1 2 Previous Next 26 Replies Latest reply on Feb 13, 2018 2:07 AM by Santhosh Kumar Miryala

    How can we speed up initial Tableau Server dashboard load?

    chris.schultz

      Without "hacking" Tableau is there any way, any setting or configuration changes, that would make the initial load of dashboards on Tableau Server much faster?   I have very high performing dashboards that take sometimes minutes to load the first time someone opens them on Server.  This is painful.   I noticed at TC14 / data14 that Facebook uses a dashboard cache hack but we're not capable of doing any hacks at this time.   Thanks.

        • 1. Re: How can we speed up initial Tableau Server dashboard load?
          Matt Lutton

          There are a number of factors at play here, so its difficult to give any concrete advice for your scenario.

           

          One thing to keep in mind is: If your dashboard is slow on desktop, it will be slow on Server.

           

          So, Filtering your data before it gets to the view is one way to make dashboard load faster. You can use the "Edit Data Source Filters" option to limit the rows returned or use quick filters sparingly to limit the initial data returned.

           

          Tableau Data Extracts (a columnar data store) will perform faster than live connections.

           

          A lot of quick filters, and overuse of "Only relevant values" filters can have an impact.

           

          There are several blog posts and threads on this forum related to this topic.  Here is one with some great advice:

          Slow loading of dashboards on the server

          • 2. Re: How can we speed up initial Tableau Server dashboard load?
            chris.schultz

            Seems like a standard response that I've seen posted on some of the other forum topics related to Dashboard performance, the difference being that I wrote "I have very high performing dashboards that take sometimes minutes to load the first time someone opens them on Server."  The Dashboards run very fast on desktop.  They run very fast on Server.  The exception is the initial load when someone first opens up the dashboard on Server. 

            • 3. Re: How can we speed up initial Tableau Server dashboard load?
              Matt Lutton

              I'm not sure if that is meant as snide, or not, but I'm simply trying to help you.  One thing to keep in mind:  We're all volunteers here, simply trying to help you locate the information you need, or help you learn yourself. All of the things I mentioned impact performance, in addition to a number of other factors.  I guess I'm not sure what you are looking for, if not advice on what impacts load time.

               

              You haven't given us much information on your dashboards, your data source(s), etc.  The number of factors that impacts load time are many.  We'd need to know a lot more to give you any specific advice for your dashboards.  Otherwise, all I think anyone can do is point you toward information that is already out there.  This is not an official support forum--you can contact support at support@tableausoftware.com

               

              Here's another thread that may help--this topic comes up often, so there are a LOT of threads on what to look at.

              Re: Tableau Server: decrease compute time for reports

              • 4. Re: How can we speed up initial Tableau Server dashboard load?
                Matt Lutton

                Good luck, Chris! Perhaps someone like Russell Christopher will be willing to help you without any additional details.  Although, I'm not sure your approach here is going to encourage folks to help.

                • 5. Re: How can we speed up initial Tableau Server dashboard load?
                  Shawn Wallwork

                  What Matthew said. Especially the Russell ping. I watched a hack-a-ton session he (well Darth Flashypants) gave and seem to remember something about rebaking your cache.

                   

                  --Shawn

                  • 6. Re: How can we speed up initial Tableau Server dashboard load?
                    Russell Christopher

                    Chris, if the dashboard is pretty fast in Desktop, it should only be marginally slower in Server.

                     

                    • How fast does the dashboard open inside Desktop running on the Server?
                    • What sort of disk IO are you getting on the Server? Are you on a SAN? Is it a VM?
                    • What version of Server are you running?
                    • 7. Re: How can we speed up initial Tableau Server dashboard load?
                      Matt Lutton

                      Thanks for confirming Russell... you're the man, and we love you for your participation on the Forums!

                      • 8. Re: How can we speed up initial Tableau Server dashboard load?
                        chris.schultz

                        Howdy Russell!   We are running Server and Desktop 8.2.2.  The Dashboard is using a packaged extract, sourced from Excel.  It's not a huge dataset and the only issue we are having is the load time when someone first opens up a Dashboard.  Other than the initial load the dashboards, all 19 of them, are blazing fast (even the one with a live query against a billion row+ Teradata table).   Attached is a pic.  Nearly everything above the std deviation reference line is the initial dashboard load in question.

                         

                        ComputeTimes.jpg

                        1 of 1 people found this helpful
                        • 9. Re: How can we speed up initial Tableau Server dashboard load?
                          chris.schultz

                          Apologies for coming off rude with my initial response but your initial post was very basic tableau 10-ish stuff and if you re-read my post it was about high performing dashboards not running fast on the initial load.  A very specific problem that doesn't merit basic things on filters and extracts.   It wasn't meant to be snide but it seemed that you were copying and pasting a very stock response to generic dashboard performance problems.

                          • 10. Re: How can we speed up initial Tableau Server dashboard load?
                            Russell Christopher

                            OK. There was a PostgreSQL bug in 8.2.1 that impacted performance, but you'd see it across the board, plus you're on 8.2.2

                             

                            Just so I'm 100% clear - the same workbook opened in Desktop on the Server machine runs how much faster than the Server version?

                             

                            Do you have bigger extracts on this server and do they respond faster?

                             

                            Was this workbook built with the legacy Excel connector or the new one? What happens if you take the extract you created of the Excel workbook then reconnect it to the workbook as a second data source. In other words, remove the "trail to Excel" in the second data source...swap data sources to make this the one that gets used and then see if it behaves differently

                            • 11. Re: How can we speed up initial Tableau Server dashboard load?
                              chris.schultz

                              When I open the dashboard in Desktop there is no load time - we're taking maybe a few hundred rows of data in the packaged extract.  I did notice that when I open up the workbook in Desktop I get the Custom SQL pass through warning and I also get a prompt to provide the password to two published data sources that are out on Server .  The Custom SQL is for the Teradata extracts that get refreshed and can be converted to views in Teradata - we are moving are users in that direction anyways.  The workbook was original built using a Desktop 8.1 version.  I opened it in 8.2.2 and saved it as an 8.2 workbook prior to publish.  I can change to data sources to point to db views instead of using custom SQL but I must say this appear to be a common problem.  Even the gurus at Facebook were talking about dashboards taking too long to load when the users first open up a workbook on server.

                              • 12. Re: How can we speed up initial Tableau Server dashboard load?
                                Matt Lutton

                                No problem; I'm here to help where I can.  Hence pinging Russell when I don't know how to help any longer.  Cheers and best of luck!  No grudges here.

                                • 13. Re: How can we speed up initial Tableau Server dashboard load?
                                  Russell Christopher

                                  Can you get a perf recorder report off this sucker and compare it to what is happening on Desktop?

                                  • 14. Re: How can we speed up initial Tableau Server dashboard load?
                                    chris.schultz

                                    I'm going to email you three performance recorder workbooks.  One for Desktop version 8.2.2 and the others are for Server 8.2.2 and Server 8.1.7.  We are in the process of upgrading to 8.2.2 so that version is in pstage mode while 8.1.7 is in production.  Based on this and other testing 8.2.2 seems to run much faster all around.  There are other factors in play as well.  While I was testing this earlier today and yesterday (before I ran performance recorder) the background process was running for a large extract.  This seemed to be bringing our two node production server to a crawl all around even though we have backgrounder and data engine on separate nodes.  During the extract refresh I was looking at the system performance of each server node.  The worker, which has background, VizQL, and other required services running on it didn't seem to be utilizing much CPU and memory, nor I/O for that matter.  The primary, which has data engine, VizQL, the repository, and other required services seemed to be utilizing high cpu and moderate memory resources.  We consider ourselves to be a heavy extract server environment with medium to low concurrency.  I will send you a snapshot of our server config in case that has something to do with our performance issues. 

                                    1 2 Previous Next