silver shoal
One of the core USP of runpod is to only pay for compute when needed, in my case, I have requests 24/7, but it fluctuates 10x throughout the day depending on the time of the day. However, every week I've noticed some GPUs/CPUs run out of supply when one user or more users have spikes in traffic. Thus, if the runpod peak load happens when I have low usage, when it climbs, there will not be any resources available to allocate.
Thus, it would be ideal to be able to reserve compute in advance for a fee, to be able to meet expected demand. Ideally set on a weekly hourly schedule and be easily updated.
deft trench