Cat5e noise/ random failures

[kingofkya]

I was wondering if anyone has experience with EDIT: cat5e or better and gigabit Ethernet. I have random issues that i am wondering might be cross talk. I have about 40 racks with a nearly identical setup. Most are just fine no complaints however there is one or two racks with significant traffic 1-10gbit range(Uplinks are 10gbit fiber unrelated) and they seam to have the most issues, but i am not sure if its just noticed more or if less traffic = less nosie = more reliable connection.

2 gigabit switches with 42ish utp cat5e or better cables each going down the side of the rack like show in attached pictures. The issue is have is randomly cables "fail" 4-5 a month seams unreasonable to me when there not really moved around. These are band new cat5 from monoprice. And i can then test the bad cable while installed and it passes all tests (second pic). But yet replacement the cat5 nearly always fixes it. The fluke however does say its a high noise and it taking time to avg noise.(or something like that)


So my question is i wonder if shielded cat5 would fix the random failures and if thats the "right solution", thoughts?

Thanks

http://kneale.us/randomFiles/forum1.jpg
http://kneale.us/randomFiles/forum2.jpg

[mariush]

Well, first thing you should obviously do is to change the type of network cable - if you have 4-5 failing each month, it would make sense to switch to another cable manufacturer or maybe choose a slightly more expensive model.
I think the cheaper monoprice cables may be using CCA (copper clad aluminum) strands inside, which may cause issues at the contact with the pins of the rj45 after a lot of heat-cold cycles

My suggestion would be to go with some Cat6 or Cat6a cables - overkill for gigabit ethernet but typically they should be better built.

It's not clear from the picture, but I wonder if it would be possible to install the switch in the middle of the servers instead of the top, so that you'd be able to use smaller patch coords (6 ft patch cords are usually $1-2 cheaper than 10-14ft cables) ... not to mention it should make more room for airflow and fewer cables in a bundle.

I'd suggest checking out this: http://www.cablesondemand.com/pcategory/150/category/MPCAT6/URvars/Catalog/Library/InfoManage/CATEGORY_6_%28CAT6%29_CABLES.htm  - it's page for Cat6 , Cat6a are more expensive and not required in your situation.

[RGB255_0_0]

I'd be looking at the switch gear if it's consistently happening. Unless you're running your cables alongside power cables you shouldn't need STP cables, assuming you run in steel conduits in high EMF areas.

Still, my thoughts would be the switches first.

[vsboost]

I tend to keep my cable bundles well away from power.

[suicidaleggroll]

There's a lot of conflicting information in your post.

You said 1-10 Gbit range, then said it's only gigabit.  You said they're cat5e cables, then continued calling them cat5.

So which is it, 1 or 10 Gbit, and cat5 or cat5e?

[blacksheeplogic]

The cables don't look well supported esp being that they are just terminated and connected to the switches with a RJ45 plug.

[xygor]

Is PoE involved?

[kingofkya]

There is also fiber 10g uplink on the switch thous are fine i was just making the point that they get more traffic in aggregate  over the copper 1gig that other racks.

I have also tried cat6 but just a few strands, the techs don't like the stuff due to fact its more rigid. Remember we talking about 10ft here. same  issue with other brands Belkin, cables2go etx, the monoprice one are the ones we use by default though because there nice and flexible.

Switches have already been replaced once brocade icx-6450 switches.


The monoprice cables are copper or they would not pass that fluke tester length would be way off at a min. And i have cut them open and checked by melting scraping the wire

No POE all just gig Ethernet.

[kingofkya]

I'd suggest checking out this: http://www.cablesondemand.com/pcategory/150/category/MPCAT6/URvars/Catalog/Library/InfoManage/CATEGORY_6_%28CAT6%29_CABLES.htm  - it's page for Cat6 , Cat6a are more expensive and not required in your situation.

Interesting thous molded connectors look identical to the monoprice ones.. I think i will get a set of thous and give them a shot

[Monkeh]

Cat6 doesn't have to be stiff, not all examples have the spine.

I'd suggest trying S/UTP or F/UTP.

As above, your cables aren't well supported. The weight of a short length of one cable doesn't do much harm, but you have them bundled up and really dragging on some ports.

[kingofkya]

The cables don't look well supported esp being that they are just terminated and connected to the switches with a RJ45 plug.

There mounted the the post on the back side to the rack see the black straps.

Then cable goes though the rack and though an organizer. Minimal stress on the rj45
http://kneale.us/randomFiles/forum3.jpg


But yeah i get your point you shoudl have seen the tangled mess that was here when I started, there was literally knots in the cat5.(thouse cables were thrown out a long time ago)

[Monkeh]

The cables don't look well supported esp being that they are just terminated and connected to the switches with a RJ45 plug.

There mounted the the post on the back side to the rack see the black straps.

Then cable goes though the rack and though an organizer. Minimal stress on the rj45
http://kneale.us/randomFiles/forum3.jpg


But yeah i get your point you shoudl have seen the tangled mess that was here when I started, there was literally knots in the cat5.(thouse cables were thrown out a long time ago)

Not much stress there, but look how the weight of 4-8 cables tied together for a couple feet is dragging on the other end..

[Tim F]

Yep, this one doesn't look too healthy:

Adding another tie for bundles of 5-10 cables nearer to where they plug in would probably solve that.

I think i'd place my bet on the end terminations being marginal.

[Morgoroth]

Years ago I was working as technical inspector (subway constructions) ans one day we start getting random false signaling and problems with networking. Problem was someone cut and stolen a big naked copper cable on electric tray.

I don't know if your norm have same requirements, but worth to check the grounding if you can't find any other problem.

[Stonent]

Most cable issues are at the connection point. Either pairs interfacing with the rj45 or the punch down on the jack.

Bluejean cables tested many cables for cat 6 compliance and they failed. The article is really interesting. They spent the money and got the Fluke cable analyzer.
https://www.bluejeanscable.com/articles/is-your-cat6-a-dog.htm

[Monkeh]

Most cable issues are at the connection point. Either pairs interfacing with the rj45 or the punch down on the jack.

Bluejean cables tested many cables for cat 6 compliance and they failed. The article is really interesting. They spent the money and got the Fluke cable analyzer.
https://www.bluejeanscable.com/articles/is-your-cat6-a-dog.htm

Shame on them for not naming and shaming.

[XOIIO]

Ooooh god that setup, that is pure porn. I'd love to have my server rack that full, but then it would cost an even more ludicrous amount of power. As it is two servers, a fan and a switch are using up almost as much power as an entire household per month.

[kingofkya]

Ooooh god that setup, that is pure porn. I'd love to have my server rack that full, but then it would cost an even more ludicrous amount of power. As it is two servers, a fan and a switch are using up almost as much power as an entire household per month.

Yeah its a 210v 20a curcit per rack quite a bit of power running 24x7

That nothing i should post a picture of the router rack 80something giga bits worth of zayo cogent level3 etc....


Yeah i think i relly have to agree at this point it has to be the cables.  Also that rack was just an example its not one of the problem ones. The real problem rack is just really hard to get a picture of, setup the same way though. BUt i think i will have my techs go look for cables like that anyways, and just get thouse stressed cables out of the bundels.

[Monkeh]

If that rack's an average example it's no wonder you have problem ones!

Do yourself a favour and pick up some support rails so the cables don't hang all the way out to the back. Maybe just some lacing bars.

[blacksheeplogic]

If that rack's an average example it's no wonder you have problem ones!

Do yourself a favour and pick up some support rails so the cables don't hang all the way out to the back. Maybe just some lacing bars.

It looks to me that whoever did the original install under bid the job. The cabling needs to be punched down, secured and then short patch cords used for long term reliability. Terminating with a RJ45 directly into the service gear will be problematic long term.

[Monkeh]

If that rack's an average example it's no wonder you have problem ones!

Do yourself a favour and pick up some support rails so the cables don't hang all the way out to the back. Maybe just some lacing bars.

It looks to me that whoever did the original install under bid the job. The cabling needs to be punched down, secured and then short patch cords used for long term reliability. Terminating with a RJ45 directly into the service gear will be problematic long term.

They're patch cables running from a switch in the same rack..

[station240]

Have you tried getting one of the failed cables reterminated with new plugs ?
That would give you an idea if it's the cable or the termination.

[kingofkya]

If that rack's an average example it's no wonder you have problem ones!

Do yourself a favour and pick up some support rails so the cables don't hang all the way out to the back. Maybe just some lacing bars.

It looks to me that whoever did the original install under bid the job. The cabling needs to be punched down, secured and then short patch cords used for long term reliability. Terminating with a RJ45 directly into the service gear will be problematic long term.

Now that makes no sense why would you add another failure point of a patch pannel. Not that i havent seen people do that... Centrilised setup (what i think your sugesting) is absoultely not happening.

If that rack's an average example it's no wonder you have problem ones!

Do yourself a favour and pick up some support rails so the cables don't hang all the way out to the back. Maybe just some lacing bars.

Anywasy the lacing bars sugestion is good somehting like this is what your thinking?
http://www.showmecables.com/product/vertical-lacer-strips-44-ru-6-pk.aspx?utm_source=google&utm_medium=cse&utm_campaign=LACE-44-1P&gclid=CjwKEAjw8bO3BRDp0bP_vL-7_lASJACL_d6wNr_fgmlZkWQETY5H6tO2f6ae3SJAWzsdpwVb8GcsPhoCRSbw_wcB

Have you tried getting one of the failed cables reterminated with new plugs ?
That would give you an idea if it's the cable or the termination.

The problem is that we have customer that need thigns fixed in 1-5min  cant be screwing around reterminating a cable.
And I have yet to beable to reproduce the issues after the the orginal cable is disconnected even if i leave it in place.

[Monkeh]

If that rack's an average example it's no wonder you have problem ones!

Do yourself a favour and pick up some support rails so the cables don't hang all the way out to the back. Maybe just some lacing bars.

Anywasy the lacing bars sugestion is good somehting like this is what your thinking?
http://www.showmecables.com/product/vertical-lacer-strips-44-ru-6-pk.aspx?utm_source=google&utm_medium=cse&utm_campaign=LACE-44-1P&gclid=CjwKEAjw8bO3BRDp0bP_vL-7_lASJACL_d6wNr_fgmlZkWQETY5H6tO2f6ae3SJAWzsdpwVb8GcsPhoCRSbw_wcBble to reproduce the issues after the the orginal cable is disconnected even if i leave it in place.

You do need some vertical support, but it's the horizontal I'm bothered by.

The servers stick out a bit further, but a few of these sticking out would let you tie off at the same height as the ports and get rid of much of the weight.

Perhaps combined with these for vertical support for the bundles. Can't be sure how convenient those would be to fix to your racks though.

E: Perhaps not quite that wide, though.

[blacksheeplogic]

It looks to me that whoever did the original install under bid the job. The cabling needs to be punched down, secured and then short patch cords used for long term reliability. Terminating with a RJ45 directly into the service gear will be problematic long term.

Now that makes no sense why would you add another failure point of a patch pannel.

How many call backs have you had on a 110 IDC within an IDF assuming the cabling was tested? Zero?

What sort of electrical/environmental stress do you think is on an ethernet cable within the data center that would cause regular failure? Did physical inspection show damage to the cables?

What's the cost of each failure to the business?

Stress on the cables and the number of failures point to a likely problem with the plug assuming there is no evidence of physical damage. A TDR trace  would show if there are any problems within the cable (unlikely).

Buy connecting the servers directly to the switch ports using a patch cord makes securing the cables more difficult as you have to allow for change and it's not practical or efficient if you have to unbundle the cables in addition there is a risk of damage to other cables.

Adding additional support/cable trays will help if the client's looking for minimal cost.