TR Forums

I want to split my ethernet signal so it can run to two routers. One wireless and one wired.

What is the best option? Switch or hub? Am I right in thinking that the switch will allocate internet traffic to where it is wanted and the hub just duplicates the incoming information both sides?

You want just one device to connect to the internet with all of the local network address translation going through it.
What you want to do is probably something like this:

wall -- modem -- router 1 -- WAP

If you need more ports than you get in that configuration, add a switch plugged into router 1. 8-port gigabit switches are quite affordable.

Sikthskies wrote:

Am I right in thinking that the switch will allocate internet traffic to where it is wanted and the hub just duplicates the incoming information both sides?

Essentially, yes. A hub can consequently produce collisions, which means that packets are lost and the source will have to time out and then re-send. There's really no reason to use a hub anymore, even if you have an old one lying around; proper Ethernet switches are extremely cheap.

Yeah, hubs were a pain in the ass back in the early 90s when real switches were expensive enough to make you pause. There's no reason to use one now.

Also, for most small networks you only want to have one router actually acting as a router (ie giving out IP addresses). Unless you want to get into partitioning subnets for some reason, I'd just turn off the routing part of your wireless "router" and have it act purely as an access point, deferring DHCP to the wired router (assuming your topology matches what JAE suggests)

AFAIK you can't even get a gigabit hub (as in, they don't exist). And 100 mbit hubs are pretty rare these days as well (they were more common back when 100 mbit Ethernet was first introduced).

In other words... you're getting a switch. Unless you want to go back to the stone age (10 mbit).

just brew it! wrote:

In other words... you're getting a switch. Unless you want to go back to the stone age (10 mbit).

Stone age to me is defined as no nub/switches and we go coax, T-connectors, and terminators. (still have like 4 cards plus cabling+connectors to build one)

Sikthskies wrote:

I want to split my ethernet signal so it can run to two routers. One wireless and one wired.

What is the best option? Switch or hub? Am I right in thinking that the switch will allocate internet traffic to where it is wanted and the hub just duplicates the incoming information both sides?

As far as I know, you cannot buy hubs anymore. If you want to use a hub, you would probably be best off physically splicing the cables so that the same cable serves 3 different points. That is basically what a hub does, except it puts that functionality into a nice neat package.

The big issue with that is that you are limited to maximum bandwidth of 37Mbps each way if all three NICs that connect to the cable are in use at once, assuming you are using "fast ethernet". With a switch, which is basically a computer with a bunch of NICs, there are only two computers in the collision domain, so they can both transmit at close to the theoretical maximum throughput of the line.

just brew it! wrote:

AFAIK you can't even get a gigabit hub (as in, they don't exist). And 100 mbit hubs are pretty rare these days as well (they were more common back when 100 mbit Ethernet was first introduced).

In other words... you're getting a switch. Unless you want to go back to the stone age (10 mbit).

I think you can use an old 10Mb hub as a 1Gb hub. As far as I know, the electrical connections are the same. You might not be able to get the same cable lengths with it in place, as the wires involved need to be made to a tighter specification to get 1Gbps operation over 100 meter distances, which to my knowledge, is what the specification requires.

Using a T-adapter like someone else mentioned in the thread would do the same thing and I believe would be compliant with the specification.

Shining Arcanine wrote:

As far as I know, the electrical connections are the same.

The pinout on gigabit changed to use all four pairs. As for the rest, I don't know what fundamental you are overlooking to think a 10 Mbps hub is capable of gigabit operation.

Flying Fox wrote:

just brew it! wrote:

Unless you want to go back to the stone age (10 mbit).

Stone age to me is defined as no nub/switches and we go coax, T-connectors, and terminators. (still have like 4 cards plus cabling+connectors to build one)

Bah, 10Base2 is modern stuff, Stone age is 10Base5 with vampire taps - now get off my lawn!

Shining Arcanine wrote:

As far as I know, you cannot buy hubs anymore. If you want to use a hub, you would probably be best off physically splicing the cables so that the same cable serves 3 different points. That is basically what a hub does, except it puts that functionality into a nice neat package.

While hubs do in fact retransmit the signal on all ports, it is not at all like just wiring the ports together physically. All twisted pair based Ethernet (10, 100, and 1000 mbit) are fundamentally point-to-point links -- for 10/100 mbit, one pair is transmit, and one is receive; for 1000 mbit there are two transmit pairs and two receive pairs. How exactly do you propose to wire 3 NICs together such that they're all transmitting and receiving on the proper pairs?

A hub also provides impedance matching, and detects collisions (transmitting a "jam" signal on all ports to ensure that all NICs detect the collision properly).

In short, splicing the cables together as you suggest will not work for a number of reasons, both physical and logical.

Shining Arcanine wrote:

I think you can use an old 10Mb hub as a 1Gb hub. As far as I know, the electrical connections are the same.

No, not even close. A hub doesn't just bridge the wires together physically, it receives signals on the receive pair(s) of each port, and retransmits those signals on the transmit pair(s) of all other ports. There's active circuitry involved, and for a 10Mb hub that circuitry won't even work at 100 mbit speeds (let alone 1000 mbit).

The electrical connections are different too -- 1000 mbit uses all 4 pairs of wires in the cable, whereas 10/100 uses just 2 of the pairs.

At best, you are confusing modern twisted pair Ethernet with "old school" 10Base2 Ethernet, which used a single coaxial cable in a shared bus arrangement.

Shining Arcane, man, I want some of what you're smoking.

Flying Fox wrote:

Stone age to me is defined as no nub/switches and we go coax, T-connectors, and terminators. (still have like 4 cards plus cabling+connectors to build one)

Nah, stone age is carrying around floppies. 5.25" floppies.
No wait, 8" floppies.
No wait, tape.
No, punch cards. Now that is a sneaker-net.

UberGerbil wrote:

Shining Arcane, man, I want some of what you're smoking.

Flying Fox wrote:

Stone age to me is defined as no nub/switches and we go coax, T-connectors, and terminators. (still have like 4 cards plus cabling+connectors to build one)

Nah, stone age is carrying around floppies. 5.25" floppies.
No wait, 8" floppies.
No wait, tape.
No, punch cards. Now that is a sneaker-net.

personally I use wax tablets.

UberGerbil wrote:

...
No, punch cards. Now that is a sneaker-net.

For any appreciable amount of data, it is more like forklift-net...

The theoretical maximum bit density for an IBM punch card was 40 bits per square inch, but you couldn't achieve this in practice because punching out too large a percentage of the holes would result in a card which lacked sufficient mechanical strength to be read reliably (the most extreme case being a so-called lace card).

Modern hard drive platters have a bit density in the hundreds of gigabits per square inch. That's 10 orders of magnitude more dense than a punch card!

cheesyking wrote:

personally I use wax tablets.

But then you need fans and heatsinks to keep them cool, because they lose their data integrity at just above room temp. Installing your wax tablets in a hot case is not advised. Though at least it makes rewrite easy. I prefer my tablets be clay for data integrity reasons, but they're WORM storage. And they're not especially suited for mobile applications (they're heavy, and can't tolerate sudden G loads). Talk about needing forklifts for significant data storage...

On the other hand, considering we're still able to retrieve accounting data from 4000+ year old clay tablets, data retention issues are clearly solved for this technology -- something that is still somewhat unresolved for most of the techs we're using today.

UberGerbil wrote:

On the other hand, considering we're still able to retrieve accounting data from 4000+ year old clay tablets, data retention issues are clearly solved for this technology -- something that is still somewhat unresolved for most of the techs we're using today.

This is selection bias at its finest--the tablets that failed are dust by now. I'm convinced that in a thousand years, monks will be able to recover data from the half-dozen or so surviving hard drives. Perhaps they will be enlightened enough to recognize lolcats for what it is: a display of collective madness which ultimately led to the downfall of civilization.

Contingency wrote:

This is selection bias at its finest--the tablets that failed are dust by now. I'm convinced that in a thousand years, monks will be able to recover data from the half-dozen or so surviving hard drives. Perhaps they will be enlightened enough to recognize lolcats for what it is: a display of collective madness which ultimately led to the downfall of civilization.

While the magnetic patterns on the platters might survive that long, the mechanical components and lubricants used thereon will have long since degraded into uselessness. They would need to build a new drive around the existing platters; if civilization has fallen, I highly doubt they'd be able to do this.

Stone tablets have the advantage of not requiring any special equipment to read the patterns on their surface.

(Jeez, we got pretty far off-topic here, didn't we? Maybe it is time for a thread split...)

just brew it! wrote:

Shining Arcanine wrote:

As far as I know, you cannot buy hubs anymore. If you want to use a hub, you would probably be best off physically splicing the cables so that the same cable serves 3 different points. That is basically what a hub does, except it puts that functionality into a nice neat package.

While hubs do in fact retransmit the signal on all ports, it is not at all like just wiring the ports together physically. All twisted pair based Ethernet (10, 100, and 1000 mbit) are fundamentally point-to-point links -- for 10/100 mbit, one pair is transmit, and one is receive; for 1000 mbit there are two transmit pairs and two receive pairs. How exactly do you propose to wire 3 NICs together such that they're all transmitting and receiving on the proper pairs?

A hub also provides impedance matching, and detects collisions (transmitting a "jam" signal on all ports to ensure that all NICs detect the collision properly).

In short, splicing the cables together as you suggest will not work for a number of reasons, both physical and logical.

Shining Arcanine wrote:

I think you can use an old 10Mb hub as a 1Gb hub. As far as I know, the electrical connections are the same.

No, not even close. A hub doesn't just bridge the wires together physically, it receives signals on the receive pair(s) of each port, and retransmits those signals on the transmit pair(s) of all other ports. There's active circuitry involved, and for a 10Mb hub that circuitry won't even work at 100 mbit speeds (let alone 1000 mbit).

The electrical connections are different too -- 1000 mbit uses all 4 pairs of wires in the cable, whereas 10/100 uses just 2 of the pairs.

At best, you are confusing modern twisted pair Ethernet with "old school" 10Base2 Ethernet, which used a single coaxial cable in a shared bus arrangement.

My networking class last semester said that all of the computers in universities used to all run off of the same cable. I do not believe think the specification changed in a way that disallows that since then.

just brew it! wrote:

Contingency wrote:

This is selection bias at its finest--the tablets that failed are dust by now. I'm convinced that in a thousand years, monks will be able to recover data from the half-dozen or so surviving hard drives. Perhaps they will be enlightened enough to recognize lolcats for what it is: a display of collective madness which ultimately led to the downfall of civilization.

While the magnetic patterns on the platters might survive that long, the mechanical components and lubricants used thereon will have long since degraded into uselessness. They would need to build a new drive around the existing platters; if civilization has fallen, I highly doubt they'd be able to do this.

Stone tablets have the advantage of not requiring any special equipment to read the patterns on their surface.

(Jeez, we got pretty far off-topic here, didn't we? Maybe it is time for a thread split...)

If the purpose of analysis is for data recovery, I find raw data retrieval more likely than rebuilding the drive. You just need the platters for that, and a lot of time. Wikipedia puts a scant 50 years between public electricity generation and the invention of the electron microscope. Can't say many will be pleased at the prospect of spending a month only to rediscover hello.jpg, but the incentive of recovering forgotten technologies is enough to warrant the effort.

Shining Arcanine wrote:

My networking class last semester said that all of the computers in universities used to all run off of the same cable. I do not believe think the specification changed in a way that disallows that since then.

Maybe they did: http://en.wikipedia.org/wiki/Thinnet
And maybe you should have paid more attention when physical topologies were being discussed.

Shining Arcanine wrote:

My networking class last semester said that all of the computers in universities used to all run off of the same cable. I do not believe think the specification changed in a way that disallows that since then.

That was either 10Base5 or 10Base2 Ethernet. All stations were connected to a common coax cable in a bus configuration. Both 10Base5 and 10Base2 have been obsolete for nearly 2 decades; 10BaseT (10 mbit twisted pair) Ethernet and later versions are completely different animals, and operate in a star configuration with hubs and switches.

I'm sorry, but you are way off base here.

Shining Arcanine wrote:

My networking class last semester said that all of the computers in universities used to all run off of the same cable. I do not believe think the specification changed in a way that disallows that since then.

Either they had a horribly bad design or the class instructor didn't know what he was talking about (or maybe you didn't get the context of what he was saying).

Even if it were possible, it would not be desirable. Back in the day, my University very carefully allocated the number of computers that were allowed per branch off of each switch to prevent excessive collisions on the "same wire". Granted this was the 10base-T days and not coax, but even with coax it wasn't a good idea to put too many computers on the "same wire".

It's possible the instructor was talking about a common backbone in a logical sense, but not a single cable in the physical sense.

UberGerbil wrote:

It's possible the instructor was talking about a common backbone in a logical sense, but not a single cable in the physical sense.

Agreed. That's what I meant by not understanding the context, above.

Contingency wrote:

Shining Arcanine wrote:

My networking class last semester said that all of the computers in universities used to all run off of the same cable. I do not believe think the specification changed in a way that disallows that since then.

Maybe they did: http://en.wikipedia.org/wiki/Thinnet
And maybe you should have paid more attention when physical topologies were being discussed.

Actually, my professor spoke about L1 for a few minutes and then went through to L2. He really did not delve into that kind of detail, as he was more interested in discussing L2 and L3 technologies. I seem to have misunderstood him.

I am glad someone here caught this for me because I was thinking of actually trying this.

Shining Arcanine wrote:

I am glad someone here caught this for me because I was thinking of actually trying this.

Well it's not like you would've permanently damaged anything (other than the Cat5/6 cables you were thinking of hacking up). It just wouldn't have worked.

Shining Arcanine wrote:

I do not believe think the specification changed in a way that disallows that since then.

The specification is IEEE 802.3 and then a set of suffix letters that specify the revision e.g. first version of 10GigE came in 802.3ae, 100GigE is coming in 802.3ba. Each version incorporates the earlier standards but over their own media type i.e. just because 100GigE incorporates the 10GigE standard, it doesn't mean that the physical layer will look anything like the same (it doesn't). In short, although it is the same standard, the technologies keep getting updated and are not backwards compatible apart from at a very high level in terms of the layout of the frame.