Where this lecture sits in the course

  • This is the last real lecture covered by the final exam; the next class is a prep/review session 0:29.
  • Instead of the usual student vote on which systems to cover, the instructor decided to cram the 15-721 (advanced) content into one lecture because he is not teaching 721 next semester 0:29.
  • Next class walks through real systems — ClickHouse, BigQuery, Snowflake, Databricks, Yellowbrick — to show how the semester's primitives get assembled into shipping products 0:57.
  • The stated payoff: when a vendor claims their system "does amazing things", you read the docs and recognize it as a recombination of known techniques, not something new 0:57 1:21.

Administrivia

  • Project 4 due December 8; Homework 6 due December 9 1:21.
  • Saturday TA session on the 7th, same time/location as before 1:44.
  • Final exam December 13; practice final released the following day 1:44.
  • The final is not cumulative — it covers material since the midterm, though foundational concepts (SQL, buffer pool) are still assumed 1:44 2:21.
  • Three-hour slot, but roughly the same number of questions as the midterm 2:32.

Lecture roadmap

The lecture is organized in four parts 16:59:

  1. Distributed query execution
  2. Distributed query planning (briefly)
  3. Distributed join algorithms (the bulk of the time)
  4. A preview of cloud systems

Framing: last class covered distributed transactional systems; this one covers what changes architecturally for analytical systems spread across nodes 3:32.