Course Overview & Logistics

CMU Database Group · 21:20 · Watch on YouTube

Andy Pavlov's logistics lecture for CMU 15-445/645 (Fall 2024), recorded ahead of the semester so the first real class can start immediately on the relational model. It covers what the course actually is — building the internals of a database system, not using or administering one — and every policy that governs it: waitlist reality, platforms (Piazza, Gradescope, Canvas), the 15/45/20/20 grade split, six homeworks, the BusTub C++ projects gated by a mandatory Project 0, office hours, late days, and plagiarism enforcement. It closes with the semester's extras: weekly industry flash talks, a September recruiting event, and an optional seminar series on composable database building blocks.

Topics

  1. Waitlist & Enrollment
  2. Course Scope & Lecture Rules
  3. Platforms & Course Materials
  4. Grading Breakdown & Homeworks
  5. Projects, BusTub & the C++ Requirement
  6. Office Hours & Late Days
  7. Academic Integrity
  8. Industry Flash Talks & Recruiting
  9. "Database Building Blocks" Seminar Series
  10. What Comes Next
flowchart TD
    L[Logistics lecture] --> A[Getting in]
    L --> B[Doing the work]
    L --> C[Beyond the course]

    A --> A1[Waitlist: out of instructor's hands<br/>445/645 now offered every semester]
    A --> A2[Scope: build DB internals, not use them<br/>C++17 required]

    B --> B1[Platforms: website, Piazza,<br/>Gradescope, Canvas]
    B --> B2[Grades: HW 15% · Projects 45%<br/>Midterm 20% · Final 20%]
    B --> B3[Project 0 HyperLogLog<br/>pass by Sep 8 or drop]
    B --> B4[4 late days · -10%/day after<br/>No Sunday office hours]
    B --> B5[Plagiarism checked against<br/>public GitHub repos]

    C --> C1[Wednesday 10-min flash talks]
    C --> C2[Recruiting event Sep 16-17]
    C --> C3[Seminar series from Sep 23<br/>composable systems, DataFusion]

Key takeaways

  • The class is about implementing a DBMS, not operating one: after the SQL homework, students build their own database system, and applied "how to use/administer a database" courses live in Heinz College and SCS instead (4:09).
  • Enrollment pressure got resolved structurally, not individually: the instructor lost waitlist control to separate 445/645 administrators, but hiring Jignesh Patel means the course now runs every semester (1:39, 2:09).
  • Grades are 15% homeworks, 45% projects, 20% midterm, 20% final — projects dominate because the course must qualify as a software systems elective; the class is curved, and leaderboard rank earns extra points (8:35, 9:04).
  • The mandatory, ungraded Project 0 (a HyperLogLog in BusTub, all tests passing by September 8, or you're asked to drop) has significantly reduced the course's attrition rate over the years it has been used (12:29, 13:24).
  • Plagiarism detection is concrete, not theoretical: public GitHub repos from non-CMU followers are collected and run through Gradescope's plagiarism checker — and those solutions are usually worse than CMU code anyway, per the leaderboards (7:26, 15:45).
  • Deadline design deliberately discourages last-minute work: no Sunday office hours (only a TA power session the Saturday before each project), 4 automatic late days for projects only, -10%/day after that, and no exceptions for job interviews (14:08, 14:56, 15:26).