Spring 2025 @cmudb.bsky.social Course: OPTIMIZE! A Journey Through Database Query Optimization
This is a survey course on how to build a SQL optimizer from scratch covering foundational and state-of-the-art implementations. All lectures available on YouTube: 15799.courses.cs.cmu.edu/spring2025/
Buckle up because we're banging into the new year with my annual retrospective of the last year in databases! Highlights include license change blowback, Databricks vs. Snowflake gangwar, @duckdb.org's shotgun weddings, and buying a quarterback to impress your lover: www.cs.cmu.edu/~pavlo/blog/...
Bad actors in this context refer to participants that become unavailable due to various reasons, such as crashes, high response latency, removal by human intervention, or deliberate non-responsiveness when certain constraints are not met (e.g., lacking an up-to-date state).
In a system with 'f' bad actors, we need at least 'f+1' good actors to ensure that the system as a whole remains good. This means the total number of actors in the system is '2f + 1'.
Monday Morning Musings ☀️
Distributed systems and consensus enthusiasts know the rule: we need 2f+1 processes to tolerate up to f failures.
Can you explain why in a way that makes anybody say 'Of course, now I get it'
Seems like a safe bet that object storage as a foundation of data systems architecture is here to stay blog.colinbreck.com/predicting-t...
Really good post. Hadn't considered that the "small data" phenomenon might be caused by incremental computation. Totally makes sense.
Also, totally agree on the gold/silver/bronze staging data take in the post.
