What's new

Ethernet Cable, Myth or Fact?

  • SNBForums Code of Conduct

    SNBForums is a community for everyone, no matter what their level of experience.

    Please be tolerant and patient of others, especially newcomers. We are all here to share and learn!

    The rules are simple: Be patient, be nice, be helpful or be gone!

jpn951

New Around Here
Hello All!

New to the Forum.

I have a 2280 square foot home and will be running ethernet cable (6a) for my home network.
I was told by a network engineer (acquaintance) that every room I plan to run cable to needs
to be the same length.

So, If the farthest room is 50' away from the switch, but the other rooms are closer, then every
room cannot exceed 50'? He explained that if you have different cable lengths you will have "Collisions"
and at the point your transfer rate would go to S***.

For instance, I have 2 PCs in my office, each PC is probably 15' away from the switch I intend to use,
He is saying that because the farthest ethernet run is 50', Ill have to use 50' of cable to each PC....
Is that true?

It seems a bit ridiculous to me.. Here I thought for the dual PCs in the office, I would run an ethernet cable
from the switch to the middle of both PCs then use a 5 port switch. At that point I can plug both PCs directly
in the 5 port switch.

And what about WAPs?
Since the switch has POE+ ports that I intend to use for the 3 Unifi NanoHDs which are in different locations,
(one is 40' away the second is 10' away and the other is 40' away) every WAP cable needs to be 40'!?

I have a server (File Sharing/RSS Feeds) and have different devices that will connect to the server via ethernet cable
on my network. I would run the ethernet cable to the server and all these devices that connect to it the same length,
but for everything else, why?

I may be wrong?

When it comes to running ethernet cable for a network, is it fact that ALL the cables need to be the same length
or is he just blowing smoke??

The house is 80' long and the switch will be centrally located in the home.
 
I would assume that almost everybody on this board is running a network that includes at least one network switch stuffed to the gunnels with multiple cables of unequal lengths - and do not suffer from data collisions slowing things down thanks to how switches work. Preventing data collisions is why we use switches (even our routers contain switches) rather than dumb hubs. But if you were to use a dumb hub, with equal length cables, you'd get collisions, so what you've been told doesn't make sense!
I'm trying to think where such an idea could come from, and can only think it has to do with radio transmitter systems where cabling needs to be of specific lengths in order to prevent standing waves - not an issue in modern networking systems.
 
It seems a bit ridiculous to me
It should be ridiculous as the only limitation is the longest length to sustain desired speeds.


There are some caveats to make it work for 10GE connections that don't apply to other standards but, all runs don't need to be the same length to work properly. The thing that stands out is the mention of sub 90M + 5M on each end to plug into the device.. In your case @ 50F / ~17M there's nothing special to take into consideration.

Now the real question is why spend on C6A when C5E / C6 would suffice?

I keep things tidy with my physically connected devices with 5FT C6 cables connecting things but, when I need to cable the Laptop using 5GE a 5E cable works just fine @ 20FT length.
 
I would assume that almost everybody on this board is running a network that includes at least one network switch stuffed to the gunnels with multiple cables of unequal lengths - and do not suffer from data collisions slowing things down thanks to how switches work. Preventing data collisions is why we use switches (even our routers contain switches) rather than dumb hubs. But if you were to use a dumb hub, with equal length cables, you'd get collisions, so what you've been told doesn't make sense!
I'm trying to think where such an idea could come from, and can only think it has to do with radio transmitter systems where cabling needs to be of specific lengths in order to prevent standing waves - not an issue in modern networking systems.

Is a "Dumb Hub" an unmanaged switch?
 
Currently I have all c7 cabling ran (varied different lengths) to a TP-Link TL-SG1016PE V3 16 Port Gigabit Switch. Previously, I had a Netgear unmanaged switch.

The TP-Link is a new switch I had just purchased. It's a managed switch however I have not been able to log into it and all I have done is plug everything in to it and it seems to work.. So, I have not configured this switch.

The strange thing about my network is that when I am transferring a file(s) to another PC on my network, the transfer rate won't break over 100MBs, most the time way less then that. But my network and PCs are getting 1.0GBs?

Shouldn't I be getting half, if not close to 1.0GBs transfering speed over the network using c7 cables?

I have a cable tester, tomorrow I'm going to check those cables and see if any of them are defective. If they are, what cable should I replace them with?

Currently, I can stream 4K movies from the server over the network to Nvidia Shields either through Plex or Kodi and have 0% issues. Well, every once in a while Kodi may say "The source is too slow" but it still plays flawless, the last thing I want to do is go with a replacement cable that cannot or struggle with streaming a 4K file.

Any help is appreciated.
 
Last edited:
It should be ridiculous as the only limitation is the longest length to sustain desired speeds.


There are some caveats to make it work for 10GE connections that don't apply to other standards but, all runs don't need to be the same length to work properly. The thing that stands out is the mention of sub 90M + 5M on each end to plug into the device.. In your case @ 50F / ~17M there's nothing special to take into consideration.

Now the real question is why spend on C6A when C5E / C6 would suffice?

I keep things tidy with my physically connected devices with 5FT C6 cables connecting things but, when I need to cable the Laptop using 5GE a 5E cable works just fine @ 20FT length.
Well, I tell you this, I will not let this "Network Engineer" get NEAR my network. I was planning on having him setup my network and make 1 of my PCs a Firewall but if he doesn't know his A$$ from a hole in the ground, that's not going to happen.

Thank you for your reply, I'm glad that I myself am not a "Dumb Hub" lol
 
Eventually I would like to run my server in link aggregation. The switch I purchased does comply. Would there be that much of a benefit or should I just leave it alone?
 
Sounds like he has shares in the cable supplier.
 
Is a "Dumb Hub" an unmanaged switch?
No, a hub works completely differently to a switch (managed or unmanged).

Even though the term "hub" is still sometimes incorrectly used to refer to an Ethernet switch (notably by ISP's :rolleyes:) I haven't actually seen anybody use a real hub for about 20 years. It's obsolete technology.

 
100MB/s is as good as you can expect (even if the theoretical speed may be a few percent higher) from a GbE switch.

To be clear, your network/PCs are not getting 1GB/s speeds, they are getting 1Gbps speeds. Potential 1GB/s speeds are where 10GbE switches are needed. And even then, it will depend on the file sizes you are transferring (smaller files will be slower vs. one/many large file(s) of 1GB or larger).


Do not (only) run 1 cable into each room (you'll regret it quickly). Run the length needed plus a metre or two on each end (will help when the ends may need re-terminating). Run at least two cables to the same outlet, and in specific areas, four cable runs will quickly prove helpful as the room usage changes over time and you accumulate more client devices.

I suggest that at the furthest reaches from the central location, 4 runs minimum (to those far locations). And in the central area/living room, I would put up to 10 runs at various locations around the room to give you the flexibility to decorate the room/furniture as needed in the future without impacting the speed/efficiency of the network either.
 
Last edited:
Currently I have all c7 cabling ran (varied different lengths) to a TP-Link TL-SG1016PE V3 16 Port Gigabit Switch. Previously, I had a Netgear unmanaged switch.

The TP-Link is a new switch I had just purchased. It's a managed switch however I have not been able to log into it and all I have done is plug everything in to it and it seems to work.. So, I have not configured this switch.

The strange thing about my network is that when I am transferring a file(s) to another PC on my network, the transfer rate won't break over 100MBs, most the time way less then that. But my network and PCs are getting 1.0GBs?

Shouldn't I be getting half, if not close to 1.0GBs transfering speed over the network using c7 cables?

I have a cable tester, tomorrow I'm going to check those cables and see if any of them are defective. If they are, what cable should I replace them with?

Currently, I can stream 4K movies from the server over the network to Nvidia Shields either through Plex or Kodi and have 0% issues. Well, every once in a while Kodi may say "The source is too slow" but it still plays flawless, the last thing I want to do is go with a replacement cable that cannot or struggle with streaming a 4K file.

Any help is appreciated.
Ethernet ports or nominally rated in Gigabits (Gb/s) or Megabits (Mb/s) per second. A 1 Gbit/s port can handle 1000 bits/8bits = ~125 MBytes/s link rate. So if you are getting sustained transfers of 100 MB/s you at about the maximum a 1 Gb/s port will handle. There is overhead in the protocols. Most switches will handle full line sync speed as they use specialized chip to do so. A 10 Gb/s switch can get up to 1.25 GB/s link rate and probably slightly less than that transfer between two ports on the switch. So a 1 Gb/s rated switch should have a total switching capacity for a 8 port device of 8*2*1 Gb/s = 16 Gb/s or higher to allow non-blocking flow of packets.

Usually, the end devices are going the be the limiting factor in transfers (PC, TV, NAS, IOTs, router, ISP drop), not the switch or the cabling (short of poor termination).

Unless your cable tester is specific to CAT7 and tests sync speeds (not your cheap type tester) + shield ground, continuity, etc. the only thing you will find is either a miswired termination or improperly terminated shield. Not enough to tell you if there is an issue with the link rate. An improperly terminated shield ground will cause it to pick up RF interference and thus slow the link down. Even if the cable termination is correct, the ground has to be correctly connected to the earth. That is one reason, most do not use CAT7. Cat5E and Cat6 are usually good enough for 10 Gb/s in a house due to the short runs.
 
Eventually I would like to run my server in link aggregation. The switch I purchased does comply. Would there be that much of a benefit or should I just leave it alone?
LAG only provides real benefit for multiple users accessing the NAS at a single time. So it would be good for a server facing the internet where thousands of users are accessing it. LAG can be used to increase data transfer between two devices that both support the higher rate - such as two 1Gb/s ports on a router LAN switch and two on another LAN switch that both devices support LAG on. All that does is shift the bottleneck elsewhere in the network - perhaps the router LAN to WAN connection or the LAN switch connection to a PC over a single Gb/s port.
 
I'm just blown away how someone can be SO wrong about these things and consider himself a "Network Engineer"

All the information I was told concerning networking from this guy was smoke and mirrors. I do appreciate all the info.

So it is normal to be getting anywhere from 25MBps - 100MBps when transfering a file(s) over the network?
These files I'm transfering are between 1GB-250GB

From what I was told, I should be getting 5-600MBps when transfering large files. So that is completely false?

If so, sounds like my network is running exactly how it should.
 
LAG only provides real benefit for multiple users accessing the NAS at a single time. So it would be good for a server facing the internet where thousands of users are accessing it. LAG can be used to increase data transfer between two devices that both support the higher rate - such as two 1Gb/s ports on a router LAN switch and two on another LAN switch that both devices support LAG on. All that does is shift the bottleneck elsewhere in the network - perhaps the router LAN to WAN connection or the LAN switch connection to a PC over a single Gb/s port.
Then I'll leave it alone. I'm not willing to spend the money for link aggregation if it will not benefit me or the network.
 
@jpn951

Let's make things easy....

1652274718280.png


I keep this on hand because I don't like doing the math converting bits to bytes. Factors of 8 when converting them makes my head hurt.

As to a LAG... there's different modes and not every device supports all of them.

1652275065539.png


For instance my cable modem supports LACP for 2 ports to be bundled together. For this I have to use mode 4 / LACP and that's the only option with my modem that will bring the links up and pass traffic. Trying other options for different traffic patterns won't work because the modem isn't smart enough to do that.

Basically in Linux / routers / switches it produces an interface file that is applied to make things work.
Code:
sudo cat /etc/network/interfaces

# interfaces(5) file used by ifup(8) and ifdown(8)
auto lo
iface lo inet loopback

auto lo:50
iface lo:50 inet static
address 192.168.0.50
netmask 255.255.255.255
network 192.168.0.0

##Outside WAN##
auto enp10s0
iface enp10s0 inet manual
bond-master bo0

auto enp11s0
iface enp11s0 inet manual
bond-master bo0

#auto enp12s0
#iface enp12s0 inet manual
#bond-master bo0


auto enp2s0 enp7s0 enp8s0 enp10s0 enp11s0 wlp3s0
allow-hotplug enp2s0 enp7s0 enp8s0 enp10s0 enp11s0 wlp3s0

iface enp2s0 inet manual
iface enp7s0 inet manual
iface enp8s0 inet manual
iface enp10s0 inet manual
iface enp11s0 inet manual
iface wlp3s0 inet manual

auto br0
iface br0 inet manual
iface br0 inet static
address 192.168.0.1
bridge_ports enp2s0 enp7s0 enp8s0 wlp3s0

auto br0
iface br0 inet manual
iface br0 inet static
address 192.168.0.2

auto bo0
iface bo0 inet dhcp
        bond-mode 4
        bond-miimon 100
        bond-lacp-rate 1
        bond-slaves enp10s0 enp11s0

With this I'm putting 10/11 into the bonding configuration to double the speed by combining the 2 interfaces.

On the LAN side I've done the same with Bridging to keep configurations simple by bundling all LAN traffic under one pseudo interface - br0.

This makes managing things a bit easier than probing each interface for the stats / monitoring. For firewall management / iptables it makes it easy to insert the br0/bo0 interfaces instead of copying 2-5 interfaces into each line of rules. In this case I made it simpler by making sets of rules that get referenced back to a single "group" for targeting traffic.

1652275734470.png


If you're planning on using pfsense or just plain Linux as a router/switch/firewall/AP it's challenging but not impossible depending on your skill level. You can make it as simple as you want or complicated and slow. The more rules you add the slow things tend to pass through it. Ordering the rules properly from top >> down requires less processing time for the packets when they hit the "router". I.E. putting all of the normal LAN traffic at the top of the list means it flies through and gets ultimate speeds.

When you get to the untrusted traffic such as the WAN then you can get a bit more specific. For me I like to keep tabs on things to a point where it's easy to spot when something might be wrong but not impact performance. By default you want to BLOCK everything and only allow things that are needed. When you first setup things though the FW is in default permit mode figuring people might not be local to the box to configure it and thus would be locked out from reaching it / all traffic would be stopped.

Anyway... this is getting beyond the scope of things here but, going DIY has some perks to keeping your traffic / data secure.
 
From what I was told, I should be getting 5-600MBps when transfering large files. So that is completely false?
Yes that's false when talking about Ethernet. WiFi connected devices are an entirely different subject.

As previously mentioned the line rate of gigabit Ethernet is 125 MB/s ignoring overheads. Real world throughput (as shown by Windows file copy for example) should be ~112 MB/s. If it's less than this then the bottleneck should be on the sending or receiving device (e.g. it's disk write speed) not the switch or the cables.
 
Last edited:
So it is normal to be getting anywhere from 25MBps - 100MBps when transfering a file(s) over the network?
These files I'm transfering are between 1GB-250GB

From what I was told, I should be getting 5-600MBps when transfering large files. So that is completely false?
5-600 mbits/s over 1gbps is a little on the realistic side depending on both sides of the link. The max you could see is ~960mbps w/ overhead which tends to be 90+MB/s.

With my setup over wifi I can hit 1.3-1.5gbps for a couple of reasons.
- I'm using an AP not a router off the shelf from the local store down the road
- I have a 2.5gbps backhaul from the AP to the "router"
- I have the disks setup in the "router" in a Raid 10 configuration that maxes out at 400MB/s+

sudo dd if=/dev/md0 of=/dev/null bs=4096k iflag=direct status=progress

6262095872 bytes (6.3 GB, 5.8 GiB) copied, 13 s, 482 MB/s
1506+0 records in
1505+0 records out
6312427520 bytes (6.3 GB, 5.9 GiB) copied, 13.1154 s, 481 MB/s

So, it depends on the network but, also the sped of the disks / enclosure you're using. Most disks these days should be able to hit 200MB/s as long as they're not SMR. That's a whole different discussion though. But, 200MB/s already exceeds your 1gbps connection so, you have room to grow if you choose to. The other side of it being the NAS can be a bottleneck as well by either the NIC being 1gbps or a low end CPU in the NAS that can't handle the speed of the disks being used.

When I hardwire into my box to do a bulk transfer I'm using 5gbps ethernet which from the green post above satisfices max speeds of the storage with room for overhead. I planned this based on the testing before shifting parts of the equation around. Testing and planning make for a better experience. It also save you some $$$ on swapping out equipment based on changing needs.

10gbps / 10GE Ethernet might be the way to go if you plan on nerding out and moving a ton of data the majority of the time. Localizing 10GE to a particular area of the network makes more sense to save on costs though. If you're planning on streaming to devices that's typically going to max out at ~100mbps / 10MB/s and won't require the faster speeds. So, the smaller segment like the NAS / primary PC's needing the higher BW could be on their own Switch and then connect to a slower switch for the rest of the network.

Cisco-3-Layered-Hierarchical-Model.gif


So, Core in this instance would be where your ISP comes into the network.
Distribution is where you divvy up the bandwidth -- where the 10GE / 1GE switches would interconnect
Access is just all of the devices

1652277370331.jpeg


So, here I would put the 10GE switch where it says switch and then use one of the ports on there to add a 8+ port 1GE switch for slower devices. Since you mentioned POE you have a couple of options since the number of mentioned devices is limited you could use POE injectors or you could do a POE switch / 4-port keeps the costs down. Then just connect the POE switch into the 10GE switch.

For my 5GE setup w/ the 2.5GE AP it was cheaper to just do a power injector or just use the AC adapter rather than pick up a switch just for that.

Anyway.. there are plenty of options to make things come together. For the cabling 2 drops to each outlet is a sufficient amount of cable. It gets pricey though to have redundancy depending on which level of cable you choose. The nice thing about running cable is if one breaks it's easier to attach the new cable to the end of the broken one and pull it from the other end instead of having to fish it through the wall like the original install. A little bit of tape and patience by the installer makes it cheaper / easier to replace instead of paying upfront for dual cables.
 
While it does make sense to have more than one cable to certain key locations there are other possibilities for other locations.

1. A simple unmanaged switch can be installed at any location giving you the option of having multiple Ethernet connected devices.
2. Using a slightly more expensive smart switches gives you the option of running multiple VLANs over a single cable using 802.1Q VLANS.
 
@CaptainSTX

To expand on that brings up 2.5 / 5GE switches if you're playing with higher end AP or NAS speeds.

The network will only go as fast as the slowest port in the path. If one room has high bandwidth needs looking into this as an option would be a good idea. 2.5 switches are down to ~$100. 5ge switches tend to be 10ge prices which are still quite high.
 

Latest threads

Sign Up For SNBForums Daily Digest

Get an update of what's new every day delivered to your mailbox. Sign up here!
Top