[Paranoia] Parallel ripping causes extreme slowdowns
Johannes Bauer
dfnsonfsduifb at gmx.de
Sun Mar 17 18:36:10 UTC 2019
Hi there,
I had a while ago built a ripmachine on a really old, crappy i686 PC
with a couple of IDE CDROM drives and some USB/IDE converters. I did
parallel ripping with it and it worked generally fine with 8 CD drives.
Now I've rebuilt the whole thing using a newer PC, newer kernel,
everything new and am using cdparanoia 10.2 to do the ripping. I have
three SATA DVD devices in parallel right now on two different SATA
controllers (two on one controller and the third on another). All drives
pass the cdparanoia tests.
When I rip with one drive, it works perfectly, the CDs don't exhibit any
errors and the drive is on full speed most of time time (it's audibly
louder than when it spins down). I can rip with one drive, a HL-DT-ST
DVD-RAM GH22NS30, I get a decent speed of ~7 minutes for ~350000
sectors, i.e., around 1.9 MB/s or an average speed of 12.8x. Full
paranoia mode is used.
Now here's the strange thing I'm seeing: As soon as I try to have more
than one drive in parallel, speed hits rock bottom. Drives are spinning
up (that takes significant time, maybe 5-10 seconds), then as soon as
data arrives they somehow stall and quickly spin down again. Rinse,
repeat. With three drives I practically get zero throughput at all
anymore, all three drives are constantly spinning up and down.
I can exclude physical interference as a root cause as I have, for
testing purposes, dislodged all three drives from the enclosure and put
them on vibration isolating foam. I'm working on a machine with decent
power (compared to the old i686 cucumber), it's a Core(TM)2 Duo CPU
E8200 @ 2.66GHz with 8 GB of RAM running a 4.18 kernel: Linux ripmachine
4.18.0-13-generic #14-Ubuntu SMP Wed Dec 5 09:04:24 UTC 2018 x86_64
x86_64 x86_64 GNU/Linux
Of course I've checked the FAQ, but my question doesn't seem to be
covered there. If there's anything I could try to narrow down the issue
(some procfs switches or buffer sizes to configure maybe? changing
something within the cdparanoia source to test a hypothesis?) please let
me know and I'll gladly try it out. I'm super desparate, this is really
annoying me and I cannot figure out for the life of me what's wrong.
All the best,
Joe
More information about the Paranoia
mailing list