3

I've been hearing conflicting statements on how much records / data size, tableau can handle. In the last week two people have told me they have dashes which are, 100m and 600m records. They do incremental refreshes.

If I have a dash with xxx million records. Do clients only receive the data that is in their aggregated view.

So, if I have a source with 200million records. In the dash it shows the aggregated total per week per product. Let's say this is 400 cells(underneath it's millions of records). Is the client only receiving 400 data points. If I then add filters to sub product or user level data, would that mean all of these data is imported due to the filters? If this is the case, how does this affect speed?

JohnandLyn Henry
  • 277
  • 1
  • 3
  • 11
  • You need to provide more information on how you are accessing the data. Are you connected to a database, reading from a file, or are you using a Tableau Extract? The extract has the capability of aggregating the data, so you would not have all the data from the source - but you have to tell it to do that explicitly. If you haven't, then Tableau queries all of the data. – Nick Aug 16 '16 at 17:20
  • See http://stackoverflow.com/a/31171086/441979 – Alex Blakemore Aug 16 '16 at 18:44

1 Answers1

4

Ultimately, Tableau can handle as much data as your datasource can handle. If you are set up so Tableau connects to a datasource directly, only the results of a query are transmitted to the user. I've got billion row datasources in BigQuery that return reasonably fast aggregated numbers to Tableau.

If your datasource is not fast then this won't give good results in Tableau.

If you are using extracts, where, in effect, Tableau pulls all the data locally, things will usually be faster, but you will have local drive and memory limits on the size of the dataset. And each user will need an extract. Unless you are using Tableau server in which case the extract can be on the server.

Dashboards built on big datasources sometimes get slow when there are a lot of filters because populating each filter requires a datasource query (which may be triggered every time you use a filter). There are strategies to speed up dashboards with this problem by using partial extracts that generate all the values used for filtering (you can sometimes use parameters for a similar speed gain). Or even just designing the filters intelligently. But speed is usually the limiting factor not the size of the source table.

The only real limit on how much Tableau can handle is how many points are displayed. And that depends on RAM. In my experience a 4GB machine will choke on a chart will a couple of million points (e.g. a map plotting every postcode in the UK). But on a 16GB RAM machine I have never found a limit other than how fast the points are drawn.

matt_black
  • 1,290
  • 1
  • 12
  • 17