Problem Report: Question about FETCH_PROBABILITY behavior.

[David H. <dho...@is...>]

CURL-LOADER VERSION: 0.47, December 2, 2008

HW DETAILS: CPU/S and memory are must: 4 core CPU in single physical
slot, 16636588 kB RAM

LINUX DISTRIBUTION and KERNEL (uname -r): 2.6.18-53.el5PAE

GCC VERSION (gcc -v): gcc version 4.1.2 20070626 (Red Hat 4.1.2-14)

COMPILATION AND MAKING OPTIONS (if defaults changed):

COMMAND-LINE:  curl-loader -f test.conf

CONFIGURATION-FILE (The most common source of problems):
Place the file inline here:

=-=-=-=-= Configuration without FETCH_PROBABILITY: BEGIN =-=-=-=-=-=

########### GENERAL SECTION ################################
BATCH_NAME= test
CLIENTS_NUM_MAX=1
CLIENTS_NUM_START=1
INTERFACE=eth0    
NETMASK=255.255.240.0
IP_ADDR_MIN= 10.56.1.1
IP_ADDR_MAX= 10.56.15.254
CYCLES_NUM=10
URLS_NUM=3

########### URL SECTION ####################################

URL=http://localhost:8080/one.html
URL_SHORT_NAME="one"
REQUEST_TYPE=GET
TIMER_AFTER_URL_SLEEP=200

URL=http://localhost:8080/two.html
URL_SHORT_NAME="two"
REQUEST_TYPE=GET
TIMER_AFTER_URL_SLEEP=200

URL=http://localhost:8080/three.html
URL_SHORT_NAME="three"
REQUEST_TYPE=GET
TIMER_AFTER_URL_SLEEP=200

=-=-=-=-= Configuration without FETCH_PROBABILITY:  END  =-=-=-=-=-=

=-=-=-=-= Configuration with FETCH_PROBABILITY: BEGIN =-=-=-=-=-=

########### GENERAL SECTION ################################
BATCH_NAME= test
CLIENTS_NUM_MAX=1
CLIENTS_NUM_START=1
INTERFACE=eth0    
NETMASK=255.255.240.0
IP_ADDR_MIN= 10.56.1.1
IP_ADDR_MAX= 10.56.15.254
CYCLES_NUM=10
URLS_NUM=3

########### URL SECTION ####################################

URL=http://localhost:8080/one.html
URL_SHORT_NAME="one"
REQUEST_TYPE=GET
TIMER_AFTER_URL_SLEEP=200

URL=http://localhost:8080/two.html
URL_SHORT_NAME="two"
REQUEST_TYPE=GET
TIMER_AFTER_URL_SLEEP=200
FETCH_PROBABILITY=10

URL=http://localhost:8080/three.html
URL_SHORT_NAME="three"
REQUEST_TYPE=GET
TIMER_AFTER_URL_SLEEP=200

=-=-=-=-= Configuration with FETCH_PROBABILITY:  END  =-=-=-=-=-=

DOES THE PROBLEM AFFECT:
       COMPILATION? No
       LINKING? No
       EXECUTION? Yes
       OTHER (please specify)?
Have you run $make cleanall  prior to $make ? Yes


DESCRIPTION:
Using the FETCH_PROBABILITY tag has counter-intuitive side effects.

When the configuration file for a single client to run ten cycles, for
three urls, and contains no FETCH_PROBABILITY tags, each url is
requested 10 times.

Final screen for the first configuration file:
   Operations:     Success       Failed      Timed out
  URL0:one        1      10     0      0     0      0
  URL1:two        1      10     0      0     0      0
  URL2:three      1      10     0      0     0      0

When a single FETCH_PROBABILITY with a value of "10" is introduced for
any single url-subsection, the number of requests for all URLs is
reduced and the url-subsection containing the FETCH_PROBABILITY tag is
requested more times relative to the subsections that do not contain a
tag.  In addition, the value of the FETCH_PROBABILITY does not seem to
predict the actual number of requests for the URL in that subgroup.
For, example I would expect that CYCLES_NUM=10 and FETCH_PROBABILITY=10
would result in approximately 1 request, instead I often see 4-10.

Final screen for the second configuration file w/ FETCH_PROBABILITY=10
defined for "URL1:two":
   Operations:     Success       Failed      Timed out
  URL0:one        3      3      0      0     0      0
  URL1:two        4      4      0      0     0      0
  URL2:three      1      1      0      0     0      0

The same thing happens if the other subsections are tagged with
FETCH_PROBABILITY=100.

QUESTION/ SUGGESTION/ PATCH:
Is the observed behavior the intended outcome? or should the expected
request frequencies be something like this:

  URL0: 10 requests
  URL1: ~1 request(s)
  URL2: 10 requests

for the second configuration?