Measure, Then Build
summary
Advocates an iterative measure, then build workflow: profile real systems to find surprising bottlenecks, then redesign. Examples include measuring existing checkers to isolate the slow random seeks, building a fast fsck with grouped indirect-block scans that yields large speedups and gets ideas absorbed into FreeBSD; implementing Squeak — a SQL-based checker prototyped in days instead of months; fault-injection to reveal inode vs data retry behaviors and bugs; checksum-based journaling that influenced ext4; and a fair-usage locks design to prevent lock-holder denial-of-service.
AI-written summary · about
chapters (18)
- 0:00Start
- 4:46Research Is Hard Work
- 5:23The Research Process
- 8:28First Inspirations
- 13:38Example: fsck
- 18:26File System Checking (part 2)
- 22:03One Level Deeper
- 22:50Evaluating Restart Workload: Random reads to large files
- 25:39Measuring D-GRAID Measuring file system reaction to D-GRAID faults
- 29:53Journaling File System
- 34:00Example: OS Locks
- 35:58Fair-Usage Locks
- 36:56Find Mentors
- 37:37Listening, Ignoring
- 45:02My First Talk
- 49:40Update Protocol #2
- 49:56ALICE Summary
- 51:29A Teaching Story
