They have deep racks if you have super deep servers, but 90% of the time the standard, run of the mill 40" deep rack is fine. Also, brackets exist to rotate the 0U PDUs 90 degrees (as shown in one of my earlier pictures), but by and large it's not a problem. We've only had to go with the 48" deep racks a few times to accommodate a very deep machine with a CMA. I prefer 2 30A @ 208V PDUs or 60A @ 208V PDUs, depending upon the density. In some cases we have the PDUs with the dual 30A inputs. It's really 2 PDUs in the same physical chassis with a common management interface. the 0U PDU gives you the advantage of a short power cord that can nearly directly to an adjacent plug instead of having to route it in/out/down/up with the network cables. That's my perspective.
We have no 120V circuits. You end up wasting another 1-2% in efficiency by running at 120V and capital costs for twice as many PDUs and twice as much copper. That's easy money to get back. On Thu, Apr 4, 2013 at 3:58 PM, Robert Hajime Lanning <lann...@lanning.cc>wrote: > On 04/04/13 07:12, Leon Towns-von Stauber wrote: > >> I've been in datacenters like that: single circuit per rack, maybe 15- >> or 20-amp, maybe even 110V. (I'm thinking of a particular one in Fremont >> right now....) >> >> I don't consider that kind of power density acceptable these days. Two >> or three 30A circuits per rack (or the equivalent) are needed to make >> sufficient use of available space. They're not hard to find. >> > > The data center I am at, I have provisioned 4x 20A 120V circuits (billed > as 2x HA pair) per rack for storage and 5x 20A 120V circuits for compute. > The fifth circuit was really because we need more outlets. The compute > rack contains 16x 1U servers, 5x 1U PDUs, 2x 1U switches, 1x 1U IP KVM, 1x > 2U patch panel. There is a 1U gap between all servers. > > I use 1U remote switchable PDUs. I don't like the 0U (vertical) PDUs. > They get in the way of deep equipment and make dealing with dealing with > the rear of the rails. > > I load every circuit to about 7A to 9A. This allows a circuit failure to > not blow its peer, as all equipment are dual homed for power. > > At $JOB[-1] we had unbalanced power usage. Some circuits running 2A, > while its peer is at 22A (yes, it took awhile, but we did blow the breaker > a few times.) All equipment was single homed. So it took downtime to > rebalance, after getting a circuit usage report from data center services. > (We only had dumb PDUs.) > > -- > Mr. Flibble > King of the Potato People > ______________________________**_________________ > Tech mailing list > Tech@lists.lopsa.org > https://lists.lopsa.org/cgi-**bin/mailman/listinfo/tech<https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech> > This list provided by the League of Professional System Administrators > http://lopsa.org/ >
_______________________________________________ Tech mailing list Tech@lists.lopsa.org https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech This list provided by the League of Professional System Administrators http://lopsa.org/
