…percentile in stats.py to be used in the start_ramp function in core.py.

…nted start_ramp code

start_ramp is now woking; it's ramping up exceeding a boundery then gradually getting closer, splitting the max distance in half til "hatch_stride" is lower than "precision"

… tab and Failures tab.

The sorting is not rendered until the next updateStats call though (so once every 2 seconds)

"current_percentile" calculates the response time percentile for the requests in a time window.
It is implemented to work in both distributed and local mode.
autotune.py listens to the events on_slave_report, on_report_to_master and on_request_success and handles a deque of the latest response times TIME_WINDOW seconds back in time.

Added variable reset to start_ramping, if reset is set to true all currently spawned clients will be killed
Added variable start_count; start_count is the number of clients that will be spawned before starting the ramping

…on current_percentile since all this functionality now is in autotune.py

removed <a> tag element on the labels and use the <th> instead

… when sorting by columns

fixed a check of the key "current_reponses" in data for the event on_slave_report so that the list of responsetimes is only appended if the key exists

…ponse times for being used in the start_ramping function

…column

…is because of that the kill_locusts name fits better if the function adding clients is named spawn_locusts

Fixed so that the ramp down will never kill less than "precision" amount of clients

…forgot to refactor

… ramp functionality

… used; this to make the UI for ramp optional as well

Changed "/ramp" route to fit with the ramp form in the web UI

…ge calibration times during the ramping

…his I made the start_ramping spawn in a separate greenlet

added "calibration_time" to the ramp form in the interface

…fter a test has already been run (the ramp_text should only be shown on the index page when no test has yet been run)

…ad of hatch_rate) in start_ramping when spawning the initial number of clients

… so now I fixed htem

The eventlistensers are only created if --ramp is used

removed unused calibration code

… comments

…rectly, fixed it

Changed so that RequestStats property fail_ratio is used in web.py instead of caluculating the fail ratio there

…ter/slaves when usering the --ramp option

…w only done once and the result put in a variable in rampstats.py

…ptable fail ratio

Before there was a problem if the acceptable fail ratio was 0

… up now