication protocols at several OSI layers.
A gateway can connect entirely dissimilar networks or it can connect dissimilar systems on the same network (thus, using a gateway does not necessarily involve internetworking). For example, a gateway might translate protocols at several different OSI layers to allow transparent communications between Internetwork Packet Exchange™ (IPX™)-based systems and systems based on TCP/IP, Systems Network Architecture (SNA), or AppleTalk. Figure 30 illustrates how a gateway is used to translate protocols to enable communications between two heterogeneous systems.
Figure 30: Gateways provide protocol translation between dissimilar systems at more than one OSI layer.
A gateway may consist of hardware or software but is usually a combination of the two. It also may provide translation at all or at only some of the different OSI layers, depending on the types of systems it connects.
Differing systems use different protocols for network communication. The following sections discuss several internetworking protocols that we have developed, adopted, or adapted, as well as how they fit into the OSI model.
NetWare Internetworking Protocols
Each of the protocols shown in Figure 31 plays a role, directly or indirectly, in NetWare internetworking. These protocols were developed to facilitate the transfer of data in networked and internetworked environments.
Figure 31: Where NetWare protocols fit in the OSI model
IP and IPX: Network-Layer Protocols
TCP/IP is the Internet's fundamental communications protocol suite. The popularity of the Internet—and of corporate intranets—has therefore made IP the world's dominant networking protocol. As a result, we have also adopted IP as its primary network-layer (OSI Layer 3) protocol. However, we continue to support their proprietary network-layer protocol, IPX.
In a NetWare environment, internetwork packet routing is accomplished at the network layer. In conjunction with industry-standard MAC protocols, NetWare IP and NetWare IPX provide the NetWare addressing mechanism that delivers communication packets to their destination and routes packets between internetworked computers.
IP and IPX base routing decisions on address fields in packet headers (provided by the MAC protocol) and on information received from other internetworking protocols. For example, IP and IPX use information supplied by Routing Information Protocol (RIP) to forward packets to the destination computer or to the next router. Similarly, NetWare Link Services Protocol™ (NLSP™) is a companion protocol to IPX for the exchange of routing information in a network.
Both IP and IPX also employ SAP, a protocol that enables networked devices such as network servers and routers to exchange information about available network services.
RIP and NLSP: Routing Protocols
NetWare routers use distance-vector and link-state routing protocols to exchange routing information with neighboring routers. In an internetwork using distance-vector routing—the traditional method of router communication—routers periodically receive information about the internetwork's topology from neighboring routers, consolidate this information within their own routing tables, and broadcast packets to other neighboring routers that summarize the information they now have.
IP RIP and IPX RIP are well-known distance-vector routing protocols. Examples of other such protocols include Cisco Internet Gateway Routing Protocol (IGRP), which is part of the IP protocol suite, and Routing Table Maintenance Protocol (RTMP), which is part of the AppleTalk protocol suite. Enhanced IGRP can handle AppleTalk and IPX—in addition to IP—routing information.
Link-state protocols adapt more quickly to network topology changes than do distance-vector protocols. Unlike distance-vector routers, each link-state router builds its own routing map: it does not rely upon secondhand summaries from other routers. Moreover, routing transmissions are made only when the internetwork changes, not at predefined intervals, which reduces network traffic. Thus, link-state protocols are better than distance-vector protocols at managing internetworking on large, complex internetworks.
NLSP is a link-state routing protocol. Examples of other link-state protocols include Open Shortest Path First (OSPF), which is part of the TCP/IP suite, and Intermediate System-to-Intermediate System (IS-IS), a router-to-router protocol that is part of the OSI suite.
Various distance-vector and link-state routing protocols can coexist on the same NetWare internetwork. Furthermore, individual routers can be configured to accept or reject individual protocols.
NCP is a set of service protocols that a server's operating system follows to accept and respond to service requests.NCP does not play a direct role in routing. However, it does provide session control and packet-level error checking between NetWare workstations and routers.
TCP and SPX: Transport-Layer Protocols
TCP and Sequenced Packet Exchange™ (SPX™) are transport-layer (OSI Layer 4) protocols. Standards at this OSI layer demand reliability from the end-to-end communication link. Accordingly, TCP and SPX provide guaranteed packet delivery and packet sequencing for IP and IPX, respectively.
Like NCP, TCP and SPX do not play a direct role in routing. These protocols are connected with internetworking only in that they guarantee the delivery of all routed packets.Return to Primer Index | Next Section