# USING PARALLEL EXECUTION <sup style={{ fontSize: '0.6em', color: '#888' }}>Enterprise</sup>

The `USING PARALLEL EXECUTION` clause allows you to instruct the query planner
to execute the query using multiple threads. This can significantly improve
performance for analytical queries on large datasets.

## Syntax

```cypher
USING PARALLEL EXECUTION [num_threads]
MATCH ...
RETURN ...
```

- `num_threads` (Optional): An integer specifying the number of threads to use.
  If omitted, Memgraph determines the optimal number of threads.

## Usage

Place the `USING PARALLEL EXECUTION` clause at the beginning of your query.

### Basic Usage

Run a query with default parallel settings:

```cypher
USING PARALLEL EXECUTION
MATCH (n:Transaction)
WHERE n.amount > 1000
RETURN count(n);
```

### Specifying Thread Count

Run a query using exactly 4 threads:

```cypher
USING PARALLEL EXECUTION 4
MATCH (n:Transaction)
RETURN avg(n.amount);
```

## Considerations

- **Enterprise Feature**: Requires Memgraph Enterprise.
- **Privileges**: Requires `PARALLEL_EXECUTION` privilege.
- **Plan Caching**: Queries with a specified thread count are not cached.
- **Fallback**: If parallel execution is not possible for the given query plan, Memgraph falls back to single-threaded execution.
- **Performance**: Best for aggregation queries that split into independent work. Operators like `DISTINCT`, `SKIP`, `LIMIT`, or hops limits require cross-worker synchronization and may yield little or no speedup. See [Parallel execution](https://memgraph.com/docs/querying/parallel-execution#performance) for details.

For more details on how parallel execution works, see the [Parallel execution](https://memgraph.com/docs/querying/parallel-execution) guide.