[vdr] Re: 18 Minutes - what does that tell?

["Richard Cieply" <CIEPLY@de.ibm.com>]

Alessio Sangali wrote:

>ja isn't noflushd supposed to keep the data in memory and write only
>when the memory is full?

yes of course it is, but I was also experimenting with hdparm
and then a ramdisk is helpfull. And was using noflushd with
ramdisk still on, that's all.

>> Another problem I had with nofushd was that it tried to store the data
>> in (virtual) memory (OK, thats what it is supposed to do :-)) and then
>> tried to flush a huge amount of data to the disk. Of course the disk
>> was sleeping. During the spinup and flush data, often it looked like
>> buffers would overflow resulting in dropouts.
>> It seems that noflushd could not receive new data (from vdr) while
>> flushing old ones - or at least not fast enough.

>which kernel version were you using?

..., 2.4.10, 2.4.16

>> Had anybody a similar problem or is somebody successfully driving a
>> vdr box with noflushd?
>> Maybe it's a good idea to stop noflushd on start of a recording and
>> start it again when all recordings are finished?

>very 'unclean', noflushd should work transparently, and vdr shouldn't
>take care of calling external programs... however if it's the only
>solution...

yes and no, because of the nature that there is a huge amount to write,
and as the postings say, there are problems running noflushd all the
time...

>btw, with ibm hdds it's possible to download a software that lets you
>set the noise/performance ratio of the hard drive: seeks become
>ABSOLUTELY impossible to hear; it's also possible to set a kind of spin
>down of the drive but I don't know if it's useful...

Nearly all new HDDs are capable of this.
Yes, it reduces the 'seek noise' but however the disk still spins
making noise, heat, power consumption.
You can use hdparm to spin disks down (and a ramdisk to store epg data :-))
but this is not good enough because of too many wakeup events like
messsages.
So noflushd is the way to go.

regards
  Richard