I am wondering if my db is somehow fragmented or messy in some way, whether this is at all possible, and whether there is a way to clean it up.
I have run the “library cleanup”. All counts are 0 there.
If the library has been cleaned up, the counts will be zero. Were you expecting something else?
What I meant is whether a db that has no issues could still be suboptimal and cause slow startup or slow Roon performance.
The quality of the metadata and the name/folder structure has an influence, according to my observations.
In what ways?
It could if not designed optimally such as missing indexes that force full table scans. Not sure if this applies to the type of db used in Roon.
Only way to find out is to start with a fresh DB (backup first of course) and see how that performs. Personally I would start with a small subset of your music (say 10% of what you have) initially to baseline what maximum performance looks like with your storage and network/internet configuration.
I curate my library in a different computer, where the “main” library sits - and it is extensively backed up. The file types are AIFF for PCM, DSF for DSD, a few FLACs for purchased MQA content, and a few other types (but very few of those).
The directory structure is:
[Source]/[Album Artist]/[Album]/
“Source” is where I got the album from: CD (ripped); SACD (ripped); HDTracks; Acoustic Sounds; Bandcamp; etc,
The metadata is curated carefully with Yate. All files have album art, album, album artist and artist, etc - they all show well in Audirvana and JRiver for example.
This is a library built over the last 10 years give or take, having ripped my CDs over a period of say 2-3 years, and having added a lot of Tidal AND Qobuz albums.
I think one peculiarity is I have a lot of albums grouped into one - meaning either different versions locally and/or Tidal and Qobuz versions. For all of my favorite albums I have both local, Tidal, and Qobuz versions. This might be unusual - if Roon didn’t have the “Group Alternate Versions” of albums feature it would be a lot less useful to me.
I could consider starting a db from scratch - the problem is I rely on my customizations heavily - and there are a ton - most singnificantly the “group together”. Roon does a TERRIBLE job of identifying same albums, so I almost always need to curate this. If Roon took track number, some possible discrepancies in track lenghts, and some form of neural network based cover art comparison it would almost perfectly group same albums. But it doesn’t.
I originally started with Tidal and when Qobuz came along I intended to migrate to Qobuz and drop Tidal. The problem is I have found many titles on Tidal that are not on Qobuz so that never happened.
Anyway… I digress.
I would like to understand what affects performance so much in my case. Audirvana is blazing fast with this same library. Yes, there’s little to no customization of library there, but still.
I got an overspec’ed NUC to make this run well and some things are better some ok.
It will be slow if you booting up as explained in the other thread. Your not going to improve on that. Are you still finding it slow after it’s all up and running? Did you check to see if your library is still doing anything in the background you will see this is in Library section as that will slow it down if for some reasons it’s re analysing your collection?
Comparing Roon to audivarna is like comparing an apple to a banana. They don’t work in the same way. Streaming is separate and not part of your local library it’s just API calls in Audivarna not so in Roon. It’s all part of your local database once added. . It also doesn’t have the same level of metadata and organisational data your adding.
@miguelito I have several computers, because even after 20 years, none of them has ever broken down on me. All have served to manage my library with MacOS, Windows XP, 7, 8, 10, 11 and Linux (Ubuntu, Mint and Manjaro). I started analog in the late 60s and digital a bit before the turn of the millennium. I’ve only known Roon since 2020, but have been “playing” with it intensively to see if it can fully replace my Foobar2000. In August 2022 I am again faced with the question renew?
My experiences range from frustration to fascination. Today I think there is nothing that is better, but also nothing that is already perfect. The integration of two music services meets my needs. It is no longer a side by side, but a database. I think 5% local and 95% via services is a good ratio to discover more.
I don’t use AIFF for PCM, DSF for DSD, MQA and I don’t want to be dependent on one vendor. OGG, MP3 or FLAC up to 96 kiloheart go on my old ears. I just sorted out OGG because Roon doesn’t reliably recognize the embedded album art.
I rip from CD, digitize vinyl, cut radio or legal sources, buy at concerts, flea markets, Amazon, Bandcamp, Qobuz… but no SACD, HDTracks… I lack the good ear for it.
I would also argue that my metadata is mostly carefully curated. I currently maintain the following fields:
filename,H,Created,Comment,Rating,Composer,url,title,urla,artist,album,albumartist,date,discnumber,tracknumber,expl,genre,label,Copyright,imagecd,imagea,Totaldiscs,Totaltracks.
For Roon filename (path),Composer,title,artist,album,albumartist,date,discnumber,tracknumber,genre,label,Copyright,Totaldiscs,Totaltracks would be helpful in safe and fast allocation
I have embedded album art 99.9% of the time and artist images 10% of the time. Album covers are recognized in almost every application and also in Roon (except OGG) and artist images are only recognized as an image in the album folder additionally (like album covers).
I now name my files and folders consistently according to this scheme:
$cut(%album artist%,50)/$cut(%album%,80) [’[’%date%’]’ ]%H%/$num(%discnumber%,2)%tracknumber%. [$cut(%artist%,50) - ]$cut(%title%,50)
This leads to structures like this:
P:\G\Groove Coverage\Greatest Hits (Groove Coverage) [2007] X
0101. groove coverage - Because I Love You (radio edit)
The shortening of the metadata to 50 or 80 characters is necessary to avoid overloading the file system. The letter at the folder end ranges from A to Z for differentiation (source, version, other). So it is a helpful hint for Roon to create more than one album.
However, I also made the mistake of throwing everything into one pot (folder). The effect was that one drive out of 7 dramatically dropped in performance.
The problem is aggravated when again and again also parts of the metadata are missing for ancient collections or the alignment is simply more difficult…
https://musicbrainz.org/release-group/38c567dd-de37-34ce-90bb-af6dc16a8275
9 Folder if all is there
Startup is slow but usage performance once up and running is great, a big leap forward compared to my mac mini.
Every now and then there’s a slowdown but I can easily chuck that to search being slow as the servers on the Roon side might be busy.
I am giving it some time to declare my woes gone with the move to the NUC, but so far it looks good.
1 Like
They have been a bit rubbish since the updates last week and I dont think they have sorted that yet.
My expectation for an end-user product is that any indexing that gets screwed up would be automatically rebuilt at some point.