1 Reply Latest reply on Aug 27, 2018 10:23 AM by Andrew Bickert

    Architectural recommendations  - Geographical Latency?

    Mathias Bengtsson



      I'm interested in people's experience from setting upp a Tableau Server environment serving users in separate locations (US, Europe, Asia). Today we have a single node installation on premise in Europe. In terms of data sources that we connect live to we have these replicated so that the user is always close to the data.


      Now I'm am looking for the best setup to get the best local performance, independent of where the user are (US, Europe, Asia).


      We are thinking a multi-node setup with three nodes in total (2 workers). The initial node will be in Europe and the workers will be installed in the US and Asia. Will this help our cause? Or will traffic be routed via Europe anyway when for example a user is logging in from Asia?


      The other idea we have is three separate installations for the three locations. But then the challenge would be that any update to data sources, workbooks etc would have to be carried out 3 times or somehow automatically replicated from the initial node.


      What is your experience, what good ideas can you share? Looking forward to your reply.



        • 1. Re: Architectural recommendations  - Geographical Latency?
          Andrew Bickert

          Hi Mathias,

          With your multi-node setup proposal, I am imagining you are going to have a gateway on the primary as well as the nodes. With this, are you planning on having a load balancer with different IP/URLs for the different Geographical Locals (ie: TableauUS going to the US gateway, TableauEU going to the EU gateway, etc.)


          Even with doing that though, it is not recommended to have nodes across geographically dispersed data centers as it can cause considerable issues. This is from the Tableau Distributed Requirements page: Distributed Requirements


          Distributed deployment across data centers: We do not recommend installing Tableau Server nodes across distributed data centers. The following examples describe some issues that are common when nodes are distributed between multiple data centers:

          • Disruption in network connectivity between nodes can cause many tasks to fail or for Tableau Server components to become unlicensed.
          • Proxies and firewalls between data centers may impede the ability of the Tableau Server nodes to communicate with each other.
          • Routing traffic between geographically dispersed data centers, can cause latency or bottleneck data transmission, resulting in poor performance and connection timeouts.



          With installing 3 separate servers, you would run into the issue of needing multiple server licenses for the 3 production machines which would increase overall cost considerably. The problem as you mentioned above, is keeping them in sync. Depending on your requirements, it may be possible to have a "Publishing Server" where you publish all of your workbooks to and then have the other servers refresh to this version at Midnight (or on a schedule). We do something similar to keep our TEST Machine in sync with Production (via scripting) but it would not be an 'ideal' solution. Interested to see what other ideas people have though. Are you planning on hosting this on your own Farm or going through a 3rd party provider.


          Kind Regards,


          1 of 1 people found this helpful