When plot with R (or some other software), it doesn’t bother you a lot for choosing colors if you have only several data series. For example, if I want to create a figure for 5 data series, I can easily choose red, green, dark green, blue and black. But what if I need to create a figure having 100 data series? Of course, we can choose different shape or pattern, like dash line, circle, triage, etc. However, to me, I feel that make the figure messy, I prefer all series in same style, except their colors. I created a Shiny-app which can generate random distinct colors in a certain color plate, would like to share it with public.
The Shiny-app on open-source edition Shiny-server has only one concurrent, which means it can run only for one user at a time point. But it can host multiple Shiny-apps, which can run synchronously. So, if we create severl Shiny-apps with different names but same function, then we can let more users use our service at same time. But users don’t how to choose the Shiny-app with small user number. This post will show you how to create a Shiny-app to redirect user to the Shiny-app with lower load.
RStudio supplies several servers for hosting user’s apps, e.g. http://www.shinyapps.io/, http://spark.rstudio.com, http://glimmer.rstudio.com. Thanks a lot to RStudio who has contributed many excellent R libraries! I want to set my shiny-server on a virtual private server (VPS), because I have some other tasks and I feel a VPS is more stable than the PC under my desk.
If you have deployed your Shiny-app on internet, you may curious about: how many users are using my app? Is the server powerful enough for hosting the app? You can get answers through the Server monitoring feature if you are using the Professional edition of Shiny server. What if we are using the Open source edition? Officially, there is no such a feature, but we can create it ourselves using Linux shell commands.
Shiny-server integrates the statistic power of R with web, convenients the statistical analysis through web. Users can do complecate statistical anlysis through their web browsers. However, most of statistical analysis of R are CPU bound computation, that means CPU utilization is high, perhaps at 100% for many seconds. This leads to the problem of thread blocking: if one user is doing some calculation, the other user must wait until that process finished! For example, if one calculation needs 10 seconds to finish, and 4 users are doing the same calculation at same time, one user may need 10 seconds to get the result, and the unlucky one may need to wait 40 seconds! That’s annoying, but the native character of statistical analysis and R (R is single-thread: it uses only one core of your CPU).
In a previous blog, I post a simple authentication method for Shiny-app, and received several comments mainly concerning that I should encrypt user password. I agree, user’s password can be intercepted when it was transferring. To secure users’ personal information, I think we should consider both server and client sides.
A month ago, I attended a friend’s wedding as a photographer (I am not good at shooting, just have a not so bad camera). I took more than 1000 pictures in high quality format. Later, I relized to share them with the guests attended wedding is a challenge: I need to go through every picture to identify if it contains the guest required pictures. After several rounds, I have to stop and try to find an efficient method.