[unisog] VoIP UPS Considerations

Peter Van Epp vanepp at sfu.ca
Mon May 26 19:50:46 GMT 2008

On Mon, May 26, 2008 at 10:30:06AM -0600, Doug Doran wrote:
> Hello,
> We are currently examining the implications of using UPSs for all our 
> VoIP wiring closets.  We have done this up until now.  I have been made 
> aware however of a few institutions that provide UPS power for the 
> building head end switches but do not provide UPS power for all their 
> POE switches.  They simply provide emergency TDM phones for each work 
> area or floor.  I am doing a survey that I will post the statistical 
> results to the list without identifying institutions.
> Please respond if you would like.
> 1.  Has your institution deployed over 200 VoIP phones? (y/n)


> 2.  Is it your practice to provide UPS power to all VoIP POE switches? (y/n)

	So far yes (the first production site is one campus in one building 
which has building UPS on diesel). We by now however may have other places
that aren't on diesel. As I recall the Ecom folks (911) recommend (as noted 
earlier no one has been able to point us at a Canadian law on the subject) 
about an hour of backup capacity on a VOIP UPS. All the trunk paths that I am 
aware of are on life safety diesel backed up UPS. At least one new building I 
know of got POTs after I pointed out the path wasn't all on life safety UPS 
and thus VOIP was a poor choice unless we were willing to upgrade the entire 
IP path (frankly not practical since there was no funding for it). 
	Certainly our PER nodes can take a trunk over a fibre (avoiding the 
cost of a multipair copper pull which can be expensive) and the copper pull
is likely factored in to the cost of a building anyway (unlike 3 or 4 
generators in unrelated buildings on the IP path :-)). 

> 3. If you deploy UPSs for POE switches what is your power out run time?

	As long as the diesel has fuel so far. As noted below I believe the 
recommendation from the 911 folks was about an hour but that is from memory
and would only apply to ours. You should probably ask your 911 operator and/or
your risk manager :-). 

> 4. If you do not deploy UPSs for all POE switches do you deploy 
> traditional phones for backup? (y/n)

	Yes, we have at least two and possibly a few more standard (carrier
provided) POTS lines in security and management offices for emergencies even
with UPS, as network problems could knock out VOIP as we have dark fibre 
between campuses that our trunks run across. 

> 5. If you answered yes to Q4 how do you determine the frequency of 
> traditional phone deployment?

	At least one per site (which I'm not sure is sufficient, but is better
than none per site :-)). When we did a single lab area here on main campus as
a pilot (some 50 or so VOIP connections across three or four rooms on one 
floor of a building) we had a single POTS line backup in a universally 
available break room (the labs are card access controlled which you need to 
consider for emergency phones) and signs in each lab area directing
people to the POTS phone in the case of an emergency. 
	We are also VOIP trunked internally although we still have T1 backup
links between the switches in place. So far our carrier(s) don't offer VOIP
trunking to the CO but when they do we will likely opt for that as a cost
saving measure (hopefully with at least 1 T1 backup circuit thopugh). 
	When we first did VOIP trunking in production (long before phones on
the desktop because that does have an immediate payback at least here :-)) we 
used to put 2  BRIs in to the analog PER node and route 911 to those two BRIs, 
partly for localization reasons, as our main trunks are on main campus and an 
emergency crew on the wrong campus because of not thinking would be a disaster 
and partly as insurance against network failure which could take out the IP 
trunks. IP and POTS are exteremely different in concept and both sides (IP and 
conventional telephone) have a hard time understanding the other. In the phone 
world a mistake can be deadly so you need to take great care with this stuff. 
A common mistake I find our IP folks making is assuming because the trunks and 
phones are IP that telephone routing is IP (or at least IP like) and that 
simply isn't true. IP failover that comes out a different place in the network 
can have disastrous concequences in a phone system the primary example being 
emergency services in the wrong physical place because they assumed that where 
the call came from in the telephone network is where the emergency is, but not 
working because your phone number isn't on the expected phone switch is also 

Peter Van Epp / Operations and Technical Support 
Simon Fraser University, Burnaby, B.C. Canada

More information about the unisog mailing list