Difference between Router Gateway Repeater Hub Bridge Switch ?

[Kelsey]

Hello Every i Want to know What is the Basics Difference between the Following Networking Devices Repeater, hub, bridge, switch, router, gateway ? All kind of Comments Are Welcome Thanks in Advance

[Marlon]

Repeater

• Forwards every frame it receives
• it is a generator,not an amplifier(i.e it removes noise & regenerates signal )
• Bi-directional in nature
• Useful in increasing ethernet size/length
• Maximum of 5 Repeaters in an Ethernet

[Marlon]

Hub

• basically a multiport repeater
• can be used to divide a single LAN into multiple levels of Hierarchy

[Marlon]

Bridge

• connect similar/dissimilar LANS
• Designed to store and forward frame
• Protocol independent
• Transparent to End Stations
• Operates in Layer-1 & Layer-2
• uses a table for filtering/routing
• does not change the Mac address in the frame

[Marlon]

Switch
A fast Bridge.

• Common forwarding approaches used include Cut through,collission free & Fully buffered

Routers

• Links dissimilar n/ws
• not transparent to end stations
• Acts on a network layer frame
• isolates LAN to subnets to manage & control traffic

Gateway

• uses Layer -7 Relay

I Hope You Doubt Are Clear

[DAIJIRO]

Repeater

• Forwards every frame it receives
• it is a generator,not an amplifier(i.e it removes noise & regenerates signal )
• Bi-directional in nature
• Useful in increasing ethernet size/length
• Maximum of 5 Repeaters in an Ethernet

Yes - a repeater repeats frames at the electrical level, layer 1. Is part of a collision domain. I cannot remember a rule of thumb for how many repeaters can be had in sequence but I think it is really governed by the timing budget. There has to be enough time for the two most widely separated hosts to mutually detect a collision (and respond with a jamming signal) within the collision period which I think for Ethernet is 64 bit times after the preamble. Maybe someone can correct that if necessary.

[DAIJIRO]

Hub

• basically a multiport repeater
• can be used to divide a single LAN into multiple levels of Hierarchy

A hub is a repeater, certainly. I wouldn't say that a hub divides a network in any way - since it is just a repeater. Still operates at layer 1 only. Still a component of a single collision domain.

The term hub at one time was split into

• repeating hub - what we now call a hub
• switching hub - what we now call a switch

[DAIJIRO]

Bridge

• connect similar/dissimilar LANS
• Designed to store and forward frame
• Protocol independent
• Transparent to End Stations
• Operates in Layer-1 & Layer-2
• uses a table for filtering/routing
• does not change the Mac address in the frame
• two types of bridges are there Transparent/source routing

A bridge is/was a device for connecting networks into separate collision domains. As you say it operates at layers 1 and 2 and it was originally a 2-port device. Normal bridging is transparent - e.g. Ethernet to Ethernet. If you are masochistic and have stuff like token ring or even FDDI I believe some translational bridges were made to convert the frames as best as was possible between one and the other.Better, IMHO, to keep each lan technology separated by routers.

Either way a bridge, while splitting the network into collision domains, is a component of a broadcast domain. Collisions stop at its ports but broadcasts pass through. It won't protect the network from broadcast storms and the like.

[DAIJIRO]

Switch

• A fast Bridge.
• Common forwarding approaches used include Cut through,collission free & Fully buffered

Yes, a switch is normally much faster than a bridge. Bridges tended to have CPUs to handle the frame forwarding and table management. Switches, on the other hand, tend to have ASICs to do what they do and switches normally have multiple ports as befits their original name of a switching hub.

Cisco support (at least on some switches) a choice between cut-through (start to transmit as soon as the frame starts to come in and the Output port is free - same speed ports only), collision-free (wait for at least the collision period before starting to transmit) and store-and-forward. More modern Cisco switches may not support all three approaches. LANs are generally so reliable that cut-through is best if in and out port speeds are the same. Note that for full duplex Ethernet which is the most common LAN technology and mode used with switches there is no possibility of a collision.

[DAIJIRO]

Routers

• Links dissimilar n/ws
• not transparent to end stations
• Acts on a network layer frame
• isolates LAN to subnets to manage & control traffic

Well, routers focus on work at layer 3. They _can_ link layer 2 network technologies which are different but are just as happy working between ones which are alike. The protocol data units (PDUs) they work on are traditionally called packets. (PDUs at layer 2 are called frames.) Routers terminate broadcast domains. Yes, routers do separate LANs but that can be incidental to their primary functions which include getting sending packets in the right direction based on their routing tables. They can usually apply a range of traffic control measures including filtering, traffic shaping and reporting.

[DAIJIRO]

Gateway

• uses Layer -7 Relay

The meaning of "gateway" has changed over the years and tends to depend on context. They generally convert one protocol to another or provide a conversion of packets between one system and a dissimilar one. A voice gateway may translate between VoIP and normal phone lines, for example.

The term originally was used for routers (but is not used that way any longer). The hangover from the old days is that end stations are configured with a default router which we still call the default gateway.

A final point on your Layer 7 comment. You are right that the highest level of the ISO model protocol stack is 7 for applications but bear in mind that the TCP/IP suite works on a reduced stack something like the following.

1. Physical - electrical and timing (and connectors)
2. Data link - basically the LAN technology - works on frames
3. Network - works on packets using IP addresses
4. Session & transport mixed together - works on end-to-end connections or transactions using in nearly all cases TCP and UDP

Above that there is only the applications. There is no presentation layer at all unless the applications include it within themselves.

[Marco-D]

You've made a good list of features already. I'll add some suggested differences.



Yes - a repeater repeats frames at the electrical level, layer 1. Is part of a collision domain. I cannot remember a rule of thumb for how many repeaters can be had in sequence but I think it is really governed by the timing budget. There has to be enough time for the two most widely separated hosts to mutually detect a collision (and respond with a jamming signal) within the collision period which I think for Ethernet is 64 bit times after the preamble. Maybe someone can correct that if necessary.

There are different rules for different ethernet media, but I believe 5 is right for 10baseT. In the usual 10baseT case it is determined by loss of inter-packet gap, and not by round trip time limits for collision detect. Even with six 150m cables you will still be below the time limit. Five is convenient as it allows for a central repeater connected to repeaters, each connected to more repeaters

[DAIJIRO]

There are different rules for different ethernet media, but I believe 5 is right for 10baseT. In the usual 10baseT case it is determined by loss of inter-packet gap, and not by round trip time limits for collision detect. Even with six 150m cables you will still be below the time limit. Five is convenient as it allows for a central repeater connected to repeaters, each connected to more repeaters

hmmm - I undersood that repeaters were permitted to,and did, eat the preamble during clock synchronisationand that this was a constraint to the number of repeaters. 5 is correct by the way. I have the idea that later repeaters synchronised more quickly and ate fewer bits but that the standard was not changed. Not at all sure tho' and to be honest no longer care

Seems like a homework question.

[Marlon]

There are different rules for different ethernet media, but I believe 5 is right for 10baseT. In the usual 10baseT case it is determined by loss of inter-packet gap, and not by round trip time limits for collision detect. Even with six 150m cables you will still be below the time limit. Five is convenient as it allows for a central repeater connected to repeaters, each connected to more repeaters

Well, there are the model 1 rules and the model 2 rules. The model 2 rules are more exact and more complicated. Yes, by now if you are still using multiple repeaters, you should at least have one switch which allows for a really large number of repeaters.