NANOGers home data centers - What's in your closet?

Alex Rubenstein alex at corp.nac.net
Wed Aug 17 02:07:15 UTC 2011


> All the actual EE's I know are most impressed with the PowerWare (now
> Eaton Power) designs.  For insance their 5110 is a line-interactive
> design built with quality components.  The last I looked APC did not
> have a line-interactive design in this price range; they were all the
> "standby" design.  Eaton does have some cheaper standby units, I can't
> make any comment about them.

Most EE's don't deal with day to day operations of datacenters.

We have been a long time (14 years) user of Exide / Powerware / Invensys / Eaton UPS systems. I have many 9315 and 9395's, ranging from 80 kva to 500 kva. They are generally good units. It is clear that as time as gone forward (9315 to 9395 transition) that the quality of build has gone done, cheaper components, etc. Simply look at the weight differences, that speaks a lot. Eaton service is expensive. Emergency parts availability is sometimes an issue. I have heard that because of less-than-stellar market acceptance of 9395, they are modernizing the 9315, which was originally to be EOL'ed. I am not sure what that means.

Our last build, after serious consideration, we decided to go with GE SG500's. In my experience, they are a much better engineered unit, with a considerably more knowledgeable sales and service contingent.

I can't speak about the 5110.


> With any UPS the key is replacing the batteries at the appropriate
> time.  Most batteries are rated for 3-5 years.  If conditions are
> right, you might be able to push that to 6 or so.  If your battery is
> older than that, replace it or you might as well not have a UPS.

This is an entirely different conversation. Battery monitoring is a requirement. Btech (and others) allow for daily visibility of battery health and failure trends. If you don't have this, you aren't serious about your datacenter. 

Batteries can fail anywhere from 4 minutes to 10 years after they are installed, and they never fail all at once.. so why replace them all at once?






More information about the NANOG mailing list