Remote Node

1. A remote node is a device that connects to a network from a point some distance away from the central host. For example, a CO (Central Office) in the PSTN (Public Switched Telephone Network) might support a number of remote nodes. Some of the nodes are dumb line concentrators that serve only to concentrate traffic over high-capacity trunks in order to reduce cabling costs. Other nodes are intelligent switching partitions that can switch basic local traffic within their own geographic domains, even though they rely on the CO for guidance in the delivery of more complex services, such as custom calling features.

2. Remote node software allows remote users to dial-in to the corporate LAN and work with the applications and data on the LAN as if they were "actually in the office." By dialing in, they become nodes on the LAN. Using a PC, Mac or UNIX workstation; a modem; and a remote access server, employees can connect from any location in the world that has an analog, a switched digital, or a wireless connection.

Switching Fabric

The term "switching fabric" refers to the component at the heart of a data communications switch that allows any input port to send data to any output port. Many different kinds of switching fabric have been used over the years, depending on the manufacturer, the size and type of the data communications switch, and the technology available at the time. Sometimes a switching fabric will directly connect to all ports, but usually there are a group of ports on a single card called a line card and the switching fabric connects the line cards together. There are many different types of switching fabrics available on the market today. An example of one of the most basic is the "crossbar" switching fabric, which consists of a matrix of rows and columns, where each row is connected to an input port and each column is connected to an output port. The resulting diagram looks like a fabric with threads crossing at right angles. A switch or "crosspoint" is located at each intersection between a row and a column. By closing the right crosspoints, each input port can be connected to each output port. Crossbar fabrics are very general, but expensive to create in large sizes because the number of crosspoints is equal to the number of input ports times the number of output ports. For instance, if you had a small eight port switch you would have eight potential input and output ports making a total of 64 crosspoints, but if you had a large switch with 100 ports you would need 10,000 crosspoints to allow every port to connect with each other. Other types of switching fabric use buffering, queuing, packet shaping, switching logic, and specialized application specific integrated circuits (ASIC) to enhance switching fabric performance. A well-designed switching fabric will reach switching speeds equal to the line rate of the port. For instance, a port with a theoretical speed of 100 Mbsp should be able to pass packets across the switching fabric to the destination port or ports at 100 Mbps, which is also known as line-rate or wire-speed switching. A poorly designed switching fabric has delays or other latency that will drop the data rate as packets travel through the switching fabric. The variety and performance of switching fabrics depend on many different variables as well as the manufacturer. However, one thing is for certain future trends in switching fabrics are hard to anticipate, but the switching fabric will always remain at the heart of the data communication switch.

Below is a simplified illustration of switching fabric which shows how any input port (i.e. line card) can transmit data to any output port, essentially linking all of the line cards together. Notice a "crosspoint" is depicted with a solid black dot indicating the intersection between a row and a column of the fabric.


Source: http://choonho.files.wordpress.com/2007/09/capture4.jpg

Internetworking

Communication between two networks or two types of networks or end equipment. This may or may not involve a difference in signaling or protocol elements supported. In the narrower sense - to join local area networks together. This way users can get access to other files, databases and applications. Bridges ad routers are the devices which typically accomplish the task of joining LANs. Internetworking may be done with cables - joining LANs together in the same building, for example. Or it may be done with telecommunications circuits - joining LANs together across the globe.

Two architectural models are commonly used to describe the protocols and methods used in internetworking. The Open System Interconnection (OSI) reference model provides a rigorous description for layering protocol functions from the underlying hardware to the software interface concepts in user applications. Internetworking is implemented in Layer 3 (Network Layer) of the model.

The Internet Protocol Suite, also called the TCP/IP model, of the Internet was not designed to conform to this model. Despite similar appearance as a layered model, it uses a much less rigorous, loosely defined architecture that concerns itself only with the higher level aspects of networking, i.e. it does not discuss hardware-specific low-level interfaces, other than assuming availability of a link-layer interface to the local network link. Internetworking is facilitated by the protocols of its Internet Layer.

Hub

The point on a network where circuits are connected. In local area networks, a hub is the core of a physical star configuration, as in ARCNET, StarLAN, Ethernet, and Token Ring. Hub hardware can be either active or passive. Wiring hubs are useful for their centralized management capabilities and for their ability to isolate nodes from disruption. Hubs work at Layer 1 (Physical) and 2 (Data Link) of the OSI Reference Model, with emphasis on Layer 1. Hubs aren't switches, as they have very little intelligence, if any, and don't set up transmission paths. Rather, hubs comprise a physical bus and numerous ports, to which are connected a bunch of wires, to which are connected individual terminal devices. As hubs are protocol-specific (e.g, Ethernet) and are not intelligent, they are very fast and very cheap. A 10Base-T hub is an inexpensive means of allowing LAN-attached devices to share a common, collapsed bus contained within a hub chassis. The connections are via UTP (Unshielded Twisted Pair), which is much less expensive than are the classic connections through coaxial cable. Unlike switches, hubs do nothing internally to control congestion. However, they typically are workgroup-level solutions which allow a large, logical Ethernet to be subdivided into multiple physical segments. For example, you could even use a small five-port hub on your desk to connect a couple of laptops and a desktop PC. Hubs can be interconnected directly, or through switches or routers, with the traffic being forwarded from the originating hub only if the destination address of the data packet indicates that is necessary to do so. Therefore, hubs do reduce congestion through the control of interhub traffic.

Bridge

1. In classic terms, a bridge is a data communications device that connects two or more network segments and forwards packets between them. Such bridges operate at Layer 1 (Physical Layer) of the OSI Reference Model. At this level, a bridge simply serves as a physical connector between segments, also amplifying the carrier signal in order to compensate for the loss of signal strength incurred as the signal is split across the bridged segments. In other words, the bridge is used to connect multiple segments of a single logical circuit. Classic bridges are relatively dumb devices, which are fast and inexpensive; they simply accept data packets, perhaps buffering them during periods of network congestion and forward them. Bridges are protocol-specific, e.g., Ethernet or Token Ring in the LAN domain. Bridges also are used in the creation of multipoint circuits in the WAN domain, e.g., DDS (Dataphone Digital Service).

Bridges also can operate at Layer 2 (Link Layer) of the OSI Reference Model. At this level, a bridge connects disparate LANs (e.g., Ethernet and Token Ring) at the Medium Access Control (MAC) sub-layer of Layer 2. In order to accomplish this feat, the MAC Bridge may be of two types, encapsulating or translating.

Encapsulating bridges accept a data packet from one network and in its native format; they then encapsulate, or envelope, that entire packet in the format acceptable to the target network. For instance, an Ethernet frame is encapsulated in a Token Ring packet in order that the Token Ring network can deliver it to the target device, which must strip away several layers of overhead information in order to get to the data payload, or content. In order to accomplish this process, a table lookup must take place in order to change basic MAC-level addressing information.

Translating bridges go a step further. Rather than simply encapsulating the original data packet, they actually translate the data packet into the native format of the target network and attached device. While this level of translation adds a small amount of delay to the packet traffic and while the cost of such a bridge is slightly greater, the level of processing required at the workstation level is much reduced.

Bridges also can serve to reduce LAN congestion through a process of filtering. A filtering bridge reads the destination address of a data packet and performs a quick table lookup in order to determine whether it should forward that packet through a port to a particular physical LAN segment. A four-port bridge, for instance, would accept a packet from an incoming port and forward it only to the LAN segment on which the target devices is connected; thereby, the traffic on the other two segments is reduced and the level of traffic on those segments is reduced accordingly. Filtering bridges may be either programmed by the LAN administrator or may be self-learning. Self-learning bridges "learn" the addresses of the attached devices on each segment by initiating broadcast query packets, and then remembering the originating addresses of the devices which respond. Self-learning bridges perform this process at regular intervals in order to repeat the "learning" process and, thereby, to adjust to the physical relocation of devices, the replacement of NICs (Network Interface Cards), and other changes in the notoriously dynamic LAN environment.

While bridges are relatively simple devices, in the overall scheme of things, they can get quite complex as we move up the bridge food chain. Bridges also can be classified as Spanning Tree Protocol (STP), Source Routing Protocol (SRP), and Source Routing Transparent (SRT).
Spanning Tree Protocol (STP) bridges, defined in the IEEE 802.1 standard, are self-learning, filtering bridges. Some STP bridges also have built-in security mechanisms which can deny access to certain resources on the basis of user and terminal ID. STP bridges can automatically reconfigure themselves for alternate paths should a network system fail.

SRP bridges are programmed with specific routes for each data packet. Routing considerations include physical node location and the number of hops (intermediate bridges) involved. This IBM bridge protocol provides for a maximum of 13 hops.

SRT bridges, defined in IEEE 802.1, are a combination of STP and SRP. SRT bridges can act in either mode, as programmed.



2. Bridge is also a verb, as in "to bridge." Imagine a phone line. It winds from your CO (Central Office) through the streets and over the poles to your phone. Now imagine you want to connect another phone to that line. A phone works on two wires, tip and ring (positive and negative). You simply clamp each one of the phone's wires to the cable coming in. That's called bridging. Imagine bridging as connecting a phone at a right angle. When you do that, you've made what's known as a "bridged tap." The first thing to know about bridging is that bridging causes the electrical current coming down the line to lose power. How much? That typically depends on the distance from the bridged tap to the phone. A few feet, and there's no significant loss. But that bridged tap can also be thousands of feet. For example, the phone company could have a bridged tap on your local loop, which joined to another long-defunct subscriber. The phone company technicians simply saved a little time by not disconnecting that tap. If you want the cleanest, loudest phone line, the local loop to your phone should not be bridged. Instead it should be a direct "home run" from your CO to your phone.

Bridging can be a real problem with digital circuits. Circuits above 1 Mbps (e.g, T-1) should never, ever be bridged. Because of the power loss, they simply won't work or will work so poorly they won't be worth having. ISDN BRI (Integrated Services Digital Network Basic Rate Interface) channels are also digital. But they were specifically designed to work with the existing telephone cable plant, which has a huge number of bridged circuits. Telephone companies typically will install ISDN BRI circuits with up to six bridged taps and about 6,000 feet of bridged cabling.

Router

1. As in software, router is a system level function that directs a call to an application.

2. As in hardware, routers are the central switching offices of the Internet and corporate Intranets and WANs. Routers are bought by everybody - from backbone service providers to local ISPs, from corporations to universities. The main provider of routers in the world is Cisco. It has built its gigantic business on selling routers - from small ones, connecting a simple corporate LAN to the Internet, to corporate enterprise wide networks, to huge ones connecting the largest of the largest backbone service providers. A router is, in the strictest terms, an interface between two networks.

Routers are highly intelligent devices which connect like and unlike LANs (Local Area Networks). They connect to MANs (Metropolitan Area Networks) and WANs (Wide Area Networks), such as X.25, Frame Relay and ATM. Routers are protocol-sensitive, typically supporting multiple protocols. Routers most commonly operate at the bottom 3 layers of the OSI model, using the Physical, Link and Network Layers to provide addressing and switching. Routers also may operate at Layer 4, the Transport Layer, in order to ensure end-to-end reliability of data transfer.

Routers are much more capable devices than are bridges, which operate primarily at Layer 1, and switches, which operate primarily at Layer 2. Routers send their traffic based on a high level of intelligence inside themselves. This intelligence allows them to consider the network as a whole. How they route (also called routing considerations) might include destination address, packet priority level, least-cost route, minimum route delay, minimum route distance, route congestion level, and community of interest. Routers are unique in their ability to consider an enterprise network as comprising multiple physical and logical subnets (subnetworks). Thereby, they are quite capable of confining data traffic within a subnet, on the basis of privilege as defined in a policy-based routing table. In a traditional router topology, each router port defines a physical subnet, and each subnet is a broadcast domain. Within that domain, all connected devices share broadcast traffic; devices outside of that domain can neither see that traffic, nor can they respond to it. Contemporary routers have the ability to define subnets on a logical basis, based on logical address (e.g., MAC or IP address) information contained within the packet header, and acted upon through consultation with a programmed routing table. In addition to standalone routers developed specifically for that purpose, server-based routers can be implemented. Such routers are in the form of high-performance PCs with routing software. As software will perform less effectively and efficiently than firmware, such devices generally are considered to be less than desirable for large enterprise-wide application, although they do serve well in support of smaller remote offices and less-intensive applications. Routers also are self-learning, as they can communicate their existence and can learn of the existence of new routers, nodes and LAN segments. Routers constantly monitor the condition of the network, as a whole, in order to dynamically adapt to changes in network conditions.

Characteristics of routers can include: LAN extension, store and forward, support for multiple media, support for multiple LAN segments, support for disparate LAN protocols, filtering, encapsulation, accommodation of various and large packet sizes, high-speed internal buses (1+Gbps), self-learning, routing based on multiple factors, route length, number of hops, route congestion, traffic type, support for a community of interest (VLAN), redundancy, and network management via SNMP (Simple Network Management Protocol).

Router protocols include both bridging and routing protocols, as they perform both functions. These protocols fall into 3 categories:

1. Gateway protocols establish router-to-router connections between like routers. The gateway protocol passes routing information and keep alive packets during periods of idleness.

2. Serial Line Protocols provide for communications over serial or dial-up links connecting unlike routers. Examples include HDLC (High-level Data Link Control), SLIP (Serial Line Interface Protocol) and PPP (Point-to-Point Protocol).

3. Protocol Stack Routing and Bridging Protocols advise the router as to which packets should be routed and which should be bridged.

Switch

A mechanical, electrical or electronic device which opens or closes circuits, completes or breaks an electrical path, or select paths or circuits. Switches work at Layers 1 (Physical) and 2 (Data Link) of the OSI Reference Model, with emphasis on Layer 2. A switch looks at incoming data (voice data, or data data) to determine the destination address. Based on that address, a transmission path is set up through the switching matrix between the incoming and outgoing physical communications ports and links. Data switches (e.g., LAN (Local Area Network) switches and packet switches) also typically contain buffers, which can hold data packets in temporary memory until the necessary resources are available to allow the data packets to be forwarded. Voice switches, of course, don't, because you can't delay voice. Switches work link-by-link, with multiple switches typically being involved in complex networks; each switch forwards the data on a link-by-link (hop-by-hop) basis. Routers are highly intelligent data switches which are capable of setting up paths from end-to-end, perhaps in consideration of the level of privilege of the user and application. Routers commonly are used at the edges of complex data networks, where intelligence is required to set up appropriate network paths. Although such intelligent decisions impose some delay on the packet traffic, they are made only at the ingress and egress edges of the network. The routers often instruct switches in the core of the network, where speed is of the essence-switches aren't as intelligent as routers, but they are faster and less expensive.