hellhollowreadingΔίκτυα και Επικοινωνίες

26 Οκτ 2013 (πριν από 4 χρόνια και 7 μήνες)

106 εμφανίσεις

A Repeater regenerates the weakened bit stream. It deals only

with bits and is a physical layer device

Traditional Ethernet has a data rate of 10Mbps. Because it is a shared medium network, this

bandwidth is shared between all stations. Collisions are possible when 2 or more stations

attempt to use the network at the same time. In a busy network, collisions are commonplace,

with the result that the bandwidth available to each node is severely reduced.

In (a) above, all stations compete for the available bandwidth. In part (b), a bridge contains the

collisions to the network where they arise. There are less collisions in each network.


is a data link device. It processes frames. It looks at addresses in frame headers to

forward frames. It connects a number of separate network segments.

In part (b) a 4
port bridge connects the 12 stations shown in (a). Now the collision domains are a

lot smaller, and fewer collisions can be expected. The bridge does address filtering. If the number

of bridge ports is equal to the number of connected stations, then each station is on its own

collision domain, i.e it only has to contend with the bridge in the use of the network. A

broadcast message arising on one segment is propagated to all segments by the bridge.The more

segments there are, the more critical the performance of the bridge becomes. A bridge is a serial

device, and must contend with the connected devices to gain access to a given segment.

A switch is like the N
port bridge shown above. See next slide.

In the configuration above the switch is like the multi
port bridge in the previous slide. The

switch is a specialised hardwire unit with several ports. In an Ethernet switch, typically

each port can have either a single station connected, in which case it has all the bandwidth, or

else a hub can be connected where the switch port’s bandwidth is shared between all the stations

connected to the hub. The switch is a high speed backplane to interconnect ports.

A switch acting as a backbone network .

The principle objective in switch design was to overcome the limitation of broadcasting where

only one message can be in transit at any given time. A switch achieves parallelism by having

multiple messages being transmitted through it simultaneously. Today, switch ports are typically

100Mbps, half duplex, and operate the CSMA/CD protocol. When only one device is connected

to a switch port, then the entire bandwidth (i.e 100Mbps) is available to the switch and device.

Both the connected device and the switch must contend for the use of the port.

A store and forward switch
: This type of device accepts a frame on an input port, buffers it,

performs checksum evaluation and forwards it to the appropriate output port.

A cut
through switch
:This type of device takes advantage of the fact that the destination

address is in the beginning of a frame, it begins repeating the incoming frame onto the

appropriate output port without waiting for the entire frame to be received.

In a bridge, frame handling is done in software. A layer 2 switch performs the same address

recognition and frame forwarding in hardware.

A Bridge can typically only analyse and forward one frame at a time, whereas a layer 2

switch has multiple parallel data paths, allowing it to handle multiple frames at a time.

A bridge uses store and forward operation. With a switch, it is possible to have either store

and forward or cut through behaviour. The best performance is with cut
through switches.

Most LANs are now based on switches.

A switch will typically communicate with attached devices via CSMA/CD at the ports. However,
internally, how a switch sets up connections between input and output ports is independent of the
technology of the ports. Switches will be provisioned with memory buffers where incoming
frames will be examined and temporarily held until they output onto the appropriate port. A given
output port may be temporarily unavailable (already in use) so buffering is necessary. There may
also be periods where the net inflow into a switch may be more than the net outflow….buffers.
Notice that there are many parallel connections through the switch simultaneously.

ATM Switches

LAN Switches

Higher Port density at lower cost than Bridges


Fewer users per segment, more bandwidth per user

segmentation, allows the creation of dedicated segments, one segment per user,
full bandwidth available, no contention.

There can be multiple simultaneous conversations through a switch. Not restricted to a
single message at a time.


Provides full duplex conversations, doubles throughput.

Can operate at 10 or 100 Mbps.

No changes to existing hubs, network interface cards, or cabling.

Allows the creation of VLANs, an aggregation of users per LAN based on needs.

Switches learn topology, forward, and filter like bridges

Media Rate adaptation 10/100Mbps

through on per port basis

Standard Ethernet, Switched

Ethernet, Fast Ethernet, Full duplex Ethernet, Gigabit Ethernet

Traditional Ethernet

is based on CSMA/CD, and is a broadcast network. It is half duplex, a

station cannot send and receive simultaneously. The transmission rate is 10Mbps. In an

Ethernet LAN, a station which accesses the network transmits at wire speed. (either 10, 100,

or 1000 Mbps). The great limitation of broadcasting is that there can be only one message

being transmitted at a time. Because of collisions, the bandwidth available to connected devices

can be seriously less than 10Mbps, and depends on the number of connected devices and how

busy they are. Ethernet does not perform well under heavy load.

Fast Ethernet

is a development of traditional Ethernet where the network transmission speed is

100Mbps. It is totally backwards compatible with traditional Ethernet. The same cabling can be

used and the access protocol (CSMA/CD) is the same. It is half duplex.

Switched Ethernet

is a further development of Ethernet. Whereas traditional and fast Ethernet

are both based on broadcasting where only a single message can be in transit over the network

at a given time, Switched Ethernet is based on circuit switching and achieves parallelism

whereby multiple messages can be in transit through the network at a given time. Several input

ports can be connected simultaneously to output ports. Switched Ethernet achieves very

significant gains over traditional or even fast Ethernet.

Full Duplex Ethernet

is a development which allows switches to both send and receive over its

ports simultaneously. In a full duplex Ethernet switch, where only a single device is connected to

its ports, there are no collisions, and CSMA/CD is not necessary. The connection between the

port and the device is a point
point link, and the device has the full bandwidth to itself.

Gigabit Ethernet

has a network speed of 1000Mbps, and is typically full duplex. Now 10Gbps

Ethernet is available.


Used to group a set of users together in a logical grouping independently of which
network segments they are attached to. e.g all marketing people that need to share
same data, and have a common interest in broadcast messages. With a VLAN, the
users can be spread over several buildings, be physically connected to different
switches, yet be logically part of the same VLAN.

VLANs can be defined in terms of the switch port to which a given machine is
connected, or based on the MAC address of the connected machine, or on the IP
address, or the type of application being used, or on a combination of the above.

VLAN traffic must always go through a router, and may be subject to security

Broadcasts are always contained within a VLAN, i.e a broadcast on one VLAN is not
visible on another VLAN.

Communication between switches

In a multi switched backbone each switch must know not only it own VLAN membership, but

it must also know the VLAN membership of other switches. For example, when a broadcast

message for the members of VLAN 1 originates with a member connected to Switch 1, then

that switch must forward that message to other members connected to other switches.

Communication between switches can be based on membership tables, frame tagging or Time

Division Multiplexed channels between the switches.

In the case of membership tables, switches exchange messages which inform each other of

VLAN membership.

Frame Tagging is based on tagging an extra header to each inter
frame to identify the

VLAN to which it belongs. The IEEE 802.1Q standard governs frame tagging.

With TDM, the trunk between switches is divided into channels, a given VLAN is

allocated a dedicated channel over which messages destined for its members travel.