Currently, the communication between servers connected in p2p is synchronous. One of the implications is that a server can only dispatch one job at a time to another server, which can be a severe limit in terms of throughput.

We propose to refactor the server to server communication tio make it asynchronus and remove that limitation. This may involve additional functionalities to ensure the resulting jobs throughput is close to optimal: - additional information may need to be exchanged between servers, for instance how many nodes are available in the remote server - we might have to implement new load-balancing algorithms or modify existing ones - etc...