Speed of protocol fish

[John Berthels]

On 09/05/06, Werner Modenbach <modenbach at alc.de> wrote:
> I recognised a limitation of speed of about 1MB/sec. which is really
> unsatisfying. So I cancelled the copy process. When doing drag and drop with
> more than 10 subdirs of my data separately in parallel I recognised the same
> limit for each transfer. But the total transfer rate of all copies was about
> 10MB/sec. . So this should not be a limit of disk access, CPU power or the
> network.
> What is the reason for the limitation? Is there anything I can do here?

I don't know the fish protocol, but the usual reason for multiple
streams going faster than a single stream is latency (e.g. time for
the other end to ack the data). You can sometimes reduce the effect of
end-to-end latency on throughput by reading and writing in larger
'chunks' - is there an option regarding that?

jb