#09 - Vector Indexes, Inverted Indexes, Filters, Tries ✸ TiDB Talk (CMU Intro to Database Systems)

Andy Pavlo (www.cs.cmu.edu...)
Slides: 15445.courses....
Notes: 15445.courses....
🛢🗲Database Flash Talk🗲🛢
Sunny Bains ( / bainssunny )
PingCAP TiDB (www.pingcap.com/)
Jump to 01:04:15
15-445/645 Intro to Database Systems (Fall 2024)
Carnegie Mellon University
15445.courses....

[@matveyshishov]

It's always a pleasure to start your morning with one of Andy's lectures.
I am still trying to wrap my mind around skip lists, however. What data patterns would make them better than B-trees?
And in general, I've been trying to understand why and how we use data structures based on specific patterns of hardware access and data flows. Not in Karska's learned data structures sense, nor in hardware accelerated DB sense, but just the step of upfront, when we already know fairly well what to expect, somehow specifying characteristics which make our system match the workload better. Kinda like LSM vs B-trees, but what if we could match some combination of hardware and software TO data, as opposed to the usual process of starting with hardware limitations? I don't see much practical application experience, and I don't trust papers (not that there are many papers, too), as I can't clearly understand their assumptions and experimental setups.