Does patch-panels affect network speed?

Have a question or want to start a discussion? Post it! No Registration Necessary.  Now with pictures!

Threaded View
Hello,

I am in the process of designing a mid-size network (~300 connections)
and I play with the idea of having 2 levels of patch-panels.

My physical design is based on the idea of not having to modify the
servers room (walls, roof, floor, concretes, etc.) if new cables are
eventualy required in the building.  Between the server rooms and the
building, we would therefore add "transit" patch-panels, with ease of
access.  The connections between the servers room panels and the
"transit" panels will contain 576 wires to cover the expected cabling
growth.  All users devices would be connected to transit panels.

Concretely speaking, it means that a desktop user will reach its
switch trough 2 patch panels exept of one.  Will that affect
performances?  You can take for granted that crimping will be done
professionally and that all panels are cat 6.

Thank you,
B

Re: Does patch-panels affect network speed?
brunogirard1970@gmail.com writes:
Quoted text here. Click to load it




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.








Re: Does patch-panels affect network speed?
Quoted text here. Click to load it

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.

Quoted text here. Click to load it

That's what I tought.

Re: Does patch-panels affect network speed?
Quoted text here. Click to load it

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

Re: Does patch-panels affect network speed?
lawrence.jones@siemens.com writes:
Quoted text here. Click to load it



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.




Re: Does patch-panels affect network speed?
Quoted text here. Click to load it

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

Re: Does patch-panels affect network speed?
lawrence.jones@siemens.com wrote in part:
Quoted text here. Click to load it

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

-- Robert


Re: Does patch-panels affect network speed?
Quoted text here. Click to load it


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

--
Mark Evans
St. Peter's CofE Aided School
Phone: +44 1392 204764 X241
Fax: +44 1392 204763

Re: Does patch-panels affect network speed?
Quoted text here. Click to load it

Agreed.  Lights mean signal and activity.  Not _clear_ signal
and good frames, at least for 10/100 on most equipment.

-- Robert

Quoted text here. Click to load it

Re: Does patch-panels affect network speed?
Quoted text here. Click to load it



I've gone into a few situations where exactly that did happen.
Split pairs, and the interface only links up at 10Mbps. Fix the split
pairs, and it works at 100Mbps with no gear (or config) change.

And there have been many posts over the years to this group with
people coming here with that exact same problem.

I'm not saying that it happens all the time, but it does happen from
time to time.

Re: Does patch-panels affect network speed?


Quoted text here. Click to load it

Only through a freak of nature.  For the umpteenth time, the
autonegotiation protocol is always done at 10Mbps, it does not have any
kind of cable test (other than that the two ends have to be able to
communicate in order to negotiate), and it does not have any kind of
fallback other than, if the two ends can't communicate, the assumption
is that the other end doesn't support autonegotiation/100Mbps so then
the link falls back to 10Mbps.  So, the only way to get a fallback to
10Mbps is if the link is so bad that autonegotiation fails, which means
the link doesn't really work at 10Mbps, either.  If the split pairs only
affect one direction and the majority of the data transmission is in the
other direction, then it's possible that you won't notice how bad the
bad direction is unless you specifically look for it.

-- Larry Jones

The hardest part for us avant-garde post-modern artists is
deciding whether or not to embrace commercialism.  -- Calvin

Re: Does patch-panels affect network speed?
On Fri, 20 Jun 2008 13:59:28 -0700 (PDT),
Quoted text here. Click to load it
[description snipped]
Quoted text here. Click to load it

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.


Quoted text here. Click to load it

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.


--
  j p d (at) d s b (dot) t u d e l f t (dot) n l .
  This message was originally posted on Usenet in plain text.
  Any other representation, additions, or changes do not have my
  consent and may be a violation of international copyright law.

Re: Does patch-panels affect network speed?
jpd wrote:
Quoted text here. Click to load it
And a few fiber runs.

Re: Does patch-panels affect network speed?
Quoted text here. Click to load it
girard1...@gmail.com> wrote:
Quoted text here. Click to load it

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.

Quoted text here. Click to load it

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.

Quoted text here. Click to load it

I know.


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

Quoted text here. Click to load it

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

Re: Does patch-panels affect network speed?
On Sat, 21 Jun 2008 07:05:38 -0700 (PDT),
Quoted text here. Click to load it

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.


Quoted text here. Click to load it

It's your show.


--
  j p d (at) d s b (dot) t u d e l f t (dot) n l .
  This message was originally posted on Usenet in plain text.
  Any other representation, additions, or changes do not have my
  consent and may be a violation of international copyright law.

Re: Does patch-panels affect network speed?
Quoted text here. Click to load it

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.


Quoted text here. Click to load it

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.

Re: Does patch-panels affect network speed?
brunogirard1970@gmail.com wrote in part:
Quoted text here. Click to load it

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.


Quoted text here. Click to load it

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



-- Robert



Re: Does patch-panels affect network speed?
brunogirard1970@gmail.com wrote:
Quoted text here. Click to load it

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

Quoted text here. Click to load it

Presumably permanently installed solid wiring.

Quoted text here. Click to load it

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.

Quoted text here. Click to load it

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

Re: Does patch-panels affect network speed?
On 26 juin, 21:46, lawrence.jo...@siemens.com wrote:
Quoted text here. Click to load it

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

Re: Does patch-panels affect network speed?
brunogirard1970@gmail.com wrote:

Quoted text here. Click to load it

It's actually an IDC rather than a crimp.

Quoted text here. Click to load it

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

--
Mark Evans
St. Peter's CofE Aided School
Phone: +44 1392 204764 X241
Fax: +44 1392 204763

Site Timeline