Execution Policies

Execution policies define the sandbox available to an authorised workload. Operators use them to limit the effect of third-party code on a data-holding node.

Policies can constrain:

• CPU, memory, storage and accelerator use;
• execution duration and concurrency;
• outbound and inbound network access;
• accessible buckets and record operations;
• secrets and environment variables;
• writable temporary storage;
• permitted result types and maximum sizes;
• artifact retention and replication;
• logging and audit requirements.

Nodes reject an execution when they cannot satisfy its authorization and policy without weakening local controls.