ReH can reproduce the load sequence of a website. To do that, it first asks a browser (Google Chrome as of the time of this writing) to capture all the requests and responses that make a given website, and then serves back the website. Read more in one of the sections below, "Endpoints".
There are three big components in the instrumentation server, and they match independent virtual machines. Let's use names for them: StationA, StationB and Instr. The Instr component should run the ReH, while both StationA and StationB should run Google Chrome with an extension installed. The extension can be found in a subrepository here.
-
After the checkout and while being at the project directory, install the instrumentation webserver in the following way:
$ cabal sandbox init $ cabal install -
You need a few third-party files for the installation to succeed. They should be:
- scripts/VNC-5.2.3-Linux-x64-ANY.tar.gz
- scripts/google-chrome-stable_current_amd64.deb
-
[TODO: Write a more detailed instance-setup procedure] Scripts and extensions to be distributed to the station machines:
$ fab apt-stations
That's it!
The ReH server is written in Haskell, which is a very decent functional and imperative language. There is a package manager for Haskell with also very decent dependency management, which is called Cabal. This project can be installed with Cabal.
ReH opens a control webserver at port 1070. This is an HTTP/2 service running with a configured certificate. On that webserver, ReH listens for requests from both external applications (e.g, the Django web server) and from the two machines running Chrome instances.
- /setnexturl/ POST to this uri the next URL to analyze (as the POST data). After this, ReH starts the analysis of the given URL. The result of this stage is some .har files produced by loading the website under different network conditions and protocols ( TODO: explain which protocols and network conditions, and where the files are created ).
to be written