Does patch-panels affect network speed?

Sounds like alot of money in ports and wires that won't ever be used, but thats your choice. I'd plan on extending my network out to the closet onto big workgroup switches.

But, having 2 levels of patch-panels assuming they are properly done won't do anything to the speed. You could have 10 chained together (although not desireable), and the loss on each connector won't be enough to cause enough problems to cause negotiation problems.

Ethernet is digital, it either works or doesn't (although maybe a extermely poorly done 100Mbps connection will fall back to 10Mbps because that speed is just so robust).

I've seen some networks that have had 4-6 panels in between some points, no problems.

[description snipped]

Wouldn't it be much simpler to add a user switch in the patch panel terminating all the user drops, and use a single higher-speed link between the user switch and the server room switch?

Saving the pulling of 575 cables between two closets might just pay for another switch.

Aside: In a professional setup, no crimping takes place: All cables are terminated on jacks and the patches are made with prefab cables.

Each discontinuity in the link (connectors causing untwisting, etc.) does deteriorate the signal. and the ethernet specification assumes for its length calculations one fixed strand run of up to 90 metres and two patch cables at either end, with the overall total under 100m.

For GigE you need cat5e, and for 10GigE you'll need cat6a unless you manage to keep total run length under 55m, then you can use cat6.

Your proposal will spend the distance between server room and transit panel for each run in addition to taking away some of the maximum length drop you can pull. Say you pull cat6, and you spend 20m from server patch panel to transit panel, that leaves you 20m for each drop and 5m for the patch cable from wall to user pc.

I think adding a switch in the transit panel is a much better idea.

And a few fiber runs.

I don't think money is a dramatic issue here. The building I am working on has 2 floors and we took the decision to construct 2 transits panels, with respectively 384 and 192 ports. The panels are placed into 4U hinged wall brackets. Total cost : 400$, plus the CAT6 wiring. Considering that this setup will avoid further fishing into the servers room walls (or, more dramatically, walls openings, sandblasting with prior removal of devices, etc.), this is well invested money.

As for adding a switch in the transit zone, it is something I have always been against of. IMHO, switches and servers belong to a network room where one can centrally manage security and establish logical design. Exept for distances limitations (which does not apply in my case), I am opposed to any switching activities outside the central server/network room.

That's what I tought.

In my design, a typical wire would have the following path

user desktop network card (rj45 connector) --> wall jack (rj45 connector) back side of the wall jack (110-block connector) --> back side of the transit panel (110-block connector) front side of the transit panel (rj45 connector) --> back side of the servers room panel (110-block connector) front side of the servers room panel (rj45 connector) --> switch (rj45 connector)

The transit panel adds an extra rj45 plastic connector.

Distance is not an issue in my case. My understanding is that the discontinuities affect the signal physical speed, but not enough to affect the data speed. The data speed (either 100/1000/10000) is anyway far away behind the signal physical speed.

I know.

No. Transit panels will be placed in the "route" that would have been followed by the cable. No additional distance is involved.

As I said in another reply, I am not very enthousiastic about havin switches outside the network room.

The wire pairs are specified to maintain a certain twisting rate to help keep the noise out. Untwist, and the signal gets noisier. If it gets too noisy you can switch to an encoding scheme with less strict requirements, but only able to achieve a slower data rate.

You can easily see this in similar cases, where people found that cable runs in networks built for 10BaseT ceased to function with upgrades to

100BaseTX, even though they ``future proofed'' the network by using cat5 instead of cat3. Often this is because of ``split-pair'' termination, resulting in no effective twisting. 10BaseT is robust against that, 100BaseTX isn't. GigE and 10GigE have progressively higher demands.

The speed of electrons in copper wire doesn't come into my argument.

It's your show.

All right, so the optimal design would be to have a network without any jacks in the wall nor patch panels? Jacks and panels do create "nodes" where cables get untwisted.

That's ironical I guess. The point here is to balance 2 opposite requirments : (1) a fast network and (2) secure devices management rules. In my experience, trailing switches all around a building is a pain when time comes to design, maintain and secure a network.

So long as the total length of the cable is less than the 100m limit there shouldn't be an issue.

Or several links.

If it dosn't it would either be a very expensive switch or a very cheap contractor. Nearly 600 cables is rather a lot.

snipped-for-privacy@gmail.com wrote in part:

Technically true, but practically not. Yes, from a signal transmission PoV, plugs crimped onto solid, direct plugged comp-router is best (fewest impedence/xt discontinuities), but only when in _PRISTINE_ condition.

When moved, plugged or unplugged, the plugs IDC work-hardens and the connection becomes unreliable. For reliability, best punch solid onto fixed jacks and use replaceable stranded patchcord to take the movements. The cable is part of the building, not some loopy extention cord. The electronics have been designed to handle these small discontinuities.

A bit of a false dichotomy. Better cable does not make the network any faster. Bad cable makes it slow. No symettry.

-- Robert

And even that limit is a bit soft. For fun, I had a Gig link brought up over some cable looping end to end, panel to panel on my datacenter. I think I got about 650-700 feet (ie. about double the

100m limit) and the gig link still came up and acted at Gig speeds over Cat5e cabling with zero errors recorded on the switch layer1..

But still, best to plan on the 100m limit as absolute to not have problems in the future.

No, it won't ever fall back. The speed negotiation is done (at 10 Mbps) based entirely on the capabilities of the devices at each end -- the quality of the connection between them doesn't enter into it at all.

-- Larry Jones

This sounds suspiciously like one of Dad's plots to build my character.

-- Calvin

Presumably a commercially made, stranded, patch cord. If not, you're asking for trouble.

Presumably permanently installed solid wiring.

This is a problem. The wiring from the transit point to the server room should be permanently installed solid wiring, but you should never crimp connectors onto solid wire; although it will work (provided you get plugs that are specifically designed for solid wire rather than the common plugs that are designed for stranded wire), it is not reliable over the long term. The "right" way to do this is to have *two* patch panels (or one that's twice as big) with the permanently installed wiring punched down on the backs and commercial patch cords connecting the fronts.

Again, a commercially made, stranded, patch cord unless you're looking for trouble down the road.

-- Larry Jones

This sounds suspiciously like one of Dad's plots to build my character.

-- Calvin

Except when all the newbies come around with their home-made cables with split pairs asking why their 100Mbps connections have failed to negotiate at 100Mbps and only at 10Mbps because of the degraded signal?

Thats what I was saying, not in terms of something like a modem that tests the different frequency signal to noise response and fallsback appropriately.

My mistake here ... The Transit and the Servers room panels will be connected together back-side to back-side (block 110 --> block 110). The cords will be connected to front sides panels only.

There are 3 other points I am doubtful about.

(1) My understanding is that a professional crimp should minimized the part of the wire that gets untwisted. In the actual organization of my server room, many cables expose a 1-inche distance between the block-110 punch and the 'beginning' of the wire`s untwisting. Does that increase the interference probability ?

(2) Since the distance between the transit and server room panels will be short, should I use a shielded lan cable?

(3) The Cisco documentation I am using for my work is not clear whether a CAT6 cable will support 10G-Ethernet or not. At one point, the book says that CAT6 is for Ethernet-Gig+ (dont know what + means here) and in another chapter, the author says that 10G-Ethernet is only supported by BASE-FX (optical fiber). So, does CAT6 supports 10G ethernet?

Thank u, BG

That doesn't happen. If both devices support 100 Mbps and can talk to each other at all, they'll negotiate 100 Mbps, no matter how dreadful the cable. The only way you get 10 Mbps is if at least one of the devices is manually forced to 10 Mbps or if the devices can't communicate at all (in which case 10 Mbps won't work, either).

-- Larry Jones

That's the problem with nature. Something's always stinging you or oozing mucus on you. -- Calvin

snipped-for-privacy@siemens.com wrote in part:

IOW, fallback from 100 to 10 is very unreliable/nonexistant. That matches my experience.

-- Robert

IME "lights on, nobody's home" is the more common result.

It's actually an IDC rather than a crimp.

This is not good, especially if the other end is in a similar state.