Simulations within hCore can be run in one of two ways: locally inside your browser, or at scale on our hCloud compute platform.
Stage of Development
Running Experiments or Large Simulations
Number of Agents
10,000+ to ∞
Free tier, pay-as-you-go, and subscription options
Exact in-browser limits will depend on the amount of processing power and RAM available to the browser tab running hCore. The prime constraint is often the power of the local machine's underlying hardware.
Toggle your execution environment from local to hCloud by first clicking on the Experiments button in the playbar at the bottom of the screen, and then tapping the hCloud banner at the top of the menu.
In the background, your browser will establish a connection to our servers which once established will let you run experiments in the cloud in the same way as if you were running them locally.
hCloud resources are billed through a composite metric, hCloud Minutes/Seconds. Currently this is calculated as the amount of time spent executing a simulation on an hCloud server per vCPU.
In our current Early Access phase, by default all simulation run using a single vCPU.
Let's take a look at two examples:
I'm building a simulation and I want to try to run a larger simulation with more agents than I can do it on my local machine. I've set the
number_of_agents in my globals, so I create an experiment with 10x the number and then toggle the Cloud Runner button to connect to cloud and run the experiment for 100 steps. Each execution of a run, from start to finish, will count towards your cloud minute (not including connect/upload/download times).
If it takes about a second to generate a state for this simulation the total billed time would be 100 cloud seconds.
After iterating on my simulation I'm confident the underlying logic is correct and I'm ready to start exploring the effect of different parameters on the outcomes of simulations. I create a linear space experiment that will sample ten times from a range of 1 to 100, and will run for 500 time steps. This creates ten separate simulations that each run for around 50 seconds to 60 seconds. In this second example the total time used would be between 8.3 to 10 minutes.
hCloud runs experiment runs in a parallelized fashion by default on HASH Free, Base, and Pro accounts. In practice this means the memory available to your simulation varies from 4-192GB, depending on your account limits.