Skip to content

Home

Jump to bottom
Lidia edited this page May 16, 2019 · 2 revisions

How to Deploy Q-Rapids Forecasting service

Requisites

The requisites to deploy Q-Rapids Dashboard are:

  • A web server, e.g. Tomcat
  • Oracle Java JRE
  • Elastic Search
  • R Server
    1. Download and install R in the machine where the R server will be deployed. For Linux, version 3.4.4 is recommended, and for Windows, version 3.5.3. For Linux, the libraries libssl and libcurl may be needed to install the R packages, install the corresponding ones for your distro. For Windows, Rtools may be needed.
    2. (Optional step) For Linux, the installation of the "prophet" package can be speeded up installing the "rstan" binaries beforehand. RStan on Linux ( Section "Installing from the repository")

Deployment Steps

The installation package should contain 5 war files and one text file containing some R scripts:

  • qrapids-forescast-rest-X.Y.X.war
  • TimeSeriesFunctions-X.Y.X.R
  1. Deploy qrapids-forescast-rest-X.Y.X.war in your Web Server
  2. Copy TimeSeriesFunctions-X.Y.X.R to an accesible folder (RSCRIP_FOLDER).Note: You can use the APACHE version of the scripts (included in the installation package), including only one forecasting method, or use the GPL version of the scripts containing the complete set of forecasting methods, available in the forecast-R_script repository
  3. Configure R script file, setting the correponding values in the following application.properties file:
    Rserve.host=<RSERVER IP>
    Rserve.port=<RSERVER PORT>
    Rscripts.location=<RSCRIPT_FOLDER>/TimeSeriesFunctions-X.Y.X.R
  4. Installing the required R packages from an R console, depending on the .R script file you have deployed:
  5. If you deployed the .R script file (APACHE version) included in the installation package:
    install.packages("devtools")
    library(devtools)
    install_version("prophet", version="0.4")
    install_version("Rserve", version="1.7-3.1")
    install_version("elastic", version="0.8.4")
  6. If you deployed the .R script file (GPL version) from qrapids-forecast-R_script repository:
    install.packages("devtools")
    library(devtools)
    install_version("forecast", version="8.7")
    install_version("prophet", version="0.4")
    install_version("Rserve", version="1.7-3.1")
    install_version("elastic", version="0.8.4")
    install_version("forecastHybrid", version="4.2.17")
  7. Now, the R server is ready to be run from the R console: library(Rserve)
    Rserve()

By default, Rserve will start the server listening on the 6311 port. You can change it if you prefer to use another port with the parameter “port” (and setting the corresponding one in the application property Rserve.port) :
Rserve(port=6312)

Check if you have the indexes for each product you have configured for the qrapids-eval. The elastic indexes accessed by this component are indexes containing metrics and factors (see detailed information at Elasticsearch Indexes)

Clone this wiki locally