Network Interconnection Devices
Many interconnection device are required in a modern network, from the interface that allows a single computer to communicate with other computers via a LAN cable or a telephone line, to the large and complex switching devices that interconnect two or more entire networks. The main categories of interconnection device used in computer networks are listed below.
Network Interface Card (NIC)
Every device on a network that needs to transmit and receive data must have a network interface card (NIC) installed. They are sometimes called network adapters, and are usually installed into one of the computer's expansion slots in the same way as a sound or graphics card. The NIC includes a transceiver, (a transmitter and receiver combined). The transceiver allows a network device to transmit and receive data via the transmission medium. Each NIC has a unique 48-bit Media Access Control (MAC) address burned in to its ROM during manufacture. The first 24 bits make up a block code known as the Organisationally Unique Identifier (OUI) that is issued to manufacturers of NICs, and identify the manufacturer. The issue of OUIs to organisations is administered by the Institute of Electrical and Electronics Engineers (IEEE). The last 24 bits constitute a sequential number issued by the manufacturer. The MAC address is sometimes called a hardware address or physical address, and uniquely identifies the network adapter. It is used by many data link layer communications protocols, including Ethernet, the 802.11 wireless protocol and Bluetooth. The use of a 48-bit adress allows for 248 (281,474,976,710,656) unique addresses. A MAC address is usually shown in hexadecimal format, with each octet separated by a dash or colon, for example: 00-90-47-03-B6-C4.
An Ethernet network interface card
As signals travel along a transmission medium there will be a loss of signal
). A repeater is a non-intelligent network device that receives a signal on one
of its ports, regenerates the signal, and then retransmits the signal on all of
its remaining ports. Repeaters can extend the length of a network (but not the
capacity) by connecting two network segments. Repeaters cannot be used to
extend a network beyond the limitations of its underlying architecture, or to
connect network segments that use different network access methods. They can,
however, connect different media types, and may be able to link bridge segments
with different data rates.
A repeater connecting network segments
Hubs are used in Ethernet networks. A signal received at any port on the hub is
retransmitted on all other ports. Network segments that employ hubs are often
described as having a star topology, in which the hub forms the wiring centre
of the star.
A hub in a star network configuration
Using a hub provides a degree of fault tolerance, because each network device
has its own connection to the hub, and if a connection fails, only a single
device is affected. Expanding the network is also easier, because many
additional devices can be added to the network using a single hub, which is
itself often connected to a network backbone. Hubs can be either active or
passive. An active hub has its own power supply, and regenerates incoming
frames before retransmitting them. Because signals are regenerated, each output
port can connect a channel of up to 100 metres (the maximum allowed for twisted
pair cables). Passive hubs simply relay the signal without regenerating it.
Managed hubs allow administrators to enable or disable individual ports
remotely, while intelligent hubs can autonomously close ports down if the
occurrence of errors in transmitted packets exceeds a certain threshold.
A 24-port hub
Bridges can be used to connect two or more LAN segments of the same type (e.g.
Ethernet to Ethernet, or Token-Ring to Token-Ring). Like repeaters, bridges can
extend the length of a network, but unlike repeaters they can also extend the
of a network, since each port on a bridge has its own MAC address. When
bridges are powered on in an Ethernet network, they start to
the network's topology by analysing the source addresses of incoming frames
from all attached network segments (a process called
). Over a period of time, they build up a
. Unless the source and the destination are on different network segments,
there is no need for the bridge to transfer an incoming frame to another
network segment. If the source and the destination are on different segments,
the bridge needs to be able to determine which segment the destination device
Bridges learn about the network segments they are connected to
The bridge monitors all traffic on the segments it connects, and checks the source and destination address of each frame against its routing table. When the bridge first becomes operational, the routing table is blank, but as data is transmitted back and forth, the bridge adds the source MAC address of any incoming frame to the routing table and associates the address with the port on which the frame arrives. In this way, the bridge quickly builds up a complete picture of the network topology. If the bridge does not know the destination segment for an incoming frame, it will forward the frame to all attached segments except the segment on which the frame was transmitted. Bridges reduce the amount of traffic on individual segments by acting as a filter, isolating intra-segment traffic. This can greatly improve response times.
Because Ethernet bridges determine whether or not to forward frames on the
basis of the desination MAC address, they are said to operate at the
data link layer
of the OSI Reference Model. Etherenet bridges are sometimes referred to as
, because their presence and operation are transparent to network users,
although they successfully isolate intrasegment traffic, reducing network
traffic overall and improving network response times.
A potential problem arises where additional bridges are added to the network to
provide a degree of redundancy and fault tolerance by providing alternative
paths through the network. If more than one path exists simultaneously between
any two segments,
can occur. The diagram below illustrates a very simple example, in which
bridges can easily become confused. A broadcast message from A, for example,
would be endlessly forwarded by these bridges, using up valuable network
bandwidth and blocking the transmission of other frames on both segments.
A bridging loop
Spanning Tree Algorithm
(STA) can eliminate loops by placing the redundant links into standby mode, to
be reactivated in the event of primary link failure, providing a new path
through the network. The algorithm must be dynamic in that, when a change in
topology occurs, the bridges must be able to register the fact that a change
has occurred and then derive a new spanning tree. The spanning-tree calculation
occurs when the bridges are first powered up, and whenever a change in network
topology is detected. Communication between the bridges is accomplished using
, which are exchanged at regular intervals (typically a few seconds). If a
bridge fails, the other bridges will detect the absence of its configuration
messages and re-run the algorithm.
The Spanning Tree Algorithm (Ethernet)
The spanning tree algorithm eliminates loops from a network by placing bridge ports that if active would create loops into a standby (blocking) condition. A port on standby can be reactivated if a link fails to provide a new path through the network. The spanning tree algorithm uses graph theory as a basis for constructing a loop-free topology. Graph theory states the following:
For any connected graph consisting of nodes and edges connecting pairs of nodes, a spanning tree of edges maintains the connectivity of the graph but contains no loops.
Each LAN segment corresponds to a
, and each bridge corresponds to an
. This enables a simple algorithm to be used by the bridges on the network to
derive a spanning tree. The algorithm is dynamic in the sense that bridges can
register a change in network topology and derive a new spanning tree by
re-running the STA. The algorithm calls for each bridge, and each port on each
bridge, to be uniquely identified. The MAC address of each port is normally
used to identify the port, with one of the bridge ports being used by the
algorithm to identlify the bridge itself, prefixed by an assigned two-byte
. An arbitrary
can be attributed by a network administrator to each bridge port on the
network, and this cost may subsequently be used by the algorithm to determine
which ports are placed into standby mode and which remain open (on the basis of
using ports that represent the lowest cost). If no cost value is assigned, a
default value is used. The diagram below illustrates a network consisting of
five sements in which redundant bridges have created loops within a network.
A network with redundant bridges
The first activity in spanning-tree computation is the selection of the root bridge , usually the bridge with the lowest identifier. The diagrams on this page use simplified bridge identifiers (B1-B5) for the sake of clarity, and the root bridge will be B1. Next, a root port (R) on each of the other bridges must be determined. This is the port from which the root bridge can be reached with the lowest root path cost (this will depend upon the cost of each bridge port that must be traversed on a particular root path). Finally, designated bridges and their designated ports are determined for each LAN segment. A designated bridge is the bridge on the segment that provides the root path with the lowest cost. The designated port is that which connects the segment to its designated bridge. In cases where two (or more) bridges have the same root path cost, the bridge with the lowest numbered bridge identifier is used.
Using the STA, all but one of the bridges connected to each LAN segment are
eliminated, removing any loops while preserving connectivity. The
following modified diagram shows the result of applying the STA to the network
illustrated previously. The STA has placed the ports connecting bridges B3 and
B5 to segment E in standby mode.
The STA places the ports connecting segment E to bridges B3 and B5 into standby mode
The spanning-tree calculation occurs when the bridge is powered up, or when a
change in network topology is detected. In order for this to happen,
communication must take place between bridges, and this is accomplished using
configuration messages called
bridge protocol data units
(BPDUs). A configuration message identifies the bridge presumed by the sender
to be the root bridge, and contains the bridge and port identifiers of the
sending bridge. It also specifies the
root path cost
from the sending bridge to the root bridge, and the age of the information
contained in the configuration message. Bridges typically exchange
configuration messages every few seconds, and if a bridge fails (causing a
topology change), neighboring bridges will detect the abscence of configuration
messages and initiate a spanning-tree recalculation by sending
topology change messages
The format of the IEEE 802.1d configuration message is illustrated below.
Bridge configuration message format
The fields of the bridge configuration message are described below.
- Protocol identifier - contains the value zero.
- Version - contains the value zero.
- Message type - contains the value zero
- Flags - only the first two bits are used. The topology-change bit , if set, signals a topology change, and the topology-change acknowledgment bit , if set, acknowledges receipt of a configuration message with the topology-change bit set.
- Root ID - identifies the root bridge using its 2-byte priority followed by its 6-byte ID.
- Root path cost - the root path cost from the bridge sending the configuration message to the root bridge.
- Bridge ID - identifies the bridge sending the message using its 2-byte priority followed by its 6-byte ID.
- Port ID - identifies the port from which the configuration message was sent.
- Message age - indicates when the configuration message should be deleted.
- Maximum age - contains the value zero.
- Hello time - indicates the time period between root bridge configuration messages.
- Forward delay - the time bridges should wait before transitioning to a new state after a topology change.
Topology-change messages are only 4 bytes long and consist of the
protocol identifier field
, and the
message type field
which contains the value 128.
Source-Route Bridging (Token Ring)
(SRB) algorithm was developed by IBM for Token Ring networks, and gets its
name from the fact that routing information is placed in all
frames by the sending device. Bridges forward frames according to the routing
the frame. A simple source-route bridging network is illustrated below.
A simple source-route bridging network
Assume that computer A wants to send a frame to computer B. Initially, A does not know whether B resides on the same ring or a different one. A therefore sends out a test frame. If the test frame returns without having been acknowledged by B, A assumes that B is on a remote segment. A will now send out an explorer frame . Each bridge receiving the explorer frame (in this case, B1 and B4) will re-transmit the frame onto any other segments connected to it. Route information is added to the explorer frame by each bridge it traverses on its way through the network. When the explorer frames reach B, each receives an individual reply, routed according to its accumulated route information. When A has received the replies, it must choose one of the routes contained therein for future transmissions to B, based on some predetermined criteria.
In most cases, the route yielded by the first reply is chosen, although the
number of hops and the maximum permitted frame size for a particular route may
also influence the selection. Once the route is selected, it is inserted into
routing information field
(RIF) of frames destined for B. A routing information field is only included
in frames that are addressed to devices on another segment. The presence of
routing information within the frame is indicated by setting the most
significant (leftmost) bit within the
Source Address field
. This bit is called the
routing information indicator
The location of the routing information field within a Token Ring data/command
frame is shown below, together with its detailed structure.
The routing information field in a Token Ring data/command frame
The routing information field consists of a single routing control field and a number of routing descriptor fields. These fields and their subfields are described below.
Routing control field
- contains the following subfields:
Routing type (3 bits)
- indicates one of three possible types of routing control:
- Specifically routed - used when the source node supplies the route in the RIF header.
- All paths explorer - used to find a remote node. Route information is collected as the frame traverses the network.
- Spanning-tree explorer - used to find a remote node. Only bridges in the spanning tree forward the frame, reducing the number of frames used for route discovery.
- Length (5 bits) - specifies the total length of the routing information field in bytes.
- Direction bit - indicates the direction in which the route is being traversed, and determines whether bridges read the ring and bridge number combinations in the route designators from right to left (forward) or left to right (reverse).
- Largest frame (6 bits) - indicates the largest frame size that can be handled by the specified route.
- Reserved bit - not used
- Routing type (3 bits) - indicates one of three possible types of routing control:
Routing descriptor fields
- contains the following subfields:
- Ring number (12 bits) - identifies a ring segment (must be unique within the network)
- Bridge number (4 bits) - identifies a bridge (only has to be unique if the bridge it identifies is parallel with another bridge connecting two rings).
The routing information field can contain a number of routing descriptor fields (up to a maximum of 14). Each route is an alternating sequence of ring and bridge numbers that starts and ends with a ring number. Bridges add their bridge number, and the number of the ring onto which the frame is forwarded, to each explorer frame they receive (the first bridge also adds the ring number of the ring from which it receives the frame, and the last bridge number always equals zero).
The switch is a relatively new network device that has replaced both hubs and bridges in LANs. A switch uses an internal address table to route incoming data frames via the port associated with their destination MAC address. Switches can be used to connect together a number of end-user devices such as workstations, or to interconnect multiple network segments. A switch that interconnects end-user devices is often called a workgroup switch. Switches provide dedicated full-duplex links for every possible pairing of ports, effectively giving each attached device its own network segment This significantly reduces the number of intra-segment and inter-segment collisions.
Workgroup switches connect together a number of enduser devices
A number of network segments on the same floor of a building (or on the same campus), each having thier own workgroup switch switch, may themselves be connected together by a higher level switch known as a floor switch. Much more powerful switches are often used to connect together a number of high-level network devices, such as floor switches, workgroup switches and routers. These devices are often called core switches, and they should have sufficient capacity to cope with the volume of traffic flowing around the network.
A core switch connects the high-level devices on the network
A network environment that consists of several interconnected networks employing different network protocols and architectures requires a sophisticated device to manage the flow of traffic between these diverse networks. Such a device, sometimes referred to as an intermediate system, but more commonly called a router, must be able to determine how to get incoming packets (or datagrams) to the destination network by the most efficient route. Routers gather information about the networks to which they are connected, and can share this information with routers on other networks. The information gathered is stored in the router's internal routing table, and includes both the routing information itself and the current status of various network links. Routers exchange this routing information using special routing protocols.
A chassis-based Nortel ERS-8600 routing switch
Computers, and other end-user devices attached to networks that form part of an internetwork, are often called hosts or end-systems. A network host does not know how to forward a datagram to a host on another network, and so it will forward the datagram to its local router (or default gateway). A datagram may traverse a number of networks, and hence a number of routers, as it travels from an end-system on the source network to an end-system on the destination network. At each intermediate router, a decision is made as to what is the optimum next hop. The process undertaken by the router in transferring the incoming datagram to one of its output ports in this way is called switching, and routers are at the heart of packet-switching networks. Unlike bridges and switches, routers do not concern themselves with MAC addresses, and instead examine the IP address contained within a datagram to determine the address of the destination network.