Platform services¶
These are the services that run the network itself. Most users interact with them indirectly, through the Parser, the guides and the SDKs. Come here when you need the exact behavior, inputs or APIs of a service, for example when writing a governance or allocation policy.
Where each service runs¶
| Service | Runs on | What it does |
|---|---|---|
| Cluster Controller Gateway | Management cluster | The network's control point for all clusters. Enforces network-wide governance policies. |
| Resource Allocator | Management cluster | Runs allocation policies to choose clusters, nodes and GPUs. |
| Tasks DB | Management cluster | Tracks long-running background operations. |
| Failure policy executor | Management cluster (or anywhere) | Runs fallback policies when something fails. |
| Cluster Controller | Every cluster | Manages one cluster: places blocks, watches nodes, enforces cluster rules. |
| Metrics system | Every cluster + management cluster | Collects and serves hardware, block and vDAG metrics. |
flowchart TB
subgraph MGMT["Management cluster"]
CCG["Cluster Controller Gateway"]
RA["Resource Allocator"]
TDB[("Tasks DB")]
FPE["Failure policy executor"]
GM[("Global metrics")]
end
subgraph CL["Each cluster"]
CC["Cluster Controller"]
LM[("Local metrics")]
end
CCG --> CC
CCG --> TDB
CCG -. "dry runs" .-> RA
LM --> GM
CC --> LM
Also on the management cluster, documented elsewhere:
- Parser: accepts specifications.
- Policies System: stores and runs policies.
- Registries: record what exists in the network.