SiteBlaster v3.0 beta is a multi-threaded application that simulates
the actions of many users; each user operating on its own thread.
There can be as many as 500 users; each operating on separate threads;
each working without knowledge of any other user.
During a test cycle, each user keeps measurements of the actual length
of time it takes to test each URL. Actual length of time is the
duration from the time the call is placed to access the URL until the
complete response is received.
When all testing has been completed, the length of time it took each
user to test the same URL is summmed and divided by the total number
of tests of that URL to produce the average length of time it took to
test the URL (i.e. response time). The reciprocal of average response
time (secs/test) is average thruput (tests/sec). These results are
reported on a per URL basis.
Similar calculations are made to determine the average user response
time and thruput but INDEPENDENT of url.
SiteBlaster v3.0 allows you to have a fixed or random delay between
each test performed by a user. This delay is NOT factored into the
user response time/thruput calculations. However, they are factored
into the system response time/thruput calculations. Therefore, the
system response time is the time duration from the time the first user
started testing and the last user finished testing divided by the
total number of tests performed by all users. Also, there are certain
inefficiencies that the OS experiences as it switches between
threads. These inefficiencies affect the response time/thruput at the
system level and are definitely reflected in the sytem calculations.
However, the extent to which they play a role at the user level is
Hopefully this helps.
On Aug 16, 4:14 am, Henry <henry...@gmail.com> wrote: