Automatic PoW scaling about kctf HOT 1 OPEN

strawman:

• if powDifficultySeconds == NUMBER, leave it at that
• if powDifficultySeconds == auto
  • default to 3 seconds (so pow is visible, but not significant)
  • whenever the number of Pod replicas is higher than the minReplicas, add 10 seconds (INCREMENT) per extra replica
  • whenever the number of replicas is equal to the maxReplicas, double the INCREMENT once (20 seconds per replica)
  • whenever the number of replicas has been equal to maxReplicas for 10 minutes or more, double the INCREMENT
  • whenever the number of Nodes is equal to the minimum, leave INCREMENT as-is
  • whenever the number of Nodes is higher than the minimum nodes, add 10 seconds per extra node
  • whenever the number of Nodes is equal to the maxNodes, double the INCREMENT once
  • whenever the number of replicas has been equal to maxNodes for 10 minutes or more, double the INCREMENT

rationale:

• if a single task is DoSing the CTF, then it will reach the maxReplicas somewhat quickly. It doesnt make sense to significantly increase the difficulty on tasks with a small number of maxReplicas, because it is more likely to be just an underprovisioned CTF, however, a CTF with a large number of maxReplicas that hits it, is very likely experiencing abnormal traffic, so increasing the proof of work linerarly to the number of replicas makes sense.
• if a single task is DoSing the CTF but the maxReplicas is higher than the resources among all nodes together, then it'll hit the maxNodes limit as well. this affects the scaling and availability of all tasks, not just one task, and it'll also double the waiting time for the tasks being throttled.
• this makes the difficulty slowly increase (linearly) and then, once nearly out of resources, it'll increase exponentially every 10 minutes

from kctf.