SongKong Jaikoz

SongKong and Jaikoz Music Tagger Community Forum

Performance / throughput

#1

I started yesterday a SongKong-Fix-Run with SongKong Pro on a collection of 95.000 song-files (most of them being MP3s).
Now, 21 hours later, it has worked on 44.000 of them.
This gives an average of 44.000 / 21 h = ~35 / minute.
So it takes nearly 2 seconds per file.
I looked up typical performance-metrics on the PC:
~ 20 % CPU taken by SongKong
~ 6 MB/sec Diskactivity overall (and 99% coming from SongKong)
~ 2.4 GB memory taken by SongKong (still 6 GB of 32 GB available from OS)

I see that it amkes a lot of calls to different servers:

Is that “normal” throughput? What are other users of “Pro”-version get? 2 Seconds per file seems somewhat “slow” for me (especially because I read that the main issue is the MusicBrainz limit with 1 call /second, but on th other hand the Pro-Version says “twice the rate of Standard” (so that would mean 4 seconds per file).

#2

The first you run through any songs through SongKong it generates an acoustid for each song (this only ever as to be done once) which is quite cpu intensive and then looks it up potential matches from Acoustid.

MusicBrainz Lookup is quicker than 1 lookup per second but often multiple different lookups need to be done depending on the resuls because its by no means a simple lookup musicbraanz song and get the right answer.

… and there are Discogs lookups as well.

Compare how long this has taken with trying to do it yourself manually and I think its pretty fast actually, especially since you can just run it overnight when not using the computer yourself. You can also pause it or stop it and run it against the remainder later since it will skip all the already matched songs

#3

Hi Paul,
thank you for your quick response on that.
I didn’t want to anger you - sorry if it sounds like that (english is not my natural language; so sometimes i maybe find the wrong words). My intention for the question was more in the direction: Is that a throughput which is typical? Or can I get it even faster somehow (more memory, …)?
You’re absolutly right: Compared to manaully tagging my files (or correct them, which makes it even worse), this is really, really fast.

So I do understand from your answer, that this is quite good behaviour and I needn’t to tweak anything, right?

Regards

Dietmar

#4

Hi Dieter no problem, but yes I would say this is good throughput.

#5

Hello Paul,
my fix-run has now finished (at ~ 36 songs/minute).
But I do not really understand the counters in the overview of the “Fix Songs” window (see attachment with screenshot):
It has loaded 94149 songs, generated fingerprints for them, but it has saved 94212 - that are 63 more than it has loaded. And it says “Fertiggestellt” (ready) are 94710 - that are 561 more than loaded.
What does this mean / how can this be explained ?

By the way: Some translations are missing / not shown correct (its german at me). I marked them in the screenshot.
“Creating Report” should be “Erstelle Report”
“Songs saved” should be “Songs gespeichert”
(not marked, but also not translated):
“Started at” should be “Gestartet am”
“File” (in the menu of SongKong) should be “Datei”
“Action” should be “Aktion”
There are also some typos.
So if you like, I can translate/fix the errors; maybe you just have a ressource file or alike for it.

#6

Hi, I think this is a bug over counting something

Thanks for your notes about translations, I do indeed have a resource file and a procedure for modifying it please take a look at http://www.jthink.net/jaikozforum/posts/list/7278.page