15 - Concurrency Control Theory (CMU Intro to Database Systems / Fall 2022)

Andy Pavlo (www.cs.cmu.edu...)
Slides: 15445.courses....
Notes 15445.courses....
15-445/645 Intro to Database Systems (Fall 2022)
Carnegie Mellon University
15445.courses....

[@agkhall]

This is one of the rawest intros to a recorded lecture I think I've ever seen.

[@user-fq6kx2ye3o]

Does the statement 'at 13:12, the strawman system is correct' hold true? Even if it only performs a single transaction, there is still a possibility of issues arising if the operations within the transaction access or modify the same record simultaneously. or am I missing something??

[@andypavlo]

Almost every DBMSs will execute queries within the same txn in sequential order. There are some optimizations you can do if you know there are no dependencies between queries in the txn (e.g., if they touch different tables, or if they need to run on different nodes). But you can only do this if the DBMS *knows* what queries the DBMS will execute ahead of time (e.g., in a UDF or stored procedure).

[@Shwed1982]

assumption is that a transaction does not do two things simultaneously by itself.

[@jamalhasanov2270]

How does the T1 proceeeded by the T2 at 43:46 yield A=960, B=1160? I guess it was possible if we had T2, then T1.

[@mephistotel87]

It's an error on the slide. The left schedule and write one should produce the same result (the same effect), otherwise the are not eqiuvalent.

[@longhtr]

Example #2 💀