Babak Tourani

@babak-t.bsky.social

Father to Lilly the Ragdoll | a nonbeliever in humans or Gods | Minimalism is my drug, especially in audio form | I deal with databases and data tools for money | former TV/Radio presenter & Basketball coach | have been figuring it out for a very long time

I love the CVs of hyped up Data and Analytics influencers. This person's first job title: "Sr. Project Lead!" Hasn't done anything before that and then **boom** "Senior Project Lead". Also, has been a "Chief AI officer" in 2018! Now, a keynote speaker at a local conference near you!

If you are into streaming and stream processing, today is the best day for finding free data: the unbounded stream of bollocks transmitted by TV commentators on Pope Francis' funeral is one of a kind! One called it a spectacle! None of them seem to have EVER watched footage of UN's general assembly!

Humans are stupid! We look at tiny creatures (in relative terms) and call them cute and associate innocence with them. “Ravens are not innocent but sparrows are.” We are sparrows compared to elephants and we have killers and psychopath criminals in our species. Why would sparrows be different?!

On #Snowflake: The EDA query on a 3.1 TBs table with 19B+ records (200 cols) with a predicate on a very high cardinality column (obviously not used for clustering) takes tens of minutes on Small WH. Rejected Request: S -> X-Large! Enabled Search Optimization on the predicate column => 5 secs!

Bild

How can you make a #metric layer tool effective, if you don't physically cut the access to your data warehouse?! If your analysts have access to both (let's say) #Cube AND the data warehouse and can choose either for building their report/visualisation, how can you prevent metric sprawl?! #datasky

What could justify storing date/time in a #database as an epoch? Not only do you make the data less accessible, but you also deprive the optimizer of your database of the ability to estimate the cardinality of the results correctly, if the epoch column is referenced in a predicate. #datasky

Last week in review: optimized performance of two DAGs by converting CTEs to Temp tables in #Snowflake and achieved 7-10X improvement in execution time! In one case SF couldn't do predicate pushdown in joins correctly so we pushed down the predicates for Snowflake through the temp table! #SQL

If "the business" regards your department/team/operation as a "cost center" (<cough> #DataTeams </cough>) that's not necessarily a stance against the cost of your operation. In many cases, that's an indication that the business does not understand what you do! Sort out your business communication!