Architecture

What are Transfer Agents in Gonka

When you send a request to a neural network via Gonka, it doesn't go directly to a GPU, but to a Transfer Agent – a special intermediary node. It's the Transfer Agent that decides which GPU will process your request. Without Transfer Agents, a network of thousands of GPUs would turn into chaos.

Simple explanation

Imagine a taxi dispatcher. You call a car – the dispatcher finds the nearest available one and sends it to you. You don't need to know all the drivers' numbers and locations. A Transfer Agent works similarly: it receives your AI request, finds a suitable free ML-node, and routes the request to it.

How Transfer Agents work

The Transfer Agent receives an incoming request from a user or application. It verifies the sender's signature and balance. It selects an ML-node with a suitable model and available resources. It directs the request to the selected node and returns the result to the client. Everything happens in milliseconds – the user doesn't notice the intermediary.

Why Transfer Agents are needed

Without Transfer Agents, every client would have to know the IP addresses of all GPUs in the network, independently check their load, and select a suitable node. This is impractical with ~4,648 GPUs in the network. Transfer Agents solve the problem of load balancing, signature verification, and routing – making the network convenient for end-users.
A Transfer Agent is the dispatcher of the Gonka network. It receives your AI request, finds an available GPU, and returns the response. Without it, the user would have to search for the right node among thousands themselves.

Want to learn more?

Understand the GNK economy or start earning right now.

Read also