PDA

View Full Version : 6.0 performance results



Free Lunch
2005-03-16, 15:59
Seeing 6.0 linked off the main page, I expected great things and gave it a try.

The linked version would not run on my gentoo system due to perl
issues. The latest nightly, v2005-03-16 (6.0b1), did.

Performance was unexpectedly much worse than 5.4.1 on my 512MB/1200
Mhz dedicated Sempron Gentoo system (using an original slimp3).

Is the database support not working yet? Of course I waited until the
server was done building the database, etc. I reviewed the server
settings and didn't see anything related.

Is this the expected performance at this time?

Browse music dir with 816 sub-dirs, 6.X:

procs memory swap io system cpu
r b w swpd free buff cache si so bi bo in cs us sy id
0 0 0 180 3496 57780 280592 0 0 0 0 1184 125 6 1 93
0 0 0 180 3496 57780 280592 0 0 0 0 1180 125 5 0 95
1 0 0 180 3432 57788 280720 0 0 128 12 1149 66 59 2 39
1 0 0 180 3432 57788 280720 0 0 0 0 1184 8 99 1 0
1 0 0 180 3432 57788 280720 0 0 0 0 1183 3 98 2 0
1 0 0 180 3432 57788 280720 0 0 0 0 1110 9 99 1 0
1 0 0 180 3432 57788 280720 0 0 0 0 1107 3 99 1 0
1 0 0 180 3432 57788 280720 0 0 0 0 1107 7 99 1 0
1 0 0 180 3432 57796 280720 0 0 0 12 1109 11 100 0 0
1 0 0 180 3432 57796 280720 0 0 0 0 1186 90 91 0 9
0 0 0 180 3304 57796 280848 0 0 128 0 1220 156 7 1 92
0 0 0 180 3304 57796 280848 0 0 0 0 1184 123 6 0 94

Each time, the music would be interrupted for approximately 3/4 of a second.

5.4.1:

0 0 0 180 3360 53604 275024 0 0 0 0 1208 161 9 1 90
0 0 0 180 3360 53604 275024 0 0 0 0 1190 149 4 0 96
1 0 0 180 3232 53604 275152 0 0 128 0 1206 74 51 2 47
1 0 0 180 3224 53612 275152 0 0 0 12 1184 14 97 3 0
0 0 0 180 3220 53612 275152 0 0 0 0 1197 144 22 0 78
0 0 0 180 3212 53620 275152 0 0 0 12 1188 155 3 1 96

Browse music dir with 104 dirs, 6.X:

0 0 0 180 3112 57608 281244 0 0 136 0 1197 167 7 0 93
0 0 0 180 3112 57608 281244 0 0 0 0 1188 129 6 1 93
1 0 0 180 3112 57608 281244 0 0 0 0 1188 106 32 2 66
0 0 0 180 3112 57616 281244 0 0 0 12 1192 92 87 0 13
0 0 0 180 3112 57616 281244 0 0 0 0 1184 121 6 1 93
0 0 0 180 2984 57616 281372 0 0 128 0 1191 133 6 0 94

Using Fishbone, a reload of the main HTML page caused a 1.5 second
drop-out, 6.X:

0 0 0 180 4564 56420 275636 0 0 0 0 1195 150 9 0 91
0 0 0 180 4564 56420 275636 0 0 0 0 1198 157 8 0 92
1 0 0 180 4436 56420 275728 0 0 88 0 1196 156 29 1 70
1 0 0 180 4308 56420 275884 0 0 160 0 1150 32 97 1 2
1 0 0 180 4308 56428 275884 0 0 0 12 1241 14 99 1 0
1 0 0 180 3732 56428 275884 0 0 0 0 1191 3 94 6 0
0 0 0 180 3788 56224 274552 0 0 0 0 1115 12 10 2 88
0 0 0 180 3788 56232 274552 0 0 0 32 1109 9 0 0 100
0 0 0 180 3788 56232 274552 0 0 0 0 1108 5 0 0 100
0 0 0 180 3788 56232 274552 0 0 0 0 1110 7 0 0 100
0 0 0 180 3788 56240 274552 0 0 0 12 1109 11 0 0 100
0 0 0 180 3788 56240 274552 0 0 0 0 1107 5 0 0 100
0 0 0 180 3788 56240 274552 0 0 0 0 1107 5 0 0 100
0 0 0 180 3788 56240 274552 0 0 0 0 1109 3 0 0 100
0 0 0 180 3788 56240 274552 0 0 0 0 1108 7 1 1 98
0 0 0 180 3788 56240 274552 0 0 0 0 1108 3 0 0 100
1 0 0 180 5588 56240 274680 0 0 128 0 1463 10 92 1 7
1 0 0 180 4756 56240 274680 0 0 0 0 1234 3 94 6 0
1 0 0 180 5544 56128 274152 0 0 128 0 1288 9 97 3 0
1 0 0 180 6292 56128 274152 0 0 0 0 1355 7 100 0 0
1 0 0 180 6292 56128 274152 0 0 0 0 1131 5 100 0 0
1 0 0 180 6292 56136 274152 0 0 0 16 1141 15 100 0 0
1 0 0 180 6292 56136 274152 0 0 0 0 1146 5 99 1 0
1 0 0 180 5716 56136 274280 0 0 128 0 1386 86 92 1 7
0 0 0 180 5716 56136 274280 0 0 0 0 1247 170 9 1 90
0 0 0 180 5716 56136 274280 0 0 0 0 1406 131 8 0 92

Those instances of 100% idle time suggest there is some serialization
in the new code?

Same reload with 5.4.1:

0 0 0 180 4876 52316 271264 0 0 0 0 1200 153 5 0 95
0 0 0 180 4876 52316 271264 0 0 0 0 1204 169 6 1 93
0 0 0 180 4748 52324 271392 0 0 128 12 1202 169 4 1 95
1 0 0 180 4168 52324 271392 0 0 0 4 1373 63 75 1 24
0 0 0 180 4672 52324 271392 0 0 0 0 1514 156 7 0 93
0 0 0 180 4544 52324 271520 0 0 128 0 1208 173 7 0 93


Final 6.X process size:

30717 slimuser 15 0 98408 93M 3720 S 6.9 19.7 52:48 slimserver.pl

Final 5.4.1 process size:

32050 slimuser 15 0 100M 96M 3080 S 6.9 20.5 0:41 slimserver.pl


Regards,

FL

Dan Sully
2005-03-16, 16:02
* Free Lunch shaped the electrons to say...

>Seeing 6.0 linked off the main page, I expected great things and gave it a try.
>
>The linked version would not run on my gentoo system due to perl
>issues. The latest nightly, v2005-03-16 (6.0b1), did.
>
>Performance was unexpectedly much worse than 5.4.1 on my 512MB/1200
>Mhz dedicated Sempron Gentoo system (using an original slimp3).
>
>Is the database support not working yet? Of course I waited until the
>server was done building the database, etc. I reviewed the server
>settings and didn't see anything related.
>
>Is this the expected performance at this time?
>
>Browse music dir with 816 sub-dirs, 6.X:

Browsing the music directory is currently unoptimized. It hits both the DB and the filesystem.

I recommend that you use the 'Browse Artist', etc links.

>Final 6.X process size:
>
>30717 slimuser 15 0 98408 93M 3720 S 6.9 19.7 52:48 slimserver.pl
>
>Final 5.4.1 process size:
>
>32050 slimuser 15 0 100M 96M 3080 S 6.9 20.5 0:41 slimserver.pl

How large is your library?

-D
--
"You can usually recover from production flaws...but you can never recover from a bad design".