Date: Fri, 06 Mar 1998 17:44:27 -0800 (PST) From: Simon Shapiro <shimon@simon-shapiro.org> To: Karl Denninger <karl@mcs.net> Cc: Bob Bishop <rb@gid.co.uk>, dmlb@ragnet.demon.co.uk, wilko@yedi.iaf.nl, julian@whistle.com, hackers@FreeBSD.ORG, "marino.ladavac@siemens.at" <lada@ws2301.gud.siemens.at> Subject: Re: SCSI Bus redundancy... Message-ID: <XFMail.980306174427.shimon@simon-shapiro.org> In-Reply-To: <19980306191749.01367@mcs.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On 07-Mar-98 Karl Denninger wrote: ... >> BTW, 48VDC P/S for a PC is about $50.00, where an AC one is $15.00. > > Yeah, well, that's because DC/DC conversion takes two forms - expensive > and > cheap but wasteful in terms of dissipation (ie: basic regulation). Yup. Many things in this earthly life tend to be that way. > AC/DC conversion by switching power supplies is so cheap now that it gets > rediculous. Of course, its also cheap enough that lots of folks don't > pay > attention to minor little things like the tolerances of the components > they > use either. Which quickly makes the cheap very expensive. > I used to be involved in Satellite Earth Station work (microcontrollers > for > same); a goodly amount of the stuff out there, particularly MCL's > Klystron > tuners and amplifiers, probably still has my code running around in them. > > The people who own these devices tend to be a bit anal about electricity. > They also eat huge gobs of it. This means generators and monster UPS > systems. Unfortunately, virtually all of *this* kind of equipment wants > 220 3-phase in rather serious amperage ratings (or worse, 480 V). > > Shutting off a hot klystron by removing power is a great way to buy a new > $25,000 tube; there is enough energy being dissipated in those things > that > if you shut them down without going through a clean power-down sequence > (ie: forced air cooling for a few minutes) almost insures that you will > damage the alignment of the plates in the tube (due to heat-related > warpage). > Given that under typical operation you have ~3KW *emitted* from the > business > end of these things in the form of a nice electron beam you really don't > want that to happen. And those are the "small" ones; some of the larger > ones had ~30KW rated outputs. Ugliest radio station I visited was a 1.8 MegaWatt short radio station. They had a direct feed from the power station down the mountain. I think they had their own set of generators at the power house. Same story. you do NOT shut these down. The desert installation I referred to was running a 2GHz microwave link at 1GW contineous. You should see what happened to seagulls who tried to fly in front of the dishes... We used the diesel-flywheel-generator-electric_motor arrangement. > What I've seen used is a monster online UPS installation with battery > room, > frequently comprised of open wet cells, and a big honking generator for > utility backup. Some of these are big enough to stand in. The amount of hydrogen boubling up classifies these rooms as explosives. > The batteries are basically there as a "smoothing" function - they keep > things going until the generator can start up, warm up somewhat, and get > to operating power. Yup. something between a filer and capacitor. They are very effective in that role. A bit too much for the typical FreeBSD server, but the lesson is the same. Have a battery somewhere in the loop, so your power input to the equipment never goes down. I do not remember the numbers, but somewhere around 6+ servers, it pays to put them in a rackmount and feed them DC. Simon To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?XFMail.980306174427.shimon>