Could you point to some resources on how the underlying (distributed) runtime is architected/works? Also, is it primarily designed to support 'bringing the code to the data', or can it also support 'bringing the data to the code' a la snowflake?