To achieve extremely fast communication (less than 10 milliseconds) between different virtual machines, the best option is to use different zones within a single region. This setup ensures high availability and redundancy while maintaining low network latency compared to spreading the instances across multiple regions or a single zone.
In a public cloud environment, the cloud provider is responsible for maintaining the physical hardware and ensuring that it has sufficient capacity to meet user demands. Hardware maintenance involves actions like repair, replacement, and upgrades of physical servers, storage devices, and networking hardware. Hardware capacity management includes monitoring and managing the hardware resources to ensure there is always adequate capacity to meet the needs of all customers using the cloud services. The cloud provider's responsibility fundamentally covers these aspects to ensure the robustness and reliability of their infrastructure.
Deploying the application on Compute Engine using preemptible instances is the most cost-effective solution for this scenario. Preemptible instances are much cheaper than regular instances and can be interrupted at any time, which aligns with the requirement that scenes can be scheduled at will and interrupted as needed. Given that individual scene rendering takes less than 12 hours, these instances provide the needed compute resources without a stringent service-level agreement (SLA) for completion time, making them ideal for your use case.
To restrict all virtual machines from having an external IP address effectively and consistently across an entire organization, it is best to define an organization policy at the root organization node. This approach ensures that the policy is inherited by all existing and future projects and folders created within the organization. This method centralizes control and maintains compliance, preventing any virtual machines from being configured with external IP addresses regardless of which team or project creates them.
Migrating the workloads to a public cloud will allow the organization to manage its mission-critical workloads consistently and centrally while eliminating the need to manage the underlying infrastructure. Public cloud providers offer global reach, centralized management tools, and the ability to scale resources as needed, which aligns with the requirements of the organization.