Struct AutoDiscoveryBackend 

pub struct AutoDiscoveryBackend { /* private fields */ }

Auto-discovery distributed backend.

Automatically discovers peers on the local network using mDNS/Bonjour and establishes connections for gradient synchronization.

Implementations

impl AutoDiscoveryBackend

pub async fn new() -> Result<Self>

Create a new auto-discovery backend with default configuration.

pub async fn with_config(config: AutoDiscoveryConfig) -> Result<Self>

Create a new auto-discovery backend with custom configuration.

pub fn peer_id(&self) -> &PeerId

Get the local node’s peer ID.

pub fn peer_id_string(&self) -> String

Get the local node’s peer ID as a string.

pub fn topology(&self) -> SharedTopology

Get the current cluster topology.

pub fn peer_count(&self) -> usize

Get the number of discovered peers.

pub async fn wait_for_peers( &self, min_peers: usize, timeout_duration: Duration, ) -> Result<usize>

Wait for a minimum number of peers to be discovered.

Returns the number of peers found, or an error if timeout occurs.

pub async fn establish_ring(&self) -> Result<()>

Ensure the ring is established, initialising it exactly once.

Uses tokio::sync::OnceCell::get_or_try_init so that exactly one concurrent caller performs the TCP connection and all others wait. If the connection attempt fails the cell remains unset, allowing the caller to retry on the next all-reduce.

pub fn is_ring_ready(&self) -> bool

Check if the ring has been successfully established.

Returns true iff establish_ring has completed successfully at least once. This is a cheap non-blocking check suitable for logging.

Trait Implementations

impl Debug for AutoDiscoveryBackend

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl DistributedBackend for AutoDiscoveryBackend

fn rank(&self) -> usize

Get the rank of this node (0 to world_size - 1).

fn world_size(&self) -> usize

Get the total number of nodes.

fn all_reduce<'life0, 'life1, 'async_trait>( &'life0 self, buffer: &'life1 mut [u8], op: ReduceOp, ) -> Pin<Box<dyn Future<Output = Result<()>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait,

Perform an all-reduce operation on a buffer.

fn barrier<'life0, 'async_trait>( &'life0 self, ) -> Pin<Box<dyn Future<Output = Result<()>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait,

Barrier synchronization.