IMO, with this approach you may just be creating more opportunities for the whole system to fail rather than providing redundancy, as your recent experience might prove.

The redundant power supplies in enterprise servers are there first a foremost to protect against power supply failures. After hard drives, and perhaps cooling fans, these have been the most common component failures that we've experienced. Note that many manufacturers, such as Dell, provide a choice of Y power cable for redundant PSUs. These are meant to reduce cable clutter when you have a single power feed, but still provide PSU redundancy.

If you were in a first-class colocation facility, where you were receiving two hightly reliable power feeds, I could see running each supply off of a separate feed.

But in your installation it means spec'ing each UPS to be able to operate every server attached to it simultaneously for the desired run time. So you purchase two of the same UPS, maintain them as best you can, then the day comes when you lose power and _one_ of them fails, what are the chances that the other will hold up all of your servers? This is exactly what you just went through. Battery failures are the most common problem. If the batteries in the first UPS are borderline, and unable to run 1/2 of your equipment, those servers will then load the second UPS, which is likely to be in a similar state of health and even more likely to fail under the increased load.



----- Original Message ----- From: "Ben Scott" <mailvor...@gmail.com>
To: "NT System Admin Issues" <ntsysadmin@lyris.sunbelt-software.com>
Sent: Monday, February 09, 2009 9:57 PM
Subject: UPS recommendations


Hi all,

 We had a power outage today.  I looked over at the server rack just
in time to see one of the UPSes light up like a Christmas tree, shriek
like an injured parakeet, and then kill itself.  (Admitted it was old,
but a graceful failure this was not.)  The servers with redundant
supplies failed over to the other UPS, which promptly went into
over-current alarm and dropped the load.  Either said UPS's management
software has been grossly misreporting its load, or two UPSes at 40%
load doesn't include enough margin during transfer.  Any which way you
slice it, it's time to buy some new UPSes.  I'm going to ask for two
entirely new 1400 or 2200 VA units (existing were 1000 VA), although
budget may be an issue.

 What do people like for UPSes, *and why*?  I don't see much
variation across manufactures in a given price band.  At a given
dollar amount, it seems I get roughly the same capacity, features,
etc.  I'm thinking differences in management software and quality of
support don't show up in a spec sheet.  Comments on that front are
especially welcomed.

 In particular, I'm interested in how to manage a multiple-server,
multiple-UPS scenario.  Our two biggest servers have redundant
supplies.  I'd like to plug each supply into a different UPS.  So each
UPS will be powering multiple servers, and each server will be drawing
power from multiple UPSes.  I imagine that makes the management
software configuration a bit trickier, specially since a lot of
management packages used to assume one-UPS-per-server.

-- Ben

~ Finally, powerful endpoint security that ISN'T a resource hog! ~
~ <http://www.sunbeltsoftware.com/Business/VIPRE-Enterprise/> ~


~ Finally, powerful endpoint security that ISN'T a resource hog! ~
~ <http://www.sunbeltsoftware.com/Business/VIPRE-Enterprise/>  ~

Reply via email to