distributed computing implementations

[Irmen de Jong]

Robin Becker wrote:
> In article <3e96efac$0$49107$e4fe514c at news.xs4all.nl>, Irmen de Jong
> <irmen at -NOSPAM-REMOVE-THIS-xs4all.nl> writes
>>To add some more to the mix, I benchmarked Pyro (3.2) :
[...]
> .....was this on the same machine? I find things seem to be awful off
> the machine even if I pass in a dotted numeric address. I'm getting the
> idea that I'm doing name lookups even if there's no apparent need. 

Yep that was on the same machine. Running the Pyro server on a different
machine on my 100 Mbit lan gives:

Connecting to server
Time to connect to server, 0.024740

Sending a long string to the server
Time to send a string of 21000 chars, 0.003866

Recieving a long stirng from the server
Time to receive a string of 22000 chars, 0.003579

Sending lots of ints to the server
Time to send 5000 ints, 4.586772 (0.000917 per call)

So you see, it's even faster when the server is running on another machine. 
This can be explained because the pickling/unpickling is split across two 
CPUs, when you run client+server on a single CPU they are fighting for cycles.

--Irmen de Jong