AzureML Web Services had a major new feature released, which makes this tool unnecessary. Now AzureML automatically scales your RRS web service to the level of actual load, minimizing the latency variability of the service responses.
The tool is kept here for references as a stress load generation tool.
========================================
Automatic service warm up for AzureML Web Service API
Once you create an AzureML Web Service, to keep it always ready for low latency responses follow this instruction:
-
Download and compile the code from this repository. You can just download the
RRSWarmer/AniStresser/bin
directory with the tool binaries -
This will produce
AniStresser.exe
-
Create a file named
run.cmd
with this content, replace {placeholders} with your id's and keysAniStresser.exe warm {your workspace} {workspace access token} {your service id} {endpoint name} input.json output.txt
-
Create input.json file with a sample payload for your service
-
Start run.cmd to test that the service works fine
-
Create a zip file with AniStresser.exe, run.cmd, and all other files needed to run the tool
-
Go to manage.windowsazure.com -> WEBSITES -> [create a new web site, it is only needed for the web job, you don't actually need to create an actual site]
- Go to that web site, and click WEBJOBS tab
- Create a new web job, in US South Central
- Use the zip file created above to supply the payload for the job
- Configure the job to run on SCHEDULE, every 6 hrs
- This is not an official tool from AzureML, no guarantees are given with it
- The tool makes service calls which may incur AzureML charges.