I am exploring the new IPE feature on 14.10. Very nice!
But sometimes the explain is showing
IPE should kick in if it reduces the overall runtime, so this query probably accessed too small tables.
I usually get the same when I try it on a VM :-)
yes, I used "only" 1 Mio rows on the PPI table.
But the SSI queries also contained a UPI access so should be pretty fast. I was wondering where to draw the line and was therefor asking for more details on these thresholds.
Once I tried with "big" tables I saw that it was used. Took me some time to find the dynamic explain option - there might be some room for improvement in the documentation - or I was just not able to find a good introduction into this...
IPE applies to queries using:
1. Single-row relation
2. Non-correlated scalar sub-queries
Single-row relation points to query condition based on USI or UPI. Which are most likely to be a Tactical queries.
Under limitations, IPE do not process Tactical queries especially if the query has caching plan.
In this scenario, I am trying to understand what type of queries fall into "Single-row relation" which gets benefit out of IPE process.