Gibabit Network

[motorbiking]

I have the floor up at the moment and I have the opportunity to install some networking for potential future use. I've ordered a real 100m cat6 and some dual and quad rj45 fact plates.

In terms of design - should I just be thinking of bringing all cables back to single point and then using a switch if I need to or is there benefit in connecting room to room as well.

Some of my devices have dual gigabit ports, so I was going to work on two lines to each room.

Its slightly academic at the moment as I only have a 4 port gbit switch with two spare ports, but for £50 it seemed worth plumbing a few cables at wall plates in.

 

[TheVictorian]

I'd go for two to each room all radiating from a central point, if that central point isnt where your router is then I would run another two to that point (one is all you need but it's nice to have a little redundancy.)

People will tell you there is no point with wireless etc etc but personally I have 2 x network points in every room and find it very useful.

 

[motorbiking]

As a tip for anyone else looking at this thread - buy a punchdown tool. I've just ordered one off ebay having thought i could get by without using a small screwdriver. Worth a couple of quid in time saved I think

 

[nflowers]

Yes, TheVictorian is correct, wire everything back to a single point (assuming the cable runs are < 100m!). Motorbiking, agreed, a proper punch down tool is essential, as well as a simple tester e.g. http://www.maplin.co.uk/p/lan-network-cable-tester-n59by or http://www.maplin.co.uk/p/rj45-and-rj11-network-cable-tester-n21cl (both discontinued but less than a fiver if you can find one nearby!)

 

[motorbiking]

Just bought one from ebay for <£3

 

[Whitling2k]

I got a punchdown tool from SF, cheapish, but not cheapest - if yoyu have more than a few to do, a cheap one if false economy.

Radial from a central point is the way to go

I would double up all cable runs , but only install one; helps future proof. Maybe even put a run of tough nylon string in too - to help pull further wires in/out the future

Also - I experimented with both Cat5e and Cat6 - and contrary to polular belief, I get much better performance with cat5e.

My reasoning was that cat6 could be faster, if installed in optimum conditions, but when installed in a real-life 1970s bungalow, sharing voids with mains cables, heating pipes etc it performed quite poor. The cat5e was less sensitive to external things, and so could take a but more rough and tumble... And was less pinikity with installation, so in the real-world near enough saturates my GB ethernet when reading off my SSD or USB3

 

[Monkeh]

Also - I experimented with both Cat5e and Cat6 - and contrary to polular belief, I get much better performance with cat5e.

My reasoning was that cat6 could be faster, if installed in optimum conditions, but when installed in a real-life 1970s bungalow, sharing voids with mains cables, heating pipes etc it performed quite poor. The cat5e was less sensitive to external things, and so could take a but more rough and tumble... And was less pinikity with installation, so in the real-world near enough saturates my GB ethernet when reading off my SSD or USB3

Cat6 is in no way more sensitive than 5e - if you had problems, it's likely due to bad cable (which would never meet cat6 specs!). It also certainly won't be faster than 5e - below 100m, in domestic conditions, both will sustain 1Gbit without any errors. Any issues are caused by installation error or non-compliant cable.

'Popular belief', being belief, has absolutely nothing to do with reality.

 

[Whitling2k]

I try not to reply to your types of post, but this one annoyed me. Quite simply, you are wrong and are being falsely presumptive. Why would I buy cheap cable to benchmark?

The termination and routing restrictions required to meet the cat6 spec mean it takes a bit longer and more thought when routing, and a higher chance of messing up; in my case for zero benefit and additional cost. Check Wikipedia for an overview, or cat6 manufactures websites.

I didn't say cat6 would be faster - I said the opposite. There are many threads on the internet where people think their cat6 upgrade will make their internet/network go faster than their old cables - I didn't believe this to be true so benchmarked it.

I ran 25m of 6 and 5e between my living room and under stair cupboard. It went under the floor, through tight gaps, and round sharp corners, and got very similar results, albeit slightly quicker/less latency with 5e.

For retrofitting my old house, I couldn't meet all cat6 requirements - I had some tight bends for example, I need it to run adjacent to mains AC and copper pipes. And I wasn't buying quality shielded cat6 when quality un-shielded cat5e met my needs.

I can transfer 1Gb/s with 5e with ease; I installed it quicker and spent less money doing so.

My above post was based on my experience and a bit of research with the simple intention of stopping people wasting money on expensive cat6 cable, expecting better performance that most retrofitters just will not see.

 

[Monkeh]

*sigh*

I don't know what cable you bought: I do know most of what's on the market is crap. Almost universally true if it's not a full box from a reputable manufacturer.

If you install cat6 to cat5e specs, assuming you don't do damage by stretching it excessively due to attempting to take sharp bends with the spine (if it has it), it will perform just fine for GigE. No noticable difference, because there will be no errors. If there are errors, there are faults: That is installation or plain bad cable, not cable type.
Shielding wouldn't help in the least: External noise is not, and almost never will be a concern in domestic environment. Hell, it's mostly used to prevent alien crosstalk in large (HUNDREDS of cables) bundles, not resist external noise from other sources. Stretched conductors, split pairs (both in the run and at termination: NEVER UNTWIST), and in the case of meeting cat6 specs, crushed/kinked cable affecting pair spacing (again, crosstalk) are the enemies, and it's just as easy to mess up cat5e. Seeing as both 5e and 6 well exceed the requirements for GigE, you have to make a REAL mess of it to prevent normal operation.

I agree, cat5e is perfectly adequate for anyone without plans to run 10GigE, but that doesn't mean it can perform better. Any actual differences will be down to poor cable or poor installation (above and beyond not installing to meet cat6 specs). A few pings and 'tests' with Windows file transfers (the LEAST accurate benchmark tool in the universe) don't reveal much.

As far as installation goes, the only practical difference is the potential (it is NOT a cat6 requirement, but makes it vastly EASIER to install to spec) of having a spine making the cable stiffer. If one avoids forcing it, this greatly helps avoid the damage both types will suffer, but it does make it more awkward if you're not used to it.

 

[Whitling2k]

Ooops - there we go with more assumptions, must remember to hail to the all knowing, only competent IT user the world has ever known:

"Windows file transfers (the LEAST accurate benchmark tool in the universe) don't reveal much"

I must be a right thicko benchmarking a network with windows file transfer... Oh wait, I didn't... you just assumed again.


seeeeeya!

 

[Monkeh]

Ooops - there we go with more assumptions, must remember to hail to the all knowing, only competent IT user the world has ever known:

"Windows file transfers (the LEAST accurate benchmark tool in the universe) don't reveal much"

I must be a right thicko benchmarking a network with windows file transfer... Oh wait, I didn't... you just assumed again.


seeeeeya!

All one can make without information is assumptions.

I would be fascinated to see a detailed analysis of the cable you installed, photos of the installation, crosstalk and return loss figures, pair skew, bit error rates indicating failed transmissions.. Anything at all to indicate you actually made cat6 unable to carry GigE, or to clarify why the cable you had was unable to.

I somehow doubt anything of the sort will be forthcoming, as you are now of the belief that cat5e is superior for your use-case (I again admit spineless cable is easier to install for most) and proof is never required for belief.

 

[Whitling2k]

Go on then, since you asked:
I used one of these - not this one but one like it from ebay - https://www.amazon.co.uk/Ethernet-Internet-Broadband-Connection-Capability-White/dp/B004BVIIAS
I 'borrowed' some cat6 cable from a mate in IT - I'm at a 15,000+ user organisation, so went on the assumption that it wouldn't be cheap cr@p, my cat5e was a screwfix bundle.
I played about firstly with Windows file transfer, then command prompts ping, then downloaded trials of Paessler's software and Totusoft's freeware packet loss and speed tests as Windows was proving quite unreliable
It was a real-world comparison, so no lab set-up. I lifted the boards behind my TV, used a combination of plastic pipe, string and bits of wood to pull the two wires about 15m through my sub-floor, through under-floor heating-pipe gaps in brick walls, through and up a stud-wall and into my ''Hub 0". I added a few m coiled up to the end to account for the more complex routing to bedrooms etc and pushed down into cat6 and cat5e (respectively) sockets.
I connected each end into my 10 port GB switch and the other into my gb laptop
then plugged my switch into my GB desktop
then played, experimented, switched, coiled, kinked etc etc. bench marking 6, 5, 6, 5 for each test
averaged results - and found that cat6e was as good if not a bit worse - from memory, my 5e performed something like 92% potential and my 6e something like 90%

I didn't photograph or save my results - I was doing this for myself, not to share, and made a mental note of my findings. It was purely to decide if it was worth paying the extra for 2x cat6 to all 5 bedrooms, dining, study, living & kitchen whilst I'm tearing back all walls, floors and ceilings. Which incidentally needs 300m cable all-in-all - i.e. doubling/tripling the cost would mean a hell of an investment.

And given the overall & average slight dip in latency and throughput, decided not to pay extra, but to run a strong piece of nylon alongside each pair of cables in case I decided to change things in the future.

And on here - I merely demonstrated my findings, that for a retro-fit, gb network, there is no benefit to paying the additional cost and working to a higher specification.

I agree with your point about assumptions, but you continuously assuming I'm being a thicko is a bit unjust.

I know what you are getting at, and more often than not you would be right. But you picked a battle with the wrong guy today. I'm a scientist, I work in science, I have studied science, my hobbies include IT, networking, computing [as well as less anti-sociable cycling, running, hiking & DIY]. My job is to test theories, see how academic science & research will work in a production and operational environment, test new ideas and operationalise experimental ideas into every-day tasks across environmental modelling, asset management, natural/anthopogenic interfaces, I know what I'm doing; and I know how good is good enough... It's what I do.

My reason for not spending £silly on cable is very just - and if I can stop other people make the same decision, I would consider it a success.

 

[Monkeh]

Go on then, since you asked:
I used one of these - not this one but one like it from ebay - https://www.amazon.co.uk/Ethernet-Internet-Broadband-Connection-Capability-White/dp/B004BVIIAS

Okay, a continuity tester. Good for verifying you've connected the pairs correctly, otherwise no value.

I 'borrowed' some cat6 cable from a mate in IT - I'm at a 15,000+ user organisation, so went on the assumption that it wouldn't be cheap cr@p, my cat5e was a screwfix bundle.

Okay, probably okay cable. At the very least, not the worst going (I have some 'cat5e' here, proudly says so on the jacket! Inside are 8 individual wires, not even the right colours, let alone pairs. It even works, just don't try and use a radio next to it.. or switch the lights on or off. Or the AC.).

I played about firstly with Windows file transfer, then command prompts ping, then downloaded trials of Paessler's software and Totusoft's freeware packet loss and speed tests as Windows was proving quite unreliable

Okay. As you found out, the file transfers are pretty useless. As for ping, on Windows it only reports '<1ms' (try and get copper anywhere near 1ms.. you need a couple hundred kilometres of fibre to do that. 5e and 6 should both be in the ballpark of 5ns/m, 0.5us for the entire allowed run), so not much use. Ping response time typically varies by ~100us on a good day to a non-Windows OS, so not much to be found there. PRTG is a bit more handy.

It was a real-world comparison, so no lab set-up. I lifted the boards behind my TV, used a combination of plastic pipe, string and bits of wood to pull the two wires about 15m through my sub-floor, through under-floor heating-pipe gaps in brick walls, through and up a stud-wall and into my ''Hub 0". I added a few m coiled up to the end to account for the more complex routing to bedrooms etc and pushed down into cat6 and cat5e (respectively) sockets.
I connected each end into my 10 port GB switch and the other into my gb laptop
then plugged my switch into my GB desktop
then played, experimented, switched, coiled, kinked etc etc. bench marking 6, 5, 6, 5 for each test
averaged results - and found that cat6e was as good if not a bit worse - from memory, my 5e performed something like 92% potential and my 6e something like 90%

Okay, so a 2% difference in apparent performance between two (assumedly, correct me if I'm wrong) Windows hosts. That should disappear over a larger test set, especially if the hardware is mixed up a bit.

And given the overall & average slight dip in latency and throughput, decided not to pay extra, but to run a strong piece of nylon alongside each pair of cables in case I decided to change things in the future.

How much of a dip in latency did you see? There really, really shouldn't be any due to the cable unless you've got a horde of bit errors. I haven't had a RX error on my workstation in the last 72 hours, over 158 million packets. According to the switch, no errors in the last 4 billion packets. Not entirely sure I believe that, but it could be right.

Bear in mind I see ~150us deviation over a moderate (4 minute) period between two well behaved hosts, a Windows host tends to be worse (both average response time and deviation). Add some load and it can get bad. I'm typing on this machine, have some video playing (over the network) on the other screen, and video encoding on the go (network source). 2.6ms worst-case response, but the average over that period is only around 250us. Shorter periods obviously can be badly skewed by a few slow responses like that.

And on here - I merely demonstrated my findings, that for a retro-fit, gb network, there is no benefit to paying the additional cost and working to a higher specification.

And I agree, there's no need for cat6 for most domestic installs. As for higher specification, I terminate both the same (and more than well enough to meet cat6 specs), you just have to be careful with the bend radius if you really do want to get there. If you don't bother terminating as neatly it'll still perform as well as 5e terminated the same way.

I agree with your point about assumptions, but you continuously assuming I'm being a thicko is a bit unjust.

I apologise if I came across that way, it was not intended. Truce?

I'm a scientist, I work in science, I have studied science

Then you'll take no offence when I say 'sample size of 1'.

 

[motorbiking]

the cable I used had a little plastic X protector running through the middle and cost about 5 quid more than 5e. None of the devices installed will require anything like 128MB/s but given the cost of 6 vs 5e. I thought I'd spend the extra.

I've installed 6 runs doubled up with double sockets everything went fine except one socket failed to light the lights in sequence. I've stared and stared at the connections at cannot see what I've done wrong. it's identical to the cable next to it. I may have to resistance check each pair. I'm not bothered as I have plenty of redundancy