[pohu]

hi guys...

just to clarify some doubts..

the device that use to set up a home/small network, should call it a "router" (be it wireless or non-wireless type) or "hub"??

Cos, l read from somewhere... "router" is something that use to link up huge network such as the internet, whereas the "hub" is use in small area network such as home and small office..

[Cyberbaby]

router and hub is two different gadget.

[reb00tz]

OK, here's a quick low-down on the different types of connection equipment...

HUB:
It's the "dumbest" of the lot - it simply repeats whatever that comes in on one port to the other ports (broadcasting). No intelligence what so ever. Also, all ports "share" the bandwidth (meaning, if there are four ports in use on a 100MBps hub, then all four share that 100MBps theoretical bandwidth.

Also, due to the fact that it simply broadcasts traffic to all ports, all ports must be connected at half duplex (HD mode for short). What this means is that all the equipment connected to that hub can only "talk or listen" at any one time. The reason for this is beyond the scope of this post. This results in all ports sharing the maximum bandwidth of the hub.

SWITCH:
This is an "improvement" over a hub - it actually "learns" what MAC address (the physical address or "name" of the connected equipment) on each port, and "creates" a one-to-one link. Thus, no traffic is broadcast throughout the switch like a hub. This means that each port can fully utilize the full bandwidth of a switch, unlike a hub; it also means all ports can operate at full duplex (FD mode).

Say, with four NICs connected, NIC A talks to NIC C, and NIC B talks to NIC D... This means NICs A & C get full 100MBps FD, and NICs B & D also get 100MBps (disclaimer below). Now, if NICs A & B both try to talk to NIC C at the same time, then the switch will "switch" the bandwidth back and forth between NICs A & B (hence the name), thus NIC C gets full 100MBps to both, but only (theoretically) half of that to each (and vice versa). Note that they're all still talking 100MBps FD rates, but due to the switching, they each only get "time slots" to talk at a theoretical rate of half that of 100MBps... Add more, and the ratio goes up, and the theoratical rate is inversely proportionate.

This is why servers usually are placed on 1GBps ports, with workstations on 100MBps ports on the same switch (which usually is pretty high-end in the first place to have mixed Gigabit and Fast Ethernet ports). This means the server can talk at full rate to any 10 of the workstations (at their full rates also) at the same time.

The "switching" mechanism also adds an extra "benefit" of switches (beyond enabling FD mode) - different rates can be hooked up to the switch, for example, if you have a Gigabit switch, you can hook up Ethernet (10MBps), Fast Ethernet (100MBps) and Gigabit Ethernet (1000MBps) ports and they will still be able to talk to each other, which won't happen with a hub.

The disclaimer is this: the maximum number of "simultaneous" one-to-one "conversations" that any switch can support is dependent solely on its backplane. This switching backplane (compared to the broadcast backplane in a hub) is what connects all the different ports together. Now, if a Fast Ethernet (100MBps) switch has a backplane with a maximum rate of 200MBps, then effectively only two 100MBps FD rate connections can be supported, or 1x 100MBps and 10x10MBps one-to-one connections can be supported, or 20x10MBps one-to-one connections can be supported (you go do the math).

Hence, the need to actually worry about backplanes when you are buying a Gigabit switch.


If I haven't lost you yet, then let's get to the interesting term, Routers. This portion assumes you understand the basic premise of IP networks, addressing and subnetting.

ROUTER:
In actual terms, a router is one that routes packets between networks. Say, you have two different networks, 10.0.0.1/24 on port A, and 192.168.0.1/24 on port B. If you place a router in between, you would probably have to program it such that any packets coming in with a destination of 192.168.0.0/24 should be "dumped" out onto port B, and any packets that it received destined for 10.0.0.0/24 should be "dumped" out onto port A.

This is very much akin to Window's own "route" command.

Routers typically work on a single network layer, which is layer 3 of the OSI (which I won't get into here). Suffice to say that routers require much more intelligence then either switches OR hubs, and as you may have realized, it operates on IPs (according to my example), one level "above" switches and hubs which operate on the physical media layer.

UNFORTUNATELY, this term has so far been "bastardized" by the common markets... "Routers" as a term is often used interchangeably (albeit incorrectly) with "Broadband Routers". We will get to that common term later.

BRIDGES:
Now, here is one term that most of us don't hear (and we usually don't need to hear). A bridge is simply a device that connects different types of networks. Back when IPX/SPX was still in use, bridges were used to connect those to a TCP/IP network.

Nowadays, bridges are used for different purposes, for example bridging a wireless LAN and a wired one. Note that a wireless bridge is somewhat different from a wireless access point (though its arguably similar in purpose).

NAT Devices:
Now, you won't find a "NAT device" advertised ANYWHERE in ANY shop. A NAT device is a hardware device that does NATting (Network Address Translation). Now, this MIGHT start to ring a bell for those not quite in the know... Sounds familiar to something that's in your (incorrectly termed) "broadband router", right?

Well, NATting is simply taking all packets from one side of a network, "fake" the "from" IP address, then sending it out to another network. Once a response is returned, it will revert the "to" address to the originating address on the originating side of the network. Other Linux geeks call it IP Masquerading...

How and why and the reason this exists is once again beyond the scope of this post.

BROADBAND ROUTERS:
Now, nearly everybody has one. If you understand the above, you would probably conclude that this is a bastardized term that simply means a hardware device that is no more than a NATting switch!

To be fair, most of these have some sort of routing intelligence built in, so I suppose the proper name would be "NATting switch router", which also concludes why I am not in the marketing department.

To add reason to insanity, let us, for arguments sake, throw wireless into the fire. Note that I say "add", meaning that the broadband router now has wireless capabilities ADDED to wired ports. Then, the "broadband wireless router" suddenly becomes a "NATting bridge switching router". Heh, I must have flunked Marketing 101 or something.

I rest my case.

*If I have anything technically wrong, I apologise. I don't have a degree in networking, just alot of accumulated crap in my brain.

[international]

The Prices are also different

[Zoomer]

Don't worry, its quite correct.

To add on:

Different types of Hubs
Hubs are physical layer devices, or layer 1.

Powered Hubs
Repeats and amplifies signals. Typically needs a power source.

Thus, length of cabling to each port can be 100m.

Most hubs today are powered.

Unpowered Hubs
Signal not amplified.

Thus, connection from any port to any other port cannot exceed 100m.
[/u]

Repeaters
Amplifies singal to overcome the 100m limitation

[yantronic]

Use a hub to share internet. Try sharing files on a hub and you will never make it.

This is because Windows recognises hubbed network cable to carry 10Mbps. Broadband never hit that yet so nevermind.

Try and transfer 400 MB on a hub and you will know why. Its a pain in the arse.

[hangyong]

Originally Posted by yantronic

Use a hub to share internet. Try sharing files on a hub and you will never make it.

This is because Windows recognises hubbed network cable to carry 10Mbps. Broadband never hit that yet so nevermind.

Try and transfer 400 MB on a hub and you will know why. Its a pain in the arse.

you know, it seriously depends very much on how many PCs are sharing the hub. if 3~4 PCs, all trying to send a file out 400MB in size, not a prob at all. I have been transfering almost 9GB file daily previously, via a 100mbps hub from PC1 to PC2, while PC3 is trying to access the web or do other things. No problems there at all.

Between a hub and a switch, there basically is not difference in speed, just that hubs will have collision prob while switch does not. And you will not be able to tell the difference, since the collision correction will not take more than 120ms.

Even if you are using a hub, and the network is busy, you have to wait in the queue as well, and once the buffer fills, you have to wait again....

[yantronic]

Originally Posted by hangyong

you know, it seriously depends very much on how many PCs are sharing the hub. if 3~4 PCs, all trying to send a file out 400MB in size, not a prob at all. I have been transfering almost 9GB file daily previously, via a 100mbps hub from PC1 to PC2, while PC3 is trying to access the web or do other things. No problems there at all.

Between a hub and a switch, there basically is not difference in speed, just that hubs will have collision prob while switch does not. And you will not be able to tell the difference, since the collision correction will not take more than 120ms.

Even if you are using a hub, and the network is busy, you have to wait in the queue as well, and once the buffer fills, you have to wait again....

Haha, I am a local machine person, still have problem understanding all of that. Thanks for the insight!

[reb00tz]

Originally Posted by hangyong

you know, it seriously depends very much on how many PCs are sharing the hub. if 3~4 PCs, all trying to send a file out 400MB in size, not a prob at all. I have been transfering almost 9GB file daily previously, via a 100mbps hub from PC1 to PC2, while PC3 is trying to access the web or do other things. No problems there at all.

Between a hub and a switch, there basically is not difference in speed, just that hubs will have collision prob while switch does not. And you will not be able to tell the difference, since the collision correction will not take more than 120ms.

Even if you are using a hub, and the network is busy, you have to wait in the queue as well, and once the buffer fills, you have to wait again....

For the most part, the limits between hub and switch are theoretical for most home users. The differences will be seen when network saturation is above 40% on a network, where a hubbed network starts to get ALOT of collisions, compared to a switched network.

[isidore]

Actually,
You can see a huge difference between a hub, switch and router quite obviously by looking at the OSI model and how each device is able to manipulate.

Layer 1 - Physical Layer. basically the transmisssion of a stream (wave) of impulses (one and off). Hubs operate at this layer, they are not aware of other layers.

Layer 2 - Data Link Layer / Media Access Control (MAC) layer. Converts the raw stream as described in layer one into a bit stream (packet). Does Flow Control / Error Correction if required. Switches operate on this layer, as they distinguish each packet not by its contents but simply by its physical MAC address, then forwrads it off the correct port.

Layer 3 - Network, such as TCP/IP or IPX or AppleTalk, etc. This builds upon the previous layer by now having a network address and as such routing (distinguishing of the IP subnet by the router and thus deciding what port to forward out on) is done on this level.

Layer 4 - Transport.

Layer 5 - Session (To join many packets into a "connection").

Layer 6 - Presentation

Layer 7 - Application. Normal everyday things like HTTP Server, DHCP Server etc run on these.

===

Thus, its hard to actually pinpoint what end of the model the "home broadband router with wireless" can be put at. From above, we can see that it is able to discern from Level 3 as it is able to discern and forward data based on the IP which it comes from. Some people call it the "Layer 3 Switch" ie, switch that can discriminate on IP addresses, some do call it a router. Then again, we notice that it can also handle some Layer 7 functions like HTTP or TELNET or even sending requests up to those DDNS servers - however do note that it does NOT discriminate on where packets are headed based on Layer 7 Protocols (something that does that would be a stateful packet firewall). I think it can be called a router since it does channel packets to 2 seperate networks. When you look at it, a broadband router comprises of a switch, a wireless chip and a mini cpu they use to do NAT / DHCP and routing in it.

By the way, a switch is also a bridge. The definition of a bridge, strictly, is actually "a device that connections 2 segments together". It does not have to be 2 seperate types of media.

[cyberengine]

Originally Posted by reb00tz

OK, here's a quick low-down on the different types of connection equipment...

HUB:
It's the "dumbest" of the lot - it simply repeats whatever that comes in on one port to the other ports (broadcasting). No intelligence what so ever. Also, all ports "share" the bandwidth (meaning, if there are four ports in use on a 100MBps hub, then all four share that 100MBps theoretical bandwidth.

Also, due to the fact that it simply broadcasts traffic to all ports, all ports must be connected at half duplex (HD mode for short). What this means is that all the equipment connected to that hub can only "talk or listen" at any one time. The reason for this is beyond the scope of this post. This results in all ports sharing the maximum bandwidth of the hub.

SWITCH:
This is an "improvement" over a hub - it actually "learns" what MAC address (the physical address or "name" of the connected equipment) on each port, and "creates" a one-to-one link. Thus, no traffic is broadcast throughout the switch like a hub. This means that each port can fully utilize the full bandwidth of a switch, unlike a hub; it also means all ports can operate at full duplex (FD mode).

Say, with four NICs connected, NIC A talks to NIC C, and NIC B talks to NIC D... This means NICs A & C get full 100MBps FD, and NICs B & D also get 100MBps (disclaimer below). Now, if NICs A & B both try to talk to NIC C at the same time, then the switch will "switch" the bandwidth back and forth between NICs A & B (hence the name), thus NIC C gets full 100MBps to both, but only (theoretically) half of that to each (and vice versa). Note that they're all still talking 100MBps FD rates, but due to the switching, they each only get "time slots" to talk at a theoretical rate of half that of 100MBps... Add more, and the ratio goes up, and the theoratical rate is inversely proportionate.

This is why servers usually are placed on 1GBps ports, with workstations on 100MBps ports on the same switch (which usually is pretty high-end in the first place to have mixed Gigabit and Fast Ethernet ports). This means the server can talk at full rate to any 10 of the workstations (at their full rates also) at the same time.

The "switching" mechanism also adds an extra "benefit" of switches (beyond enabling FD mode) - different rates can be hooked up to the switch, for example, if you have a Gigabit switch, you can hook up Ethernet (10MBps), Fast Ethernet (100MBps) and Gigabit Ethernet (1000MBps) ports and they will still be able to talk to each other, which won't happen with a hub.

The disclaimer is this: the maximum number of "simultaneous" one-to-one "conversations" that any switch can support is dependent solely on its backplane. This switching backplane (compared to the broadcast backplane in a hub) is what connects all the different ports together. Now, if a Fast Ethernet (100MBps) switch has a backplane with a maximum rate of 200MBps, then effectively only two 100MBps FD rate connections can be supported, or 1x 100MBps and 10x10MBps one-to-one connections can be supported, or 20x10MBps one-to-one connections can be supported (you go do the math).

Hence, the need to actually worry about backplanes when you are buying a Gigabit switch.


If I haven't lost you yet, then let's get to the interesting term, Routers. This portion assumes you understand the basic premise of IP networks, addressing and subnetting.

ROUTER:
In actual terms, a router is one that routes packets between networks. Say, you have two different networks, 10.0.0.1/24 on port A, and 192.168.0.1/24 on port B. If you place a router in between, you would probably have to program it such that any packets coming in with a destination of 192.168.0.0/24 should be "dumped" out onto port B, and any packets that it received destined for 10.0.0.0/24 should be "dumped" out onto port A.

This is very much akin to Window's own "route" command.

Routers typically work on a single network layer, which is layer 3 of the OSI (which I won't get into here). Suffice to say that routers require much more intelligence then either switches OR hubs, and as you may have realized, it operates on IPs (according to my example), one level "above" switches and hubs which operate on the physical media layer.

UNFORTUNATELY, this term has so far been "bastardized" by the common markets... "Routers" as a term is often used interchangeably (albeit incorrectly) with "Broadband Routers". We will get to that common term later.

BRIDGES:
Now, here is one term that most of us don't hear (and we usually don't need to hear). A bridge is simply a device that connects different types of networks. Back when IPX/SPX was still in use, bridges were used to connect those to a TCP/IP network.

Nowadays, bridges are used for different purposes, for example bridging a wireless LAN and a wired one. Note that a wireless bridge is somewhat different from a wireless access point (though its arguably similar in purpose).

NAT Devices:
Now, you won't find a "NAT device" advertised ANYWHERE in ANY shop. A NAT device is a hardware device that does NATting (Network Address Translation). Now, this MIGHT start to ring a bell for those not quite in the know... Sounds familiar to something that's in your (incorrectly termed) "broadband router", right?

Well, NATting is simply taking all packets from one side of a network, "fake" the "from" IP address, then sending it out to another network. Once a response is returned, it will revert the "to" address to the originating address on the originating side of the network. Other Linux geeks call it IP Masquerading...

How and why and the reason this exists is once again beyond the scope of this post.

BROADBAND ROUTERS:
Now, nearly everybody has one. If you understand the above, you would probably conclude that this is a bastardized term that simply means a hardware device that is no more than a NATting switch!

To be fair, most of these have some sort of routing intelligence built in, so I suppose the proper name would be "NATting switch router", which also concludes why I am not in the marketing department.

To add reason to insanity, let us, for arguments sake, throw wireless into the fire. Note that I say "add", meaning that the broadband router now has wireless capabilities ADDED to wired ports. Then, the "broadband wireless router" suddenly becomes a "NATting bridge switching router". Heh, I must have flunked Marketing 101 or something.

I rest my case.

*If I have anything technically wrong, I apologise. I don't have a degree in networking, just alot of accumulated crap in my brain.

whoa....so detailed...beri gd...but NAT and DHCP configuration are onli support in proper routers like Cisco 2600 which dun includes those we use @ home......

[lennardseah]

Originally Posted by cyberengine

whoa....so detailed...beri gd...but NAT and DHCP configuration are onli support in proper routers like Cisco 2600 which dun includes those we use @ home......

nonsense.

most home routers have them too

[lennardseah]

BRIDGES:
Now, here is one term that most of us don't hear (and we usually don't need to hear). A bridge is simply a device that connects different types of networks. Back when IPX/SPX was still in use, bridges were used to connect those to a TCP/IP network.

the correct definition should be:
A bridge device filters data traffic at a network boundary. Bridges reduce the amount of traffic on a LAN by dividing it into two segments.
Bridges operate at the data link layer (Layer 2) of the OSI model. Bridges inspect incoming traffic and decide whether to forward or discard it. An Ethernet bridge, for example, inspects each incoming Ethernet frame - including the source and destination MAC addresses, and sometimes the frame size - in making individual forwarding decisions.

Bridges serve a similar function as switches, that also operate at Layer 2. Traditional bridges, though, support one network boundary, whereas switches usually offer four or more hardware ports. Switches are sometimes called "multi-port bridges" for this reason

[isidore]

Originally Posted by lennardseah

nonsense.

most home routers have them too

:shoot4:

LOL!

[lennardseah]

Originally Posted by isidore

:shoot4:

LOL!

nice smilies pack