Clever Geek Handbook

Understanding bokeh server

I cannot find any mention of how many sessions bokeh-server

can handle.

I would like to include some plots in my web application and would like to get an idea of ​​how a single bokeh server will handle my ~ 100 user traffic at any given time. Each user page can have up to 10 bokeh. I would use redis

as a backend

My stack looks like this (all on one VPS core, 1G RAM):

  • nginx

    (web server)
  • uwsgi

    (applications server)
  • flask

    (web framework)
  • redis

    (save data in memory)

How is the config option bokeh-server

--multi-user

reproduced in my use case? I am having trouble understanding the volume of the bokeh session

.

+3


source to share


1 answer


IMPORTANT . The question is above, and the answer below is about the old, long gone, first generation Bokeh server, before Bokeh 0.11

. For information on using the current Bokeh server, which is more stable, better, and easier to use and better documented, see:

http://bokeh.pydata.org/en/latest/docs/user_guide/server.html







a few thoughts:

regarding loading - unknown, but it's not so much the number of users as the amount of your data, since most of the overhead is json serialization / deserialization. One user can load bokeh server if json content is gigantic. But under normal use I would expect 100 users to be no problem

Note. If you are only using one core, I will not help much nginx.

regaring multi user - This means different users can register with their own username and password. This means that users will not stomp on each other with documents. In the case of a single user, the bokeh session always connects to the bokeh server as the user's "defaultuser", in the multiuser case, users must register and log into the session using their credentials. "multiplayer" is more important when users are posting content, since (IIUC) you are only one click of the content on the server, this shouldn't be a problem.

+5


source






More articles:


All Articles