linux v9 sleep wake timers

[chairmanmao]

Jeff responded to a question I posed back in March of this year about sleep/wake timers in linux v9, to the effect that they weren't in the linux implementation but would be reasonably straightforward to implement (I asked the question because my sage server runs infrequently enough that I don't want the server live 24/7, in order to save electricity).

Has anyone implemented the necessary functionality in the last few months? Or researched it some more to see what's involved?

[Narflex]

I can tell you no one has added it because I would have seen the change come through GitHub.

[Fuzzy]

Does ubuntu server (used as an example here) even have the ability to detect when it is idle? In the windows version, Sage does not put the computer to sleep - that's all on windows to do that. All Sage does is tell windows when it needs the computer to wake up, and tell windows when it cannot go to sleep.

I'm thinking the furthest we might want Sage to do on Linux is to perhaps write a simple file (or environment variable) with it's power wishes, and an external cron launched script might be able to read that file (or environment variable) to determine if it can sleep, and until when, and utilize rtcwake to do so. Doing it this way would then allow the script to be modified as wished to check other things before putting the system to sleep.

SAGEPOWER=WAKE_AT_UTC mm-dd-yyyy hh:mm:ss

SAGEPOWER=KEEP_AWAKE

Sage already runs a process continuously to determine if it can sleep and when it should wake, so adding this simple output of that value shouldn't be too difficult.

[SomeWhatLost]

I have a bit of a philosophical question, but with modern CPU's and their power efficiency out of the box, plus the fact that they adjust their clock to fit the load/etc, is "sleep" still a big deal?
obviously "sleeping" is still better than being awake, but is it really that much better than idling these days?
just curious really... anyone do any measurements with a kill-o-watt or anything?

[Taddeusz]

Unless you're system is running full tilt all the time, which is highly unlikely, it's like running a 100-150W lightbulb 24/7. Of course that is highly dependent on your components and how many hard drives you have.

[EnterNoEscape]

I did look into this last year for OpenDCT and we don't support it there because all of the suggestions I found involved using scripts to tell the program what's going on with the system and a utility to tell the computer when to wake back up. There's comments in the code on how we could implement it, but it felt too hacky for my taste. I was planning on contributing whatever I did there to the SageTV project too, but like I said, it never panned out.

[KryptoNyte]

Originally Posted by SomeWhatLost

I have a bit of a philosophical question, but with modern CPU's and their power efficiency out of the box, plus the fact that they adjust their clock to fit the load/etc, is "sleep" still a big deal?
obviously "sleeping" is still better than being awake, but is it really that much better than idling these days?
just curious really... anyone do any measurements with a kill-o-watt or anything?

I'm at around 150 watts with my SageTV server in its lowest power state (when not in sleep). It's on a second floor office in my home, and the central AC unit can barely get cool air to this room to begin with. For me, this is the difference between sitting in a 75 degree office (in sleep) or an 82 degree office (sleep disabled) over the course of a nasty hot day, and let's face, there is no shortage of nasty hot days.

In the winter time, I'm happy to leave it running.

[SomeWhatLost]

thanks for the replies, I was just curious, I though idle would be a bit (ie much) less power... especially since the drives can sleep/spin down independently of the computer...

[Fuzzy]

a typical (low end) 600W computer power supply is likely to burn 40-60W or so on it's own when turned on, because they are designed for peak efficiency at about 50% load, not 0%. Going to S3 sleep and running just the standby portion of the power supply is a pretty hefty savings, even if the CPU itself is only drawing 10-15W at idle, from the wall it could be pulling 80 easily.

[SomeWhatLost]

that seems like a crappy power supply design... but I guess you are right...
I just checked my Sage server (no current recordings going on), it is at 120W with the occasional spike to 150W...
of course it is also my unraid server, (all drives are currently spun up for some reason? except for the 2 sage recording drives...), a windows VM running a batch VIDEO_TS -> MKV conversion on all my dvd's, and a few other things going on...
but average CPU load is <20% (although cores 5 & 6 spike to about 50%, they are the windows VM) so its technically mostly idle... 120W is more then I thought it would be at idle oh well... at least I eliminated 2 other boxes doing this VM thing, so I hope its a net gain (well power loss?) over all...