On Tue, 24 Oct 2000, Christian Varnholt wrote:
> I thought I have to work in such a way, because there may be the risk
> that I loose some data. The I/O to disc is may be slow and with the
> RT-Task I am sure the results are written at the correct time. The fifo
No, why do you think I/O to disc will be faster when started from RT task? It
would only be, if you write real time disk driver, completely dedicated to
your RT task, writing on separate disk. In real world all processes compete
for disk access. Linux buffers it a little bit, and it results in better
overall performance.
> has only the function of a buffer, if the LINUX write-task is to slow.
> What happens, if some LINUX programs use all the CPU power ? The
> RT-tasks are still working, but what happens to my LINUX-tasks ?
>
Exactly. RT-tasks are still working. If you would write to disk from them,
they would block waiting for disk as well. You can not have the pipe faster
than its slowest element.
Just increase priority of your Linux task writing to disk (even use soft
RT). Use separate disk on separate controller for collecting RT
data. Optimize it with hdparm. Use some kind of "sync" flag, either when
mounting filesystem or when opening the file. Use a few MB or more of
buffers.
Best regards,
--
Tomek
-- [rtl] ---
To unsubscribe:
echo "unsubscribe rtl" | mail [EMAIL PROTECTED] OR
echo "unsubscribe rtl <Your_email>" | mail [EMAIL PROTECTED]
---
For more information on Real-Time Linux see:
http://www.rtlinux.org/rtlinux/
