12 - Query Execution I (CMU Databases Systems / Fall 2019)

[@Sam-sk3dz]

Cannot understand the last part of JIT compilation and the introduction is so short. Why no students coming up questions? Is that class closing to the ending?

[@andypavlo]

The reason why I don't go into detail is because I have an entire lecture about it in the advanced class: kzbin.info/www/bejne/eqfLopqjh7aod80

[@Sam-sk3dz]

@@andypavlo Thanks. I'll follow on and try to understand.

[@harshavardhanunnibhavi5993]

What is the name of the system that would generate a single function instead of the tree for the query plan?

[@andypavlo]

I don't remember what you are referring to. Do you have a time in the lecture that this comes up?

[@antonlastauka4898]

@@andypavlo I guess it is here: kzbin.info/www/bejne/gmaxmYBtoNxpe8k You are describing the system that is developed internally in CMU

[@siddharthabiswas2147]

What is special about OLAP queries ?

[@angus10292]

the slide on 49:53 , how does the DBMS figure out all the tuples it needs ? Isn't that the expected output from the DBMS ? or in other words, how can the DBMS know which tuples it needs before actually fetching them from disk ? I guess this is determined from the record ID , just before the access methods fetch pages/tuples ?

[@djgreenmoon55]

Say you have a B+ tree index on age, you know which leaf nodes include tuples with age < 30.

[@AshishNegi1618]

Instead of : read leaf node of Index and for each tuple fetch the data page => random IO for every tuple.. First sequential scan all leaf nodes of index => get all page ids that you need .. then fetch data pages in sorted order of page ids..

[@AshishNegi1618]

in index-scan, index should have all the columns for the filtering/where clause. Once final keys are identified, other columns are fetched from main table.

[@ameypatel5774]

How does buffer pool bypass work? If the sequential scan operator will not store the fetched pages in the buffer pool(aka memory), how will it even scan it?

[@andypavlo]

The memory used for the bypass is private to the query (i.e., allocated on the heap by the worker). Other queries cannot see the contents of those pages. With a buffer pool, the pages are visible to all queries.

[@eytehs]

I don't like to show code... SQL does not count because it is beautiful - Andy Pavlo

[@lingjin8917]

what is the mean of offset in the "Late materialization" for the column storage. offset of what?

[@ar_rahman_90]

Column offset in a page

[@lebronsun8205]

in column store, its like the index of every record