2 years ago
Hello! I'm evaluating Railway for my project, but I have a concern about its resource management model. I understand that Railway charges are based on usage, which is appealing. However, we don't live in a world of infinite computing resources, so this leads me to these questions:
Resource Reservation: Does Railway reserve a constant 8 CPU cores and 8 GB of RAM for each container, or are these resources shared between node deployments?
Resource Guarantees: How does Railway ensure that the resources my container needs at a given time (e.g., 8 GB of RAM) will be available when my application needs them? Are resources shared across deployments, and if so, could resource contention occur if a node on which a deployment is running becomes overloaded?
I'd appreciate some insights into how Railway manages resources to address these concerns.
1 Replies
2 years ago
Hi! I hope the experience has been good so far, we'd love your feedback post-evaluation.
Resource Reservation: Our scheduling system accounts for resource reservation when deciding which host to place workloads on, i.e. resources are effectively guaranteed at the time of your deployment according to your account's plan limits (8vCPU/8GB RAM for Hobby, 32vCPU/32GB RAM for Pro).
Resource Guarantees: Plan limits are treated as hard constraints in our scheduling algorithm, so hosts are never over-provisioned with workloads (if we know a host won't be able to meet the demands of a service, we won't place the service on that host).
Let us know if you have any other questions.