FlightAware Discussions

Dump1090 power consumption

Folks, this is curiosity.

I have been experimenting with the Pi vcgencmd measure_temp to see the effect of running various services. I notice the temp varies by around 5C when dump1090-fa.service is stopped or started.

Appreciate ambient temperatures are a factor but does dump1090 use more processing power during daytime with much aircraft traffic and less at night or does the service use CPU power regardless of signals received?

Geoff

Does this graph from my receiver answer the question?

It’s the overall CPU utilization of the device. There is nothing else running on it.
Over day i do have up to 120 aircraft on screen at the same time where low traffic during night can be between 2 and 10

That is very interesting @foxhunter Looks like your CPU obviously works harder when aircraft traffic is high then tails off around same period when it quietens down.

Although I am quite close to London Heathrow I only have an indoor unit working on a Zero. I’ve only monitored usage using htop and my dump1090 seems to be taking around 50%

Is that graph from a Pi3 or 4?

Geoff

The graph is taken from my Raspberry Pi3B using the Flightaware pro plus stick with an additional filter, linked to an Vinnant 7.5dBi Antenna

I also have an Airspy stick which is using more CPU based on the settings.

There are several users here active using a Zero for feeding in general, nobody reported real issues with it.

I am not using dump1090-fa but readsb from user @wiedehopf as a dump1090 alternative.
Currently i don’t know if this has a higher or lower CPU usage

Most of it yes.

Less coverage usually means the CPU is less dependant on time of day / traffic.

1 Like

Are you kidding?
You can’t run the airspy on a pi zero.
Only makes sense RPi 3B and later.

readsb tends to use around as much CPU as dump1090-fa but you can configure the preamble threshold in readsb if you want to.
It also will automatically use less CPU on a pizero and in return not get quite as many messages, maybe 5 % less.
dump1090-fa will in the next version do something similar for the pizero.

1 Like

Just because these two sentences are close to each other doesn’t mean they are related .
I was operating the Airspy on my 3B and the second sentence belongs to the general statement that people are using a Zero for feeding.

Updated it to make it less misunderstandable

1 Like

Understandable, this is the problem with the typed word as opposed to the spoken word. I think many have been guilty of posting something then realising a typo or punctuation alters meaning.

Geoff

A further addition to this, purely as an experiment and as aircraft traffic is very low during the early hours of the morning I created a script to check processor temperature and then shut down all my servers for 6 hours, take a CPU temperature then restart.

Interestingly, the ambient temperature in my bedroom of late has been around 23 - prior to shutting down the CPU temperature is around 49-50c

I then shut down Dump1090 - fr24feed - adsbx mlat and socat - piaware - lighttpd and finally sshd.

Just before restarting after 6 hours the CPU temperature shows around 35c so a drop of 15c

Stop and start done via crontab

Geoff

If your adsbx is still using socat it’s quite out of data, maybe update? (GitHub - adsbxchange/adsb-exchange: ADS-B Exchange Linux Setup Scripts)

Anyhow in regards to that test.
Most of those programs process messages from dump1090 so you should shut down dump1090 last if you really want to know anything about their CPU usage …
Also this test will depend on air traffic.
If you’d really want to change CPU usage i believe there are options for dump1090-fa now to reduce CPU usage quite a bit.

To me, the easiest way to determine if CPU load is due to software or traffic is simply to disconnect the antenna.

You can watch the CPU load.
You can watch the CPU temp.

Plug the antenna back in after an hour to confirm and you’ll have an answer.

To me, the easiest way to determine if CPU load is due to software or traffic is simply to disconnect the antenna.

OK, shall try that. I’m not looking to conserve anything, merely curious as to what uses what.

Geoff

If your adsbx is still using socat it’s quite out of data, maybe update? (GitHub - adsbxchange/adsb-exchange: ADS-B Exchange Linux Setup Scripts )

Thanks, shall do that.

If you’d really want to change CPU usage i believe there are options for dump1090-fa now to reduce CPU usage quite a bit.

It is more curiosity and practice with scripts rather than a need but will look at dump1090-fa settings.

Incidentally, again out of curiosity I connected my Pi to a USB device that measures volts and amps. With everything running (headless) it consumes around 440mA and when I shut down all servers it drops to 130mA

Thanks,

Geoff

Weirdly, disconnected antenna and viewing htop it made no noticable difference to dump1090 CPU usage which remained around 45-55% Actually forgot to re-attach antenna and it was off for 8 hours.

Forgot to check temperature but guess if CPU remained around 50% then temperature would be stable too.

Geoff

1 Like

Actually quite odd, out of curiosity I followed [geckoVN] suggestion in a later post in this thread, disconnected my antenna, of course aircraft dropped to zero on map but viewing htop the dump1090 CPU usage remained about the same. Forgot to re-attach antenna and even after a few hours CPU usage for dump1090 remained at around 45-55%

Geoff

It’s the same if you do not have any aircraft during nighttime. The decoder tries to decode data which are not there. I would not expect any noticable CPU load change

Guess I misinterpreted the graph you showed in the earlier message. I thought the highs and lows of the graph were busy and quiet times of aircraft traffic.

Geoff

For me, the dry heat dissipated by electronics inside my house serves as dehumidification heat.
If I didn’t have my PC running 24/7 I would have to buy a dedicated dehumidificator.

This is in the morning. NOT raining or anything.

I thought you were saying exactly the opposite above and showed a graph to make the point?

Yes, but that’s the overall CPU usage. Interestingly the ADS-B CPU usage did not change over the given time

As i am testing currently that’s the charts from my secondary receiver

image
image