Previous | Table of Contents | Next |
Almost every production network has to be split into multiple physical segments because of the limitations imposed by the technical specification of your network implementation. A network may also be split into separate segments to improve performance. Of course, once you physically split a segment, your networked workstations cannot access any resources outside of their local segment, and this is where routers and bridges come into play.
Routers
A router is used to physically combine multiple segments into one logical segment. Routers work at the transport or protocol layer and will retransmit data packets from one segment to another segment based on unique network addresses. Routers utilize an algorithm to build a list of network addresses in their local segments to provide the ability to only route data packets that require their intervention. If you have multiple routers in your network and one fails, the other router can often manage to find an alternate path to make sure the data packets reach their destination.
Of course, this performance does not come without cost as routers are often expensive items. Many routers are based on Intel CPUs, although the fastest use proprietary RISC-based CPUs. A router requires quite a bit of processing power because of the time limit imposed by the network implementation. For instance, if the time-out limit for an acknowledgment has already been reached, it does no good to route a data packet to another computer. By the time the packet reaches the computer, it has already requested that the packet be resent from the host computer. One of the less expensive alternatives to a dedicated router is to use the routing capabilities of Windows NT to route IPX/SPX and TCP/IP data packets. Another item to consider with routers is that, if they cannot identify a particular packet type, they will discard it. And discarding a packet can cause you numerous problems.
Bridges
Not all network protocols, such as NetBEUI, are capable of being routed, so an alternative method of combining segments is required. This is what a bridge is designed to accomplish. A bridge works at the MAC layer and either retransmits all packets regardless of their network address (transparent routing) or based on a specific network address (source routing). Some bridges combine both of these abilities in a single box. This is commonly referred to as a transparent source routing bridge. Bridges are most commonly utilized in Lan Manager-based networks and may still be utilized when the network is upgraded to Windows NT Server.
I had an interesting problem caused by a bridge with Windows NT Server 3.1 and Windows for Workgroups 3.11 clients. This particular implementation utilized a source transparent bridge to combine network segments. When the Windows for Workgroups client attempted to establish a session with the Windows NT Server, the bridge intervened and the Windows for Workgroup client issued both a source routed packet and a transparent packet across the network segment. The Windows NT Server received both packets, issued an acknowledgment, and began its communication with the Windows for Workgroup client based on source routed packets. But once the NT Server realized that it did not need to include the source routing information, it switched to transparent routing. And this caused the Windows for Workgroup computer to terminate the session it had established with the NT Server. Essentially, this meant that no resources on the NT server could be accessed by the Windows NT client.This particular problem was fixed in a service pack, although it was not specifically mentioned in the bug fix list. Im only mentioning it here so that you are aware of the problem, in case it creeps in again. And also for you to be aware that if you use a bridge, you can eliminate this type of problem by using either source routing or transparent routing, but not both simulaneously.
Switching Hubs
A switching hub is usually used to improve network performance by providing each connected network client with its own network segment. Most times, a switching hub is used between a server and several clients in full-duplex operation. If you have 10Mb network adapters on your network clients capable of full-duplex operation, for example, these clients can communicate at 10Mb upstream and 10Mb downstream, for a total bandwidth of 20Mb. Switching hubs are also used to join two network segments of differing transmission speeds. The server may have a 100Mb network adapter while the clients have 10Mb adapters. By connecting your server to the 100Mb port of your switching hub and your clients (or additional hubs) to the 10Mb ports, you can join the dissimilar segments into a single physical segment. This provides high-speed access to your server and increases network throughput to your clients.
Previous | Table of Contents | Next |