Module simnet

Expand description

A simulator capable of simulating Hyperactor’s network channels (see: [channel]). The simulator can simulate message delivery delays and failures, and is used for testing and development of message distribution techniques.

Structs§

BetaDistribution: A beta distribution scaled to a given range of values.
LatencyConfig: Configuration for latencies between distances for the simulator
SimNet: SimNet defines a network of nodes. Each node is identified by a unique id. The network is represented as a graph of nodes. The graph is represented as a map of edges. The network also has a cloud of inflight messages
SimNetHandle: A handle to a running SimNet instance.
SimulatorEventRecord: DeliveryRecord is a structure to bookkeep the message events.
TorchOpEvent: A pytorch operation

Enums§

LatencyDistribution: A distribution of latencies that can be sampled from
SimNetError: SimNetError is used to indicate errors that occur during network simulation.
TrainingScriptState: The state of the python training script.

Traits§

Address: This is used to define an Address-type for the network. Addresses are bound to nodes in the network.
Dispatcher: Dispatcher is a trait that defines the send operation. The send operation takes a target address and a data buffer. This method is called when the simulator is ready for the message to be received by the target address.
Event: The unit of execution for the simulator. Using handle(), simnet can schedule executions in the network. If you want to send a message for example, you would want to implement a MessageDeliveryEvent much on the lines expressed in simnet tests. You can also do other more advanced concepts such as node churn, or even simulate process spawns in a distributed system. For example, one can implement a SystemActorSimEvent in order to spawn a system actor.

Functions§

simnet_handle: A handle for SimNet through which you can send and schedule events in the network.
start: Starts a sim net.
start_with_config: Starts a sim net with configured latencies between distances

Type Aliases§