SchedLab
A learn-first experiment that makes fragmentation and head-of-line blocking visible. Replay the same abstract cluster under FIFO and opportunistic backfilling.
The teaching model
Four racks, eight abstract slots per rack, and jobs that require a contiguous block within one rack. Free capacity can exist without a placement for the first waiting job.
A controlled comparison
The guided lesson compares strict FIFO and opportunistic backfilling from the same starting conditions. It reports observed wait, started and completed jobs, jobs still queued, and average occupied capacity over a selected horizon.
Honest metrics
Waiting jobs are reported separately. Average waiting time includes only jobs that started, so the view does not invent an eventual wait for unfinished observations. The abstract slot model makes no GPU performance claim.