Scale and Size Oso Fallback
The size of your deployment will depend on your environment and workload. The minimum system requirements is sufficient to start a Oso Fallback node and support a test workload but may not be sufficient for your full production workload.
Scaling the CPU
Each Oso Fallback node is designed to operate independently so you may increase overall capacity by increasing the number of CPUs per node or adding more nodes. When thinking about autoscaling, it's important to plan for any start up time to download the latest copy of rules and data. We recommend monitoring average CPU across all Oso Fallback nodes and increasing capacity when it exceeds 80%.
Sizing the memory
At baseline the Oso Fallback node uses minimal memory. However, we use in-memory caches to speed up query evaluation. Depending on the diversity of your queries, the cache may need to be larger. We recommend monitoring the max memory utilization observed across all Oso Fallback nodes and increasing the memory allocation per node if it exceeds 80%.
Sizing the disk
The Oso Fallback node downloads the copy of rules and data to the provided disk. When there is a newer copy of the rules and data available, the Oso Fallback node downloads the newer version while the current version is still in use; this results in a temporary doubling of the disk utilization. We recommend monitoring the disk utilization and provisioning more disk if it exceeds 65%. Because the Oso Fallback node fetches data from disk to answer queries, we also recommend using low latency, high throughput disk volumes.