Skip to main content
Grid distributes state and execution so that unrelated work runs at the same time. This is the property that makes the chain scale: a busy entity does not slow the rest of the network.

​
Independent activity runs in parallel

When Alice sends a transfer and Carol sends an unrelated transfer, neither waits for the other. Their activations land on different entities, possibly on different shards, and execute concurrently. This is a direct consequence of how state is partitioned. the Grid has no single global state machine that every transaction must serialize through. Each entity owns its own state and runs on a shard alongside many other entities. Independent entities mean independent execution.

​
A cross-entity flow is a chain of activations

A token transfer is two executions, not one:
  1. The sender’s component decrements the sender’s balance.
  2. The sender’s component sends an activation (a continuation) to the recipient’s entity.
  3. The recipient’s component increments the recipient’s balance.
Each side writes its own state. The flow finalizes in a single block when the chain is short, or across a small number of blocks when it spans many shards. See Shards and clusters for the cross-shard mechanics.

​
A busy entity doesn’t slow the chain

When many activations target the same entity (a popular merchant collecting payments, a popular agent receiving requests), those activations queue at that entity and are processed in order on its shard. The rest of the network keeps running. Contention stays local to the hot entity, not global to the network.

​
What you observe as a client

From a client’s perspective, the parallel execution model is mostly invisible:
  • Submit an activation. Wait for finality. Read the receipt.
  • Independent transfers don’t interfere with each other.
  • A spike in unrelated traffic doesn’t slow your activation.
You don’t manage parallelism, schedule shards, or coordinate cross-entity execution. The runtime handles it. Your job is to submit the work and wait for finality.
  • Shards and clusters: the partitioning model that enables parallel execution.
  • Components: the unit of execution that owns its own state.
  • Activations: the unit of work submitted to a component.
  • Finality: when you can trust an activation is settled.