distributed computing implementations

[Bengt Richter]

On Fri, 11 Apr 2003 23:04:01 +0200, Irmen de Jong <irmen at -NOSPAM-REMOVETHIS-xs4all.nl> wrote:

>Robin Becker wrote:
>
>> sigh :(
>> for real distributed computing I wouldn't want to hold the sockets open.
>
>Sorry if I missed that requirement. I was just responding to Duncan's
>CORBA timings...
>
>> I just tried the original server.py/time-client.py pair again this time
>> on a freeBSD system
>> 
>> freeBSD-->localhost 500 secs
>> freeBSD-->192.168.0.3 50 secs (a local win32 machine).
>> 
>> 
>> I guess I'm just not doing what I think I'm doing.
>
>I really can't explain these numbers either? Localhost is 10 times slower
>than over the network???

IIRC, there was some discussion some time ago about this being a particular
FreeBSD pessimality problem that wasn't true of Linux, but someone was fixing it?
But FreeBSD has good networking reputation I thought, so this needs double check,
for sure. Maybe it was for an old version?

>
>Unless your client & server both struggle for CPU cycles on the same machine 
>at the same time...
Or the struggle involves unnecessary scheduling waits of some kind?

Regards,
Bengt Richter