Running one instance of a CPU-bound process uses about 1000 CPU-seconds per hour.
When I start 10 identical processes in parallel (via Bash console), each process only uses ~500 CPU-seconds per hour. Total usage is ~5000 CPU-seconds/hour.
During all tests I am far below my daily CPU-seconds limit and not in the tarpit.
Is this expected behavior? Is there any per-user CPU share / concurrency limit or scheduler effect that reduces CPU available per process even when under the daily quota?
Can you provide us with more details so we could investigate / try to reproduce this?
Well I actually can't really reproduce this with normal Python code, except perhaps to a minimal extent with lots of SQL queries. But my actual code writes and reads a lot from databases, so it could simply be that performance suffers because too many queries are being made at the same time. So you could forget about it thanks!
