- Resource monitoring is treated as mandatory at this scale; the rest of the video depends on reading CPU, network, disk, and memory limits correctly.
8:24 - Core utilization measures how much of a time window one CPU core spent doing work instead of sitting idle.
9:25 - A core that was idle for 30 minutes during a 1-hour window has 50% utilization.
10:09 - CPU utilization can be reported two ways: summed across cores, where a 4-core CPU can show 400%, or normalized to 100%.
10:2710:47 - A single tight loop in Node.js consumes one core because one thread can only use one core at a time.
11:3913:10 - On the uploader's 12-core machine, a single-thread loop left much of the total CPU idle.
12:37 - Spawning 12 worker threads runs 12 loops at once, which can drive all 12 cores to full utilization.
14:0215:32 - More threads can speed CPU-bound work, but they introduce concurrency problems such as race conditions and may require coordination tools such as semaphores.
14:23
Why this matters later
- A low request rate with idle CPU means the server may be blocked on network, disk, a database, process overhead, or connection limits rather than raw CPU.
8:24 - A fully saturated CPU means the next improvement probably needs less work per request, more cores, a faster runtime, or a different architecture.
15:47