Scaling up Shiny apps

Xiaochen Cai

unread,

Jul 21, 2016, 10:45:08 AM7/21/16

to Shiny - Web Framework for R

Hi,

I am building a Shiny tool which runs heavy computation job. What are the options to scale up Shiny apps? The only options I know are:

a) parallel computing: the overhead for preparing parallel is significant sometimes..
b) using C/C++/C#: has anyone tried to call C# from R on Linux?
c) faster CPU
d) lots of memory
e) physical server (I am running on a VM now)

Any suggestion would be very helpful. Thanks very much!

Joseph McMurry

unread,

Jul 26, 2016, 2:28:29 PM7/26/16

to Shiny - Web Framework for R

Isn't R single threaded?

Tareef Kawaf

unread,

Jul 28, 2016, 7:33:04 AM7/28/16

to Shiny - Web Framework for R

The core of R itself is single threaded, but there are libraries that allow for parallel execution across multiple cores/machines, and algorithms that can leverage multiple cores. The answer to Xiachen's question is really dependent on what the code is trying to do. If the code is single threaded, then a faster CPU on a physical server may make a difference, but my experience has been that the choice of libraries you use (for example there are many libraries that use Rcpp to call out C++ code) which are 100-1000x faster than their counterparts in R.

HTH

Reply all

Reply to author

Forward