Grounding and Bonding in a data center?

Status
Not open for further replies.
J

jimmycher

New member
Location
Ashburn VA
Can someone give me a simple explanation on why I need to ground server racks? I've had electricians say as long as the servers are plugged into a grounded outlet, it shouldn't matter.

I believe a short will always go to ground, but unless the racks are grounded, the short might go through me before it is goes to ground ??

My intial request was ignored, but recently a small AC unit leaked water onto the racks while I was pulling cables.

The TIA/EIA-942 Telecommunications Infrastructure Standard for Data Centers have guidelines/requirements, but that is primarily with ESD and uptime, not safety.

As I must persuade management, specific safety examples would be helpful.

Thanks
 
jimmycher said:
Can someone give me a simple explanation on why I need to ground server racks? I've had electricians say as long as the servers are plugged into a grounded outlet, it shouldn't matter.

I believe a short will always go to ground, but unless the racks are grounded, the short might go through me before it is goes to ground ??

My intial request was ignored, but recently a small AC unit leaked water onto the racks while I was pulling cables.

The TIA/EIA-942 Telecommunications Infrastructure Standard for Data Centers have guidelines/requirements, but that is primarily with ESD and uptime, not safety.

As I must persuade management, specific safety examples would be helpful.

Thanks
Click to expand...
If you search through existing threads, or look at some of the Mike Holt video clips on grounding and bonding, you will see that what is important for electrical safety is a path which can allow enough current to flow to trip the breaker or fuse if there is a fault.
The EGC will do that just fine. And a short will NOT go to ground. It may go back to the source of the current (service neutral or transformer secondary) via the EGC, the earth, or more often both. But the earth path is going to be higher resistance than the EGC and so will carry relatively little of the fault current.
Worst case, if the fault is a dead short from line to grounded metal, the voltage there to "real" ground will be half of the line voltage until the breaker trips.

As for the AC dripping, that is the reason that the code requires a drip pan or other device between the A/C or other plumbing and building electrical equipment. The NEC does not require that over your server racks, but it would be a good idea anyway. :)

Getting back to your original question: If the servers have cords with a grounding conductor and that grounding conductor is connected internally to the server case, then that will in turn provide a ground/EGC path for the rack. Similarly if the receptacles are mounted to the rack in a way that provides good electrical contact, then they will provide the ground/EGC path for the rack. But it does not hurt to provide a separate grounding conductor/EGC to each rack. That would be company policy rather than NEC, but would have some benefits. Among other things, it would not be dependent on incidental metal to metal contact which might be hindered by paint or aluminum oxide. I have also seen data centers which use plastic washers on the rack mount screws to keep from damaging the paint!
 
Last edited:
jimmycher said:
Can someone give me a simple explanation on why I need to ground server racks? I've had electricians say as long as the servers are plugged into a grounded outlet, it shouldn't matter.

I believe a short will always go to ground, but unless the racks are grounded, the short might go through me before it is goes to ground ??

My intial request was ignored, but recently a small AC unit leaked water onto the racks while I was pulling cables.

The TIA/EIA-942 Telecommunications Infrastructure Standard for Data Centers have guidelines/requirements, but that is primarily with ESD and uptime, not safety.

As I must persuade management, specific safety examples would be helpful.

Thanks
Click to expand...
There is nothing in the NEC which specifically requires server racks to be grounded with a bonding jumper under Article 250, Part V. However, section 250.96 use of the word "frames" can be interpreted to include server racks... but even that states where they "...are to serve as equipment grounding conductors, with or without the use of supplementary equipment grounding conductors..."
 
I wrestle with this myself. I see no guarantee of grounding the rack via a power cord equipment grounding conductor. That would ground/bond that piece of equipment, but not necessarily the rack. Some rack rails are powder coated, and many rack items have painted rack ears. So you've got a screw holding a painted object to a painted rail. I've never seen anyone scrape paint when installing rack equipment.

If the rails are plated and conductive, as opposed to painted, that helps.

Why the rack needs to be grounded, I'm not sure. The only way I can see to energize it would be if an internal power cord gets pinched or nicked and it manages to energize the rack frame. Many racks also have a ground screw provision, so I'm not sure how effective that will be without bonding the rack. So use those only if you know you've got a good bond to the frame (or use them to ground your rack if you have ground bars available for that purpose).

I found some green screws from Panduit for bonding equipment to the rack frame (RGTBS1032G). They have teeth under the heads and thread forming screws, so they should bite through a painted chassis. I try to use these whenever we bolt a PDU to a rack. That will hopefully ensure that the rack frame gets bonded since the PDU has a power cord out to a grounded wall receptacle.
 
Check out Motorola R-56 from a thread in CampFire Chat .......couple of members have links to it. It's a grounding/bonding spec for communication sites that boils down to anything that does not move gets a copper conductor cadwelded to it and all of them get brought back to a common point.
 
iwire said:
I know nothing about it but here is a pfd of the document John suggested.

www.radioandtrunking.com/downloads/motorola/R56_2005_manual.pdf?
Click to expand...

A very impressive document.
A good reference and good spare time reading.

A quick skim turned up:
1. In terms of galvanic corrosion, a connection to steel in concrete will be very similar to a connection to copper.
2. Do NOT exothermically weld to tower guy wires, even though they need to be grounded!

:)
 
Last edited:
Status
Not open for further replies.
Top