After playing around with R studio server for a while, I decided to write a followup to my previous blog post. I want to go over a few of the strong points of using RStudio server to access a remote machine using R.
Advantages of RStudio server
Before discovering RStudio Server, I would use either plain ssh with X enabled, or NoMachine NX to access a remote machine. In contrast to plain ssh, RStudio server ahs the advantage that it does not require a running X server at the host. This is mainly a problem under Windows, where you need to install Cygwin to get the X server running (assuming using putty for the ssh connection). Although it works, I don’t really like installing a lot of additional software just to get a remote connection with graphic support. Under Mac things work a little bit more out-of-the-box, but natively Mac OS X does not run X, and it fires up a seperate program where all X stuff happens. This also does not feel ideal, although it is much better than using windows. RStudio server does not require an X session, just a good web browser. Plots and pdfs can also be shown inside the browser.
Using a tool like NoMachine only requires a small client application, in addition to a server-side installation of NoMachine. This works quit well, although performance across firewalls can become slow. Although I have not used RStudio server across heavy firewalls, I suspect it performs a little better.
When being trapped behind a corporate firewall, the only traffic allowed to leave the firewall is over a proxy that only allows traffic over port 80. NoMachine and ssh can be used through tunnels, but RStudio Server is purely based on web traffic, and I think this might work well.