Slashdot Mirror
Open Power Management Console
Scott Haugaard writes "There's a cool open code/open hardware power management device page has the schematics, programs, and pictures of a homebrew remote power console. Use this to cold boot those unresponsive servers remotely, or to shut them down at night via scripts and bring them up again before your users log on to save power. Solder, TCL, LEDs, wires galore. Manage power to 16 servers from two wires on a serial port."
Oh, wait. A link from Slashdot shuts down your servers *anyway*.
--
Evan
"$30 for the One True Ring. $10 each additional ring!" -- JRR "Bob" Tolkien
For those under-developed area of the world where electricity is still a luxury (California for example)
--
Je t'aime Stéphanie
=-=-=-=-=
=-=-=-=-=-=-=-=-=
Oh bother.
Take a step back and take a look at this circuit. Now take a step forward. Now take a step back again... nevermind.
Take a look at what this circuit really DOES. I examined the schematic, and having designed computer-driven power relays before, I think I have a good idea of what this circuit actually does. Think about this: it introduces a single-point failure node into the power supply of up to 16 computers. So now if this circuit fails, it can potentially take down a whole room full of systems. Or even worse, if a cracker gets into it, he can blow your systems away at will. I recently had the displeasure of having an old linux test box get cracked through the old FTP exploit, boy that skript kiddee enjoyed himself rebooting my system over and over, as I later determined from reading the logs.
Anyway, I used to build similar power relays, I just did the computer front end, and some expert electrics guys assembled the solenoid driven switching system. I worked at a stereo shop back in the mid-70s, they wanted to put all their demonstration stereos on relays, then have a computer turn them on and off, in preset configurations. So I got an Apple II with a Corvus 5Mb hard disk, I set up a UCSD Pascal system, and wrote a nice little program to browse through presets and create your own.
Now came the hard part. I set up 4 PSIO boards, which were these cool boards with 4 parallel ports each. So I had 64 bits of parallel I/O, so I could control 32 L/R stereo powered audio channels.
On the audio side, the electrics guy wired up a series of 64 switchable solenoids capable of carrying about 500watts of power each, but were capable of being triggered by the low voltage of a parallel port signal. One bit per relay. I just wrote 1s or 0s to the port, and the relays all triggered to their new state, on or off. Hit the whole array with 0s to turn everything off, 1s to turn them on. The electrics guy wired the whole thing up with thick copper Monster Cable, and there were plenty of worries over grounding, which was way beyond my abilities, but solved by the application of thick braided cables. I didn't want to go anywhere near that assembly, since it was hooked up to high-powered stereo equipment, like Phase Linear 400s, a few assorted tube amps, etc. These guys were demonstrating high-end audiofile stuff, and the speakers were the real key. They wanted to be able to switch different amplifiers to different arrays of stereo speakers. You could turn the relays on or off in pairs, or leave 2, 4, or more pairs of speakers on to listen to them in parallel. And there's where the problem started.
I discovered that my conceptually clean idea of the relays going on and off in a crisp square wave transition was not reality. The relays took longer to close than to open. The switch bounced closed and took a moment to stabilize. We didn't realize this when we first started playing with it, and every time they tried to switch settings, it blew up the audio amps (and they were mighty peeved at blowing expensive PA400 amps!). We had to pull everything out, run a test signal through the relays and look at the output on an oscilloscope, where we discovered that if we switched amp relays, there was a brief period where both settings were open, causing the amps and speakers to run parallel, short circuiting everything and melting it all down in one huge imploding glumph. Ooops!
So I had to rewrite all the Pascal switching routines in 6502 assembly language. I could keep the user interface, and just pass the switching preset strings to the ASM routine. The program turned off all the relays at once, then ran a short ASM delay loop, then wrote the new switch presets. Whew, that was a bitch, I had to time both the relays's switch performance as well as the ASM delay loops on the oscilloscope, and got everything timed out perfectly. But it was pretty dicey there for a while.
So you see why I'm skeptical about power management through computer controlled relay systems. IMHO, it's like setting up a rack of servers connected to a light switch on the wall, and there's a security guard standing by it who is only supposed to allow authorized users to flip the switch. Except he's probably just as likely to bump the damn switch and blow out the whole server farm.
I keep all my servers in the basement of my parent's house, and use my mother as the control mechanism. She is much more difficult to control than a PERL script, because she doesn't understand the concept of warning lights or switches. So I've hooked all of them up to speakers and I use mp3 files of songs she knows that correspond to warning messages. This leads to conversations like this:
"Hullo, mum? Is the server making noise?"
"Um, yes. I think."
"Is it the Barenaked Ladies."
"No...it sounds like Johnny Cash."
"Is it 'Johnny Yuma?'"
"No, that one about the car. 'One piece at a time.'"
"Okay, that means the switch is dead. You need to jab a pencil into the little grey box with the red triangle on it."
"Do you mean the one with the big sign that says 'don't touch?'"
"No, the one with the red triangle."
She's still easier to work with than BiznessOnline.com.
Hey freaks: now you're ju