Hey Neil -
Tableau Extracts aren't really meant to be a "near real-time" solution, You're probably best served by going with a live connection to the data source.
Just curious - what sort of data are you working with? Why the requirement of 5-minute latency?
I'm working with trading data that is updated every 5 minutes.
I am currently using a live connection. However, what is annoying about the live connection is every time the user opens the report it defaults to the view when the report was uploaded. They then have to refresh it manually.
Out of curiosity, why are extracts not suitable for near real-time solutions?
Weird – you should get the same “default” behavior regardless of whether you’re using an extract or a live connection. Maybe post some screenshots of what’s happening or something?
In any case, why don’t you have your users leverage the “Remember My Changes” feature? This will allow them to “lock in” their changes…they can even save the changed report as “their default”.
Hey Neil -
Extracts are generally not appropriate for RT or near-RT data because it's difficult to keep datasets refreshed unless you're in a near constant refresh cycle. This can lead to issues / questions like:
- I'm always refreshing, and I can see that said process is relatively expensive - I see contention between my workbook rendering and extract refreshing now. What to do? I don't want my rendering speed to suffer just cause I'm extracting 24x7 (solution: more hardware, more licensing $$, which customers don't like)
- My refresh takes more than five minutes because I'm dealing with <insert data volume here>. However I'm refreshing every 5 minutes. How do I manage this? Can I have multiple schedules operating on the same data source so I get an overlap? (no.)
- I'm refreshing my data every <x> minutes but I'm still picking up older data from the cache. How do I fix? (turn off caching, which turns into....) Hey! Now that I've turned off caching, performance is going down! (yes, you turned off caching).
Don't get me wrong - where there's a rule, there's an exception - there may be certain scenarios (small data sets which refresh really quickly and don't need for caching) where you could make this work and it might be appropriate to have a higher refresh rate than 15 minutes...but this can very quickly turn into a Gordian Knot and I suspect the developers didn't want to give users a convenient way to shoot themselves in the foot in this regard.