Jia Guo

> cc @atris since this is related to the query resiliency and cost estimation Listed are some potential approaches we are evaluating/benchmarking. It's just a sketch pad; the design is...

Update -- we would like to share the design doc here https://docs.google.com/document/d/1Z9DYAfKznHQI9Wn8BjTWZYTcNRVGiPP0B8aEP3w_1jQ/ Please feel free to review/comment

This can potentially be a beginner task

Hey @heatclub it's not yet picked up, do you want to do both? You can take a look at files changed in https://github.com/apache/pinot/pull/8753. 2) is more exploratory as you probably...

Thanks for the clarification question @jugomezv. For 1) I'm thinking to add retry downloading on untar failed for the old (not streamed) segment download code path. As currently the new...

> Oh I think I get it, your fix is was only for streaming, ok I get it. Yes for sure we need that: will it also be an option...

Designing the one request - multiple (sequential) queries model for statistical functions. Planning to use mini-batch stochastic gradient descent for regression algorithms 2. 3. 4.

NaN could be due to Math.sqrt(negative_number) or 0.0/0.0 We have recently discovered this by-definition impl of covariance/correlation has numerical stability issue when E[x^2] ~ E[x]^2 >> 0 (see [1](https://en.wikipedia.org/wiki/Algorithms_for_calculating_variance#Covariance) [2](https://github.com/trinodb/trino/blob/1866a23e3b0377144c1820de892c0de2762351a8/core/trino-main/src/main/java/io/trino/operator/aggregation/state/CorrelationState.java)...

Hey @heatclub sorry for the delayed reply, I was in smth. Yes your understanding is basically correct, we need to retry download upon download failures (which we already do), and...