Cat5/6 100 meters
-
@scottalanmiller said in Cat5/6 100 meters:
@Dashrender and I have been discussing this offline and I think there is a missing piece that people don't remember and that is if we use a hub (aka repeater) we are in CSMA/CD networking, not Switched Ethernet which is what all of us are used to since the late 1990s. This changes everything about the network. Modern Ethernet cannot have a hub (repeater), it's not allowed (nor are they made.)
But in the old days, we had them. But there are still length limitations of the segments (but not 100m) caused by collision detection algorithms. Exactly how this works is mired in myth so it is hard to discover the exact implications. But you can't simply extend CSMA/CD indefinitely as the cabling has to support the collision detection time, in addition to withstanding the signal degradation over distances. And, of course, dealing with impedance.
The last thing I recall about stringing switches together before the latency would kill you was 5... you could have 5 segments device - switch - switch - switch - switch - device and you'd still be OK.
This likely plays into the collision/timing thing Scott and I were talking about.
Of course I'm not sure how the internet at large completely defeats this. i.e. why doesn't ethernet timeout while waiting on a response? or is there actually one by the local router while the router is waiting on the internet traffic itself?
-
@Dashrender said in Cat5/6 100 meters:
@scottalanmiller said in Cat5/6 100 meters:
@Dashrender and I have been discussing this offline and I think there is a missing piece that people don't remember and that is if we use a hub (aka repeater) we are in CSMA/CD networking, not Switched Ethernet which is what all of us are used to since the late 1990s. This changes everything about the network. Modern Ethernet cannot have a hub (repeater), it's not allowed (nor are they made.)
But in the old days, we had them. But there are still length limitations of the segments (but not 100m) caused by collision detection algorithms. Exactly how this works is mired in myth so it is hard to discover the exact implications. But you can't simply extend CSMA/CD indefinitely as the cabling has to support the collision detection time, in addition to withstanding the signal degradation over distances. And, of course, dealing with impedance.
The last thing I recall about stringing switches together before the latency would kill you was 5... you could have 5 segments device - switch - switch - switch - switch - device and you'd still be OK.
That's the 5 hop rule. Ethernet has always been like that. I never knew the reason why, but you always need a router at 5 hops.
-
@Dashrender said in Cat5/6 100 meters:
@scottalanmiller said in Cat5/6 100 meters:
@Dashrender and I have been discussing this offline and I think there is a missing piece that people don't remember and that is if we use a hub (aka repeater) we are in CSMA/CD networking, not Switched Ethernet which is what all of us are used to since the late 1990s. This changes everything about the network. Modern Ethernet cannot have a hub (repeater), it's not allowed (nor are they made.)
But in the old days, we had them. But there are still length limitations of the segments (but not 100m) caused by collision detection algorithms. Exactly how this works is mired in myth so it is hard to discover the exact implications. But you can't simply extend CSMA/CD indefinitely as the cabling has to support the collision detection time, in addition to withstanding the signal degradation over distances. And, of course, dealing with impedance.
The last thing I recall about stringing switches together before the latency would kill you was 5... you could have 5 segments device - switch - switch - switch - switch - device and you'd still be OK.
This likely plays into the collision/timing thing Scott and I were talking about.
Of course I'm not sure how the internet at large completely defeats this. i.e. why doesn't ethernet timeout while waiting on a response? or is there actually one by the local router while the router is waiting on the internet traffic itself?
Strong switches is bad, but not a killer. Switches are fast today. You could have 20 and not cause "issues." But you'd notice.
-
@travisdh1 said in Cat5/6 100 meters:
@Dashrender said in Cat5/6 100 meters:
@scottalanmiller said in Cat5/6 100 meters:
@Dashrender and I have been discussing this offline and I think there is a missing piece that people don't remember and that is if we use a hub (aka repeater) we are in CSMA/CD networking, not Switched Ethernet which is what all of us are used to since the late 1990s. This changes everything about the network. Modern Ethernet cannot have a hub (repeater), it's not allowed (nor are they made.)
But in the old days, we had them. But there are still length limitations of the segments (but not 100m) caused by collision detection algorithms. Exactly how this works is mired in myth so it is hard to discover the exact implications. But you can't simply extend CSMA/CD indefinitely as the cabling has to support the collision detection time, in addition to withstanding the signal degradation over distances. And, of course, dealing with impedance.
The last thing I recall about stringing switches together before the latency would kill you was 5... you could have 5 segments device - switch - switch - switch - switch - device and you'd still be OK.
That's the 5 hop rule. Ethernet has always been like that. I never knew the reason why, but you always need a router at 5 hops.
The hop rule is for routers, not switches. Router hops are drastically slower than switches.
-
@scottalanmiller said in Cat5/6 100 meters:
@travisdh1 said in Cat5/6 100 meters:
@Dashrender said in Cat5/6 100 meters:
@scottalanmiller said in Cat5/6 100 meters:
@Dashrender and I have been discussing this offline and I think there is a missing piece that people don't remember and that is if we use a hub (aka repeater) we are in CSMA/CD networking, not Switched Ethernet which is what all of us are used to since the late 1990s. This changes everything about the network. Modern Ethernet cannot have a hub (repeater), it's not allowed (nor are they made.)
But in the old days, we had them. But there are still length limitations of the segments (but not 100m) caused by collision detection algorithms. Exactly how this works is mired in myth so it is hard to discover the exact implications. But you can't simply extend CSMA/CD indefinitely as the cabling has to support the collision detection time, in addition to withstanding the signal degradation over distances. And, of course, dealing with impedance.
The last thing I recall about stringing switches together before the latency would kill you was 5... you could have 5 segments device - switch - switch - switch - switch - device and you'd still be OK.
That's the 5 hop rule. Ethernet has always been like that. I never knew the reason why, but you always need a router at 5 hops.
The hop rule is for routers, not switches. Router hops are drastically slower than switches.
many tracerts will show you many more than 5 hops while navigating the internet.
-
@Dashrender said in Cat5/6 100 meters:
@scottalanmiller said in Cat5/6 100 meters:
@travisdh1 said in Cat5/6 100 meters:
@Dashrender said in Cat5/6 100 meters:
@scottalanmiller said in Cat5/6 100 meters:
@Dashrender and I have been discussing this offline and I think there is a missing piece that people don't remember and that is if we use a hub (aka repeater) we are in CSMA/CD networking, not Switched Ethernet which is what all of us are used to since the late 1990s. This changes everything about the network. Modern Ethernet cannot have a hub (repeater), it's not allowed (nor are they made.)
But in the old days, we had them. But there are still length limitations of the segments (but not 100m) caused by collision detection algorithms. Exactly how this works is mired in myth so it is hard to discover the exact implications. But you can't simply extend CSMA/CD indefinitely as the cabling has to support the collision detection time, in addition to withstanding the signal degradation over distances. And, of course, dealing with impedance.
The last thing I recall about stringing switches together before the latency would kill you was 5... you could have 5 segments device - switch - switch - switch - switch - device and you'd still be OK.
That's the 5 hop rule. Ethernet has always been like that. I never knew the reason why, but you always need a router at 5 hops.
The hop rule is for routers, not switches. Router hops are drastically slower than switches.
many tracerts will show you many more than 5 hops while navigating the internet.
Yes, of course. But they only show routers, not switches. Switches are not considered a "hop". Traceroute cannot detect a switch.
-
No reason NOT to use fiber. For that distance, fiber is cheaper, faster, more reliable.
-
For 100m, singlemode fiber simplex is 12$. Transceiver are like 7€ each. You can’t beat it with copper.
-
@Francesco-Provino said in Cat5/6 100 meters:
No reason NOT to use fiber. For that distance, fiber is cheaper, faster, more reliable.
Not faster, same speed. But otherwise, yeah. Let to go wrong, just easy.
-
@scottalanmiller said in Cat5/6 100 meters:
@Francesco-Provino said in Cat5/6 100 meters:
No reason NOT to use fiber. For that distance, fiber is cheaper, faster, more reliable.
Not faster, same speed. But otherwise, yeah. Let to go wrong, just easy.
Indeed, assuming same potential bandwidth of 1gb/s, while fibre will go around 0.1c faster than copper (contrary to popular belief, it doesn't go the speed of light, because it's still bouncing around all the way down, it isn't going straight like... like from a star or whatever), it still has to be changed from light to copper and back on either end.
Over long distances as well (and I mean quite a long way), the light itself can modulate slightly and depending on all sorts of things (lossless vs lossy with various other checks) in the end most setups you aren't gaining a whole lot unless you take advantage of either the ability to go longer distances (like the OP) which does increase speed because by the time you reach 150 meters with copper vs fibre, copper adds probably 50ms or so depending on the line.
So over 150m fibre may be slightly faster than copper, but not really significantly noticeable. Given how crappy a lot of cat 5/6 cable is, it might up being pretty noticeable at such a distance.
-
I’ve used an EdgeRouter X with a POE injector to power it because there was no power at the halfway point. If you have a spare injector from one of their access points or need to install on where the switch sits in the middle the EdgeRouter X has pass through POE on the last port.
-
To clarify, the EdgeRouter X can be put into switch mode.
-
@tonyshowoff said in Cat5/6 100 meters:
So over 150m fibre may be slightly faster than copper, but not really significantly noticeable. Given how crappy a lot of cat 5/6 cable is, it might up being pretty noticeable at such a distance.
I think at that length it is actually slower because of the conversion time on either end.
-
@syko24 said in Cat5/6 100 meters:
I’ve used an EdgeRouter X with a POE injector to power it because there was no power at the halfway point. If you have a spare injector from one of their access points or need to install on where the switch sits in the middle the EdgeRouter X has pass through POE on the last port.
PoE doesn't go all that far, though.
-
@tonyshowoff said in Cat5/6 100 meters:
Indeed, assuming same potential bandwidth of 1gb/s, while fibre will go around 0.1c faster than copper (contrary to popular belief, it doesn't go the speed of light, because it's still bouncing around all the way down, it isn't going straight like... like from a star or whatever), it still has to be changed from light to copper and back on either end.
What a lot of people think is happening with copper is something weird, too. When in reality, the copper is an electro-magnetic waveguide media, just like optical cable is. Light and RF are both the same thing, just different wave lengths. So copper cabling is way more like optic fiber than people realize. Both are taking EM waves and guiding them down the length. So both use the same calculation for figuring out the speed. Both have signals that would move at the speed of light in a vacuum (c) and both are slowed by the bouncing effect and the media itself.
So copper can easily be faster than optical, because they both start from and are affected by the same things.
-
Fiber, fiber, fiber - anything else is just absolutely crazy.
You can buy pre-terminated fiber in custom lengths from fs.com. 2 pairs of single mode fiber for indoor use, 150 meter / 500ft, will be around 115 USD / 100 EUR.
If you can't attach fiber modules in the switches, two media converters will do the job flawlessly for very little money. About $35 / 30 EUR per standalone gigabit converter. That puts a gigabit 150m fiber link at a total cost of less than $200 / 200EUR.
-
@scottalanmiller no. Single strand single mode can carry 40+Gbps today.
-
@Francesco-Provino said in Cat5/6 100 meters:
@scottalanmiller no. Single strand single mode can carry 40+Gbps today.
I think you are misunderstanding the meaning of speed as we were discussing. We are talking about the speed of the signal going down the line, not the bandwidth. People often call bandwidth speed, but it is not.
Also, copper can do 40GB/s plus today, too. The bandwidth is determined by the protocol, the speed is determined by the media.
https://www.mellanox.com/related-docs/prod_cables/DS_40GbE_Passive Copper Cables.pdf
