ntfy: high battery usage on Cellular but not WiFi (Pixel 8)

thomasg

Hi. I've been running ntfy as a single-connection notification for some weeks now.
The ntfy developers and some users see a typical energy usage of < 1% with typical device usage for the ntfy client.
I however have been consistently seeing around 50% battery usage for my ntfy client.
The device is a Pixel 8 (not A, not Pro).

I'm usually on the road, but stayed at home for a couple of days recently, and noticed that the ntfy battery drain dropped from ₅₀% to <1%.
I've tried varying usage patterns over a couple of days, and it seems clear to me, that the device usage itself is not the difference, but the difference is strictly if I'm on WiFi or cellular.

When I'm at home, and turn WiFi off, ntfy battery usage increases by a large margin. When I'm on the road, ntfy usage is always high. When I'm at home and have WiFi on, the total phone runtime will increase from slightly less than one day, to at least 3 days with typical use and instant notifications.

On WiFi as well as on Cellular, ntfy notifications are always received immediately (within 3s of sending it), regardless of the device state, the only difference is the battery drain.

I'd expect the runtime on Cellular with WiFi off to be much better, than on Cellular + WiFi on, since cellular modems are generally more power efficient (fewer CPU wakeups) and used at the same time anyway.

Long story short: Anyone else seeing this? Any ideas what the reason could be, and/or how to debug it?

tango

thomasg

It could well be that your cellular signal is poor where you are, that would cause phone to use more battery power.

BluePixel4k

thomasg Anyone else seeing this? Any ideas what the reason could be, and/or how to debug it?

I had this problem too.
But the battery usage stats only showed about 10%. But I found out (after uninstalling ntfy) that my P8 now uses much less power than before (on cellular).

We should report this to ntfy.

EmLeX932

By any chance when you have high battery drain. are you connected to 5g networks?

Ive noticed on my phone that if i'm on 5g my battery drain is insane, like an hour can take 5-8% in some cases just for the cellular modem.

but if i lock my phone to LTE only i have great battery life and a empty owner profile can last days. This is on a pixel 7 but figured its worth mentioning. because google really did not put a power friendly modem into pixel 7's not sure about 8

what do your battery stats say if you turn it to show by system. when you have large drain on cellular data?

thomasg

The problem isn't that the phone uses more power by itself, the problem is that ntfy appears to either use a lot more resources or more probably wake-up events on Cellular.
The battery monitoring clearly shows the increased draw specifically for the ntfy app, not the Modem subsystem.

5G on or off doesn't make a difference, I tried.

thomasg

BluePixel4k I talked to the (friendly) folks at the ntfy matrix channel (including the main developer), and nobody there is having the same issue. I don't know if anyone of them uses a Pixel (8) though.

For all of them, the ntfy client is reported at < 1% battery usage.

xxx

BluePixel4k the battery usage stats only showed about 10%. But I found out (after uninstalling ntfy) that my P8 now uses much less power than before (on cellular).

I think this is some kind of 'normal':
ntfy uses a little more battery than the threema push service on my P6.
Molly-up and ntfy needed more than threema alone but much less than Molly with websocket.

Viewpoint0232

The Pixel 8 has very bad battery life on a cellular connection. This is a hardware problem.

Did you only notice this difference (cellular vs Wifi) after installing ntfy? I wouldn't just rely on what the battery use statistics in the settings app say, because apparently they are unreliable.

thomasg

Viewpoint0232 It might well be, that the Modem is inefficient, but I don't think that's the main issue.
After all, the built-in battery monitoring shows a lot of battery usage specifically for the ntfy client.
I'm not particularly informed about how the per-app battery usage is measured exactly, but I assume it mainly consists of measured CPU usage and wake-up events, and likely won't include the draw of the cellular modem.

I can't say that I ever noticed the issue without ntfy running, but before I was using ntfy I was using matrix without Push, but the default Pull configuration, which draws a lot of power on Cellular as well as Wifi.

Before installing ntfy I wasn't aware that the phone with my typical usage could achieve anything over 1 day's use, let alone more than 3.

de0u

thomasg I talked to the (friendly) folks at the ntfy matrix channel (including the main developer), and nobody there is having the same issue

Can they provide you with some things to measure? For example, some statistics to extract from the system log about how much work the ntfy code is doing?

Perhaps (?) you are seeing more battery draw because you get more notifications per day?

Edit: or perhaps (?) your device flips between cellular and Wi-Fi more times per day?

thomasg

Apparently I can't edit, so here's another post...

I should also note, that in the battery "system" stats, WiFi and Cellular are typically in a similar range. For example, today, I'm at 41% battery usage for WiFi and 29% for Cellular (I've been home a bit more than on a typical weekday), while ntfy is listed at 35%.

bitestring

I have the same issue with my Pixel 8a. For me, the culprit seems to be IMSService. It holds a wakelock on the CPU for days and never releases it. This cause the phone to not enter deep sleep, causing severe battery drain. Restarting the phone fixes it until something triggers this bug.

Edit:
IMSService is responsible for LTE/5G communication over IP (VoIP, WiFI calls, Video calling etc.).