I think you can probably break down users into multiple categories:
1) All Users - Anyone with an account on Tableau Server.
2) Active Users - Anyone who actively uses Tableau Server.
3) Concurrent Users - Number of people who use Tableau Server at the same time. This changes throughout the month/week/day, of course.
Full disclosure, I'm not a Tableau Server expert, but I think that concurrent users are the most important metric. So you'll probably want to get an idea of the number of concurrent users during normal times as well as the number of concurrent users during peak times. Do you have any idea what those numbers are?
1 of 1 people found this helpful
The actual answer is "it depends"
It depends on ...
- how many concurrent sessions (as you & Ken have mentioned)
- how "hard" are the users hitting the server
- how complex are your workbooks
- how many different workbooks you have
and more technically ...
- how much RAM on the server
- how much disk space
- how fast are the CPUs
- data is live connection or extracts
- network latency between database server and tableau server
And it all boils down to ...
- what is the acceptable response time when a user opens a workbook, navigates dashboards, changes a filter or parameter ?
Almost 5000 users on an 8 core sounds like a LOT!, but if they are spread around the world and have light usage on Tableau throughout the day and dashboards aren't really complex it "might" be OK.
From memory there is documentation that Tableau can provide as a guide, but you use things like TabJolt to test it. Spin up a non-prod instance of your Tableau Server and see how it performs when you have 10, 100, 500, 1000 users working simultaneously. Based on those results you can get the answer you're after
Much better answer than mine!
It's just experience I've done this with 2 big clients over the years. One I said that they could stay on 8 core for a very long time. One I said they could "probably" stay on 8 core for the next 12 months then they might need to upgrade. They presented at a local TUG about 14 months later saying they were really starting to see the slow-down in their 8 core and were considering going to 16 core (that's a big investment!!).
It's not a perfect science, because things like "dashboard complexity" can change very quickly. But TabJolt can help as well
I'm actually in the middle of something similar right now at work. Spent all day working with TabJolt. It was a blast!!