SaladCloud

Salad Container Engine takes a container image, resource requirements and replica count, then schedules instances across the chosen pool. Job queues distribute work to workers, while a gateway can expose a service. Community Cloud, Dedicated and private fleets use different capacity arrangements.

Queue-based jobs and live endpoints need different deployment choices. A queue lets your application retry unfinished work; a live service also needs replicas and health checks so requests can reach a ready instance. Build for these needs rather than relying only on a GPU memory calculation.

Community Cloud usage is prepaid. Running instances are billed by the second; allocation, image downloads and cold starts are not charged. Priority affects price and interruption exposure. Reserved Dedicated capacity uses a separate contract.

Choose the GPU class and pool for the workload. Community machines can leave the network, so a container must tolerate rescheduling rather than depend on one permanent host.

Deployment documentation and cloud support cover container groups, billing and service configuration. Enterprise support is a separate offering.

Keep durable inputs and outputs outside replaceable workers. A distributed container service needs a different recovery design from a single VPS with persistent local files.