4 Replies Latest reply on Nov 25, 2013 11:40 AM by Jonathon Stewart

    Suggested # of VizQL Processes

    Jonathon Stewart

      Hey there,


      We are using Tableau Server with over 2000 users, but at a given moment we probably have about 50 or so folks viewing reports at once.  However, we have about 7 days throughout the year that the number of concurrent users spikes - to nearly 300 at one time.  This has crashed our Tableau Server on a couple of occasions.


      We have a single 8-core license running on a virtual server with 64GB RAM.  I know this article recommends 2 VizQL processes for normal traffic.  Are 2 VizQLs good enough for 300 concurrent users?  If not, should I only increase the # of VizQLs when necessary or should I just increase to 3 or 4 for daily use and be done with it?


      Lastly, we don't have a budget for additional core licenses but we could split our existing 8 cores among 2 or 3 virtual server nodes. Do you think that would help with the spike in traffic?




        • 1. Re: Suggested # of VizQL Processes
          Russell Christopher

          Hey Jonathon -


          You should defintely move to Tableau 8.1 -- specifically, the 64-bit version. You'll get much, much more mileage out of each VizQL in 64-bit land. Each process has a bigger cache, can spin up more threads and in general is just all around "awesomer". It's pretty breathtaking, actually.


          However, if you have 300 concurrent users hitting the server at the same time, it's highly unlikely that any combination of VizQLs on 8 cores (no matter how you split them up and no matter how many vizqls are running)  is going to gracefully handle the load in question. You simply need more cores -- which you said you can't do.


          That being said, go with 64-bit and 2 vizqls on your single machine - since they now can each consume much, much more RAM than in their 32-bit past you won't see crashes due to 32-bit memory address space out of memory issues like you probably were getting before when under stress. I would not split your cores up across multiple machines - scale up before you scale out. The only reason I'd consider doing this is if I HAD to have high availability. I would essentially be  (knowingly) trading peak performance for high availability in this scenario.


          We hopefully will have more "official" guidance around 8.1 64-bit mode performance available for everyone soon.


          But to reiterate, 8 cores is not enough to handle 300 concurrent users in a performant manner. If you want to deliver good performance during that spike, more cores! more cores!


          Good luck!

          1 of 1 people found this helpful
          • 2. Re: Suggested # of VizQL Processes
            Cristian Vasile



            You could try first a poor man approach to cover that 7 days ferocious users' requests, and save vast majority of the visualizations/dashboards as .PDF files and deliver them by email.


            If you have enough IT resources to test and explore, then you could try to split that 8 cores X1VM as 4 cores X2VMs for HA as Russell underline and to verify if load balancing the traffic would improve overall response time.


            Adding more memory for 64bit Tableau VMs will help a lot. Other suggestion is to check if make sense to improve that machine(s) laod adding one or two PCIe ssd card and move temporary files and folders on ssd disks.




            • 3. Re: Suggested # of VizQL Processes
              Russell Christopher

              To be clear -


              Wouldn't split this machine up into VMs unless you NEED high availability.

              • 4. Re: Suggested # of VizQL Processes
                Jonathon Stewart

                Thanks for the detailed responses - very helpful.


                We are currently migrating our server to a cloud hosted environment.  Once we have completed the transition we will upgrade to Tableau Server 8.1.  Unfortunately, I don't think that adding cores is possible at this time. We'll have to see if we can spread out the traffic during these high volume days.