36:44 that's oversimplified. there is more into that. coordination, distributed resource management, failure detectors. it's a bit easier of your system is only querying data.
It's said we don't see anything on the screen. While it's not required, but sometimes might be helpful, when Andy point out something on slides. Not complaining, just mention. Andy is the best lecturer I know on RU-vid.
Great lectures Andy, appreciate you uploading all the videos! I’ve been meaning to ask you this: what software do you use to prepare the slides? Doesn’t seem like powerpoint. But if it is, mad skills.
Best Introduction to OLAP databases, It took me 2 days to fully understand the paper but I am so excited to read other papers and finish all the lectures bit by bit. Thanks for the videos Professor :)
@andypavlo Is this new semester the best place to start as of 2024? I see that there are differences compared to 15-721 Advanced Database Systems (Spring 2023)'s playlist.
A note on 38:30 : "The queryplan of Postgres is a tree; you only have one parent, whereas a with a DAG you can do some computation here for a nested query and send it to two different parts of the tree." While the query plan is always shown as a tree, it does support multiple references to materialized queries when CTEs are involved, effectively making the query plan of one SQL query a DAG. You'll see this as e.g. a "CTE Scan" node: EXPLAIN WITH a AS (SELECT DISTINCT relname FROM pg_class) SELECT min(b.relname), max(c.relname) FROM a b, a c; It's not super optimized as it doesn't have indexing on the intermediate table, but it does mean it does allow some deduplication of work effort and technically makes it a DAG (even if CTE scan isn't done in parallel with the construction of the materialized CTE).
Yes but with a DAG you don't need to materialize it, you can even pipeline... What you described is essentially just like two base table scans from the same table. Sure, you can draw those as a DAG too but that's not what you're actually doing...
Seems that not having the Microsoft OLAP stack of Power BI is a big omission. The Tabular model in Excel Power Pivot, Analysis Services, and Power BI have been using columnar storage since 2010.
