I am about to upgrade our server room's power plan and would kindly ask you for your help with a proper, optimal design / advice on "good practice" approach. Please note that our business is small, but would nevertheless want to have a solid solution.

What I have so far:

  • 2 x APC SMX3000rmhv2u rack UPS w/ management cards
  • 2 x simple PDUs built into the rack cabinet
  • 1 x 3500W inverter generator for emergency

Any of the UPS's would be enough for equipment in the server room, but would like to use of them for redundancy and if possible, extended runtime.

Which setup would make most sense?

Utility,Generator ==> ats ==> UPS1, UPS2 ==--> PDU ==> load

This setup joins the output of both UPS units with a Y cable. It looks simple and for that I like it, but I wonder if it's ok to run UPS outputs in parallel just with a Y cable. Reading through the documentation of Eaton and APC they seem to call this set-up a parallel-capacity, but I am not sure what component shall be joining those two UPS outputs. Would this approach work? Would it add capacity (double the runtime)? Would it be redundant?

Utility,Generator ==> ats ==> UPS1, UPS2 ==> ats --> PDU ==> load

Or this one? This one adds redundancy, might be an overkill with ATS's, but it seems like a safe bet.

Another option I see is to have both UPS units directly connected to Utility, then their outputs go to ats, from ats to pdu and then to load. In case of power failure I'd need to manually unplug one of the UPS units from utility and plug it into generator.

Unfortunately, I can't post pics :/


Your first setup (with the Y cable) is the basic idea for a parallel UPS system, but you won't actually use a Y-cable; the UPSes would be hard-wired to an external paralleling system which should also provide a bypass connection -- basically a circuit breaker panel used in reverse. The UPSes would also need to communicate with each other to stay in sync for frequency and internal bypass operation. I haven't seen those sorts of functions available on smaller units like the one you listed, only on larger datacenter-scale units. You would get a bit more runtime with both units running, but it's important to make sure you're tracking your load and runtimes assuming one of the UPSes is unavailable (disconnected for maintenance, or has failed) to preserve the redundancy.

Better designs have dual PSUs in all of the equipment, and use smal rack-mount ATSs only for exception cases. Each PSU connects to separate PDUs, each fed in turn by separate UPSes. Ideally each UPS then connects to separate ATSes (and, in higher tiers of datacenters, those are fed from separate generators and separate utility feeds.)

In your case, I'd do: {Utility, Generator) ==> ATS --> { UPS1, UPS2 } ==> { PDU1, PDU2 } ==> { PSUs, or ATS } --> Single-cabled devices as needed

This does leave the upstream ATS as a single point of failure, but I think that's unavoidable when hardwired to smaller generators like yours.


I'd recommend sizing for the runtime and protections you actually require. Please look into extended-runtime UPS solutions compatible with your units. You can add extra runtime via additional batteries.

Is all of your equipment on dual power supplies? Are the power distribution modes configurable on the servers?

  • Thanks for answer. Only servers have dual PSUs, bunch of network and backup equipment does not. – Charlie Jun 22 '15 at 7:25

Your Answer

By clicking "Post Your Answer", you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.