When the rumours started to get legs about this i figured it would be a very welcome function, making sure maintenance was performed when it didn’t interfere with playback etc.
First of all, the KB articles regarding background work isn’t reflecting the facts around this.
When new music is added to your library, Roon immediately identifies and analyzes each file to understand its audio characteristics and contextualize it in your library.
No, that is not true. If maintenance is set to scheduled, analysis will NOT happen until the slot takes place.
Now, what is bothersome is the fact that processor throttling doesn’t seem to affect scheduled background work, why this omission?
The second issue i have is that it seems truly obsessive to burn lots and lots of CPU cycles each night for something that have VERY LITTLE meaning for the library experience.
Is is generally a waste of power the way this is done, as 99,5% of my library is untouched and still we reiterate every single night.
(My library is untouched since about a week, and yet we spend huge amounts of CPU cycles every night…
Please revisit this function and lets make it less intrusive. I couldnt care less if i had a button to do this on demand, which is actually what i do.
(Non scheduled, background work set to “limited”)
As it is designed now, it makes very little sense.
And please, let’s make the scheduling option possible in a 24h time format, or are we back in the 60’s?
The context might be a bit unclear here, but as can be seen i run Windows (11 24H2 IoT Edt LTSC) on a fanless/headless dedicated server.
Media is mainly local, with about 11Tb of files (15K albums, 230K tracks, majority in RedBook FLAC ~ 12.500 albums, rest is 50%HighRez PCM and 50%DSD) and a small Qobuz favorite section (around a 100 albums)
Around 600 albums are unidentified.
This doesn’t make sense to me, at least not on a dedicated server. Can you clarify why? Maybe I’m missing something.
I’m thinking:
By definition, I don’t use Roon when the background work is scheduled, so surely it can use whatever resources it needs? (Chances are that it’s scheduled during a time when I’m asleep)
By combining a schedule and throttling, the chances that it can’t finish the work during the scheduled window increases.
It seems to work intensively for about 3hours each night, and i neither want to pay for more electricity than needed nor need to worry about unnecessary heat being generated.
As far is i can tell, absolutely nothing has been accomplished during these three (ish) hours each night. For what i know, Roon could be selling CPU cycles to bitcoin mining?
No logs are presented or any feedback “congrats! 3 albums was identified last night!”
Basically, i dont want my library maintained or fettled with after initial identification and analysis. The architecture should be changed and unidentified albums could be checked (cant possibly take three hours burning CPU) and changed identified albums could be notified upon request by each Roon Server.
It’s not good architechture to re-identify the complete library each night (or whatever it’s doing for three-four hours).
I can see the point, but this would have to be achieved by having to do less work. Assuming a constant amount of work to do, the same energy is used whether it’s throttled or not, the difference is only the time it takes. Previously, it did the same work spread out over various points in 24 hours.
That’s yet another different wish though. I definitely do want new credits being processed etc. I assume most users want Roon to be Roon and identify stuff, link metadata, etc.
What I do wish for is a button for “don’t try to identify” because my homespun techno mixes won’t be on MusicBrainz. And maybe this could go along with a property to say that an album is finished and nothing more can/should be found.
Something like this would indeed be nice.
I think we first need to know what it is that it’s doing.
Your reflections are welcome, and i like the idea of an “accepted” flag when one is not interested in retrying identification or basically don’t want the metadata updated (when one has edited or groomed the metadata).
I would like to set that flag for my entire library of identified albums basically… And a few of those odd albums which are right as rain, but cannot be found in the online databases.
And if this was possible, less work would be needed to reiterate every night.
Perhaps worth noting that edited parts already are supposed to remain untouched:
How editing impacts future metadata improvements
Before making an edit, it is important to understand how this will impact future metadata updates. Our metadata is always improving, and those improvements are automatically added to your database. However, if you make an edit, some of these improvements may not automatically be added in the future.
Editing an album will freeze its structure, meaning we won’t automatically reorganize the track/album relationship in the future as our algorithm improves
Editing a field prevents future updates to that field in order to maintain your edits
Using “identify album” pins tracks/albums to a specific release forever, so even if we come up with a better one because of a new data source, or new data from one of our sources, you continue to have the one you manually selected
It’s perhaps not an entirely simple problem. If I edited a bunch credits, do I want other credits in the same album to become static, even if there is something new to add? Probably yes. Maybe this could be left to for the user to decide (but comes at a cost of potential user confusion when things don’t update, and it may not make all that big of a difference)
I guess you have convinced me even more that I want to know what exactly it’s doing for hours every night before I form an opinion
According to previous statements, “prefer” does not mean “use only”, just to, well, prefer it if both local and online metadata is available. So I doubt that this means that it won’t check for other metadata online.
That’s a shame it does not leave the albums alone and stop trying to identify. I only have 37 unidentified albums, but these will never ever be identified.
I can imagine people with tons of bootlegs could potentially clog up their queries.
Crossing fingers that future updates gives us some more advanced options to the nightly scheduled updates. Especially if it’s a system with sub-par cpu/RAM
Yeah, it won’t matter for 37 in terms of work or speed, but it’s a very serious problem for those who have many hundreds or thousands, and it could be quite easily fixed, one imagines.
I didn’t say, but this is not the case here as the server is an okay machine for number crunching, Intel Core i5 11400, 32Gb DDR 3200Mhz and a good Samsung 250Gb NVMe m.2 system drive. No throttling by OS or setup, so in the daytime when playing non upsampled music it’s idling away at around 1.2Ghz. When i’m stressing it by upsampling or updating it boosts a few cores into the 4Ghz territory.
When i have restructured my library or added lots of new music i generally allow it to use 2-3Cores (out of the 12) for identification and analysis and that generally scores as a total usage of around 15-20% and pushes the CPU temps up to around 50°-55°C after a couple of hours. In daily use the temp stays at 40°C
This isn’t critical, but i’d like to know what Roon is spending all of these CPU cycles on, and basically “what’s in it for me?”
Like Marian, the facts of Performance Monitor each day since a few weeks has peaked my interest in either getting to know what is actually being done. But until (if?) that is revealed i’d like to be in control of what my 24/7 machinery is doing!
