A cloud engine is a set of nodes (machines) that run the Internet Computer protocol together. This set of nodes is called a subnet. Every node keeps a full copy of your app, its data, and its backend logic, and every node runs the same work. The nodes check each other on every request, which is what keeps your app tamperproof and online even if a node or a whole data center goes down.
That full-copy design is why capacity does not add up the way it would on a normal cloud.
Why capacity is not the sum of the nodes
Each node runs the whole app, not a slice of it, and the same data lives on every node at the same time. So if your engine has four nodes, its memory and compute are what one node provides, not four times that. Four nodes with 100 GiB each give the engine 100 GiB, not 400 GiB.
Adding nodes does not pool their CPU or memory into one bigger machine. It adds copies, which is what keeps your app tamperproof and available.
Memory
- Memory is the memory your apps are using on the engine.
- Effective memory (in the info note) is the engine's per-node capacity, so you can see how much headroom is left.
To add memory, move your nodes to a larger class. Adding more nodes of the same class does not raise the ceiling, because the data is copied to every node rather than spread across them.
CPU
- CPU is the average utilization across the engine's live nodes, shown as a percentage. It is not a sum. A reading of 40% means the nodes are, on average, using 40% of their compute.
- vCPU per node (in the info note) is the size of each node.
For write-heavy work, sustained high CPU means the nodes are near their compute ceiling, and the fix is a larger node class, not more nodes.
Scaling out vs scaling up
- Add nodes (scale out): more fault tolerance, and more capacity to serve simultaneous reads. It does not add memory, and it does not raise the per-node ceiling for writes.
- Move to a larger node class (scale up): more memory and CPU on every node, so more capacity for your app.
Related: Can I see and change the nodes on my engine? Glossary.
Comments
0 comments
Please sign in to leave a comment.