[OSM-Devserver] 100 gzip-Prozesse

[Nic Roets]

Ok there were a few (10) zombie processes from a smaller test that I killed.

One osmosis writing to 190 files vs 190 gzips writing to one file each:
Fortunately the free RAM is now much larger, and Linux can buffer the output
for longer, so it can write more after each seek. So it's less seeks.

Linux is quite good at switching between task that communicate by pipes. 12
years ago I did a test and on the same hardware Linux pipes were 5 or 10
times more efficient than Windows messages. With all the multicore and
multiprocessor computers on the market I'm sure a lot of work has been done
to make it efficient so that Linux developers are encouraged to split
processes where possible.

On Tue, Feb 9, 2010 at 7:32 PM, Jochen Topf <jochen at remote.org> wrote:

> On Tue, Feb 09, 2010 at 07:26:01PM +0200, Nic Roets wrote:
> > The gzip's are mine. They're niced and ioniced, so I hope they did not
> cause
> > any problems.
>
> They still need memory and its probably really unefficient to run so many
> at the same time, because they will continually step on each others toes,
> especially when the disk have to seek back and forth between the places
> where they are reading and/or writing the data for all those processes.
>
> Now there is only one gzip but many 惡at愀 running? They don愒 look like
> they are connected to anything.
>
> Jochen
> --
> Jochen Topf  jochen at remote.org  http://www.remote.org/jochen/ +49-721-388298
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.openstreetmap.de/pipermail/devserver/attachments/20100209/08eac18d/attachment.htm>