> * Do you use UPSes for all network gear across campus?
We've built up over the years to having UPSes in 90% of our network
closets across campus. We've aimed for about 1 hour of backup
historically in our larger closets, but after recent upgrades we're
down to 15-30 minutes. In many cases these are on generators so this
Our primary concern historically has been to avoid equipment failures
due to bad power. Nowadays, life safety concerns are more important
due to VoIP, security/card access/fire alarms, and other
infrastructure communications needs (e.g. building management
To this end, all UPSses we are buying now are of the online,
double-conversion type. We have mainly APC Smart-UPS RT1500, RT2200
and APC Symmetra RM 2-6kVA units. The data centers have large 3-phase
Symmetra units and generator backup. Historically, we've used APC
Smart-UPS SU1400 and SU2200 units.
> * Is IT responsible for the upkeep, or is facilities/physical plant responsible?
> * What is your experience with keeping up with many UPSes?
All UPSes have SNMP monitoring cards that we monitor with Nagios,
Cricket and email alerts, which is key to keeping up with them and
fixing any battery issues before they get so bad that the batteries
expand and become stuck inside the units rendering the UPSes trash.
The newer ones also have the environmental monitoring capability, so
we keep track of temperature and in some cases humidity in our network
closets. On more than one occasion this has helped us avoid network
downtime due to failure of air conditioning units, which we've come to
experience as the #1 risk to network downtime these days, so catching
these issues quickly has been helpful. Mechanical systems are just
not as reliable as electronic systems.
We've been getting about 10-15 years lifespan out of the UPS
equipment, and between 2-6 years typically for batteries, depending on
environmental factors. A bunch of our APC SU2200's are still going
after 14 years--their original batteries lasted 9 years before we
started seeing failures. I think the older UPSes with the larger
12v18Ah batteries are more durable than the newer ones which all seem
to come with the smaller 12v7Ah or 12v5Ah batteries.
We rebuild all of our battery packs in-house (but not the actual
batteries themselves). It is a lot of work--our rate of battery
failure is maybe 1-3 a week during "peak" season when the temperatures
are high (not all of our closets have cooling unfortunately). Part of
the reason is that we've unfortunately not been able to budget for
proactive battery replacement, but we've been able to get maintenance
budget allocated to keep up with replacements after failure.
In our last few network equipment upgrade cycles, we've also made sure
to add UPS upgrades and/or battery replacements as part of the network
> * If you don't use UPSes, what is your experience with this? Have you seen switch failures after power issues?
We still have some locations with no UPSes. We see many more network
issues in locations without UPS backup, although not too many
"permanent" failures that were not recoverable by reloading the
switches software/configuration. Our newest switches don't like to
lose power suddently, or they corrupt their filesystem (Juniper EX
line) but this is somewhat mitigated by their dual-resilient root
partitions which back each other up.
There is also the "peace of mind" that comes from not having the
"unknown variables" of power issues come into the picture when
debugging network issues. We've had weird cases in the past where
brownouts have caused strange, hard to debug behavior in network
Participation and subscription information for this EDUCAUSE Constituent Group discussion list can be found at http://www.educause.edu/groups/.