On Tue, Jan 09, 2007 at 07:06:17PM +0530, R. Ramkumar wrote:
Theoretically, the end throughput could depend on whether the
application is singly or multiply threaded, if the bottleneck happens
to be end host processing rather than the network. In that case, data
from the net needs to wait while the existing data is processing (say
written to disk). Such issues become important especially in LAN's.
For example, microdc, a command line direct connect client happens to
be much faster than others on the IITM lan because of its highly
threaded design. (Well... at the expense of other apps in the system,
I am not getting into issues of multithreading really being effective
only if there are multiple processors and stuff like that...)
