J
jlukar
Hi,
I have a load tester that connects to the server and sends through
1000's of messages and using the returned message from the server
calculates the round-trip numbers (milliseconds).
One implementation of the server has bottleneck. Another
implementation does not have as big of a bottleneck.
However the slower server simply is slower to accept the messages while
it is doing other things and hence the round trip numbers do not show
this fact.
It is like trying to stuff some rice grains down a narrow pipe.
Can anyone tell me what is the low level behaviour of writing to the
socket OutputStream ? if my loader is sending more messages than the
server can accept, will the message linger around the network layer
(possibly on the file handler behind the socket) untill the serve can
accept it ?
If this is not the case and my loader blocks untill server accepts,
then that can be misleading using the round trip approach.
I have a load tester that connects to the server and sends through
1000's of messages and using the returned message from the server
calculates the round-trip numbers (milliseconds).
One implementation of the server has bottleneck. Another
implementation does not have as big of a bottleneck.
However the slower server simply is slower to accept the messages while
it is doing other things and hence the round trip numbers do not show
this fact.
It is like trying to stuff some rice grains down a narrow pipe.
Can anyone tell me what is the low level behaviour of writing to the
socket OutputStream ? if my loader is sending more messages than the
server can accept, will the message linger around the network layer
(possibly on the file handler behind the socket) untill the serve can
accept it ?
If this is not the case and my loader blocks untill server accepts,
then that can be misleading using the round trip approach.