8 Replies Latest reply on Mar 8, 2018 4:46 AM by John Sobczak

    Hyper or Just Hype?

    philip.watkinson

      We are currently running TD 10.5 on our test server and it is taking well over an hour to render one extract. I understand that we are to expect superior performance from this new Hyper engine, but so far we have not seen it. Is there something that we are missing? Please, we need some help or at the least some direction.

        • 1. Re: Hyper or Just Hype?
          Carisa Chang

          Hi Philip,

           

          You should expect to see the same or better performance than previous versions of Tableau Desktop that were not using Hyper, but how much of a performance increase you see will depend on a few different factors. Do you happen to know how long it took to create the extract using an earlier version of Tableau Desktop? Or how long the same query takes to run directly on your database? Hyper can't build an extract faster than the database can send the data, so the speed of the database can be a common limiting factor.

          • 2. Re: Hyper or Just Hype?
            philip.watkinson

            Hi Carisa,

             

            Thank you for this key information, which apparently was not covered in the recent "Hyperfest" roll-out of the TD 10.5. So if I understand you correctly, the effectiveness of Hyper is dependent on the speed of the database, correct? If this is accurate, then what steps do you suggest that clients take in order to achieve optimum efficacy? I suspect that there are many more users out there who are also frustrated with this scenario and might benefit from a list of possible improvements.

            • 3. Re: Hyper or Just Hype?
              Carisa Chang

              Hi Philip,

               

              Happy to help! Hyper will stream in the data as fast as the database can provide it - and then create an extract. The extract creation/querying is where Hyper really sings, but it is dependent on the database serving the data up first. There aren't any optimization steps needed for Hyper itself - it's been highly tuned already. But if you're looking to speed up the query on your database, I'd think the steps will vary depending on which database you are using, the different driver options available, and the structure of your data. Usually, some sort of indexing or using views on the database can speed up its response time.

               

              If you test the query on your database - and it comes back much faster than Hyper is making an extract - please open up a case with Tableau Support to look into why.

              • 4. Re: Hyper or Just Hype?
                Patrick A Van Der Hyde

                Philip,

                 

                You might want to test a quick illustration of the improvement on our community visualizations. I upgraded both to utilize Hyper extracts recently -

                 

                Ideas

                 

                Crow's Nest

                 

                Changing the time frames is probably the most intensive thing you can do with these visualizations.  The Tableau community database this runs against contains data rows for every poster, response, and thread that has ever occurred in the community (past 6 years) and it is parsed down to weekly data of specific questions that have been replied to or not.  The views used to take 30-40 seconds to respond but now they are nearly instantaneous.  Note for the Crows Nest view, no data will be returned until you select a large time period like "last 3 years" since we have at least one reply to every post in the forums over that time period (thus the null data). 

                 

                It's just an example but I was very impressed to see the improvement in speed on these public facing visualizations.  I suggest some testing between versions and .tde and .hyper extracts to determine yourself.

                 

                Regards

                 

                Patrick

                • 5. Re: Hyper or Just Hype?
                  Devin Brown

                  I have only performed two new extracts with 10.5, but it also took me at least 50% longer to create the extract into HYPER than TDE. I can understand the database itself limiting gains in extract speed, but I can not understand the database having an affect on a HYPER extract taking longer than TDE.

                  • 6. Re: Hyper or Just Hype?
                    Patrick A Van Der Hyde

                    Hello Devin,

                     

                    In most instances - the extract in Hyper should be faster than TDE formats.  The logic behind this statement is here - The New Data Engine in Tableau Server 

                     

                    If you run into an issue preventing the use of Tableau or hindering your workflow, please submit a support ticket to have your specific extracts reviewed with a Support Engineer. 

                     

                    Patrick 

                    • 7. Re: Hyper or Just Hype?
                      Devin Brown

                      In addition to significantly slower extraction, I am looking at two identical data sets: 6,666,754kb as .TDE, 10,912,896kb as .HYPER. I'm not feeling the hype, to be honest.

                      • 8. Re: Hyper or Just Hype?
                        John Sobczak

                        Is it possible that Hyper's "In-Memory" data might conflict with PC desktop hardware that uses Intel's Rapid Storage Technology RAID 1 Volume mirroring?   I'm seeing garbage string data for the Excel Sample Superstore data when connecting live or as an extract using Desktop 10.5   I have even tried  re-downloading 10.5 Desktop and it still occurs.  I'm trying to find another RAID 1 machine to see if it is consistent on the same hardware.  I do not have this issue using 10.5 on my laptop.