0

Lets say I have a set of Agents in a distributed network without a centralized unit. I want to communicate them with P2P. So every Agent is a peer right?

The network should build itself and when a new Agent want to take part in this network or wants to leave, the whole thing should still run. Even if there are currently no Agents. (Is a possibility in my case).

So how can an Agent discover another Agent who wants to take part in this Network?

I thought quite a long time about this, and for myself I came to conclusion that a decentralized implementation is not really possible, but I wanted to ask the community.

Dominik Brandstetter
  • 33
  • 6
  • 1
    There are some patterns to build systems like you want. A good day to discover agents in a system is by looking up in a service directory (yellow list) or in a central registration (white list). You can make these lists distributed, replicated and so on. You should take a look at the FIPA standard for agent communication. – darlinton Jul 28 '16 at 10:51
  • Yes, Jade implements this FIPA standard, but I looked on the guide and on some of the examples of Jade and they all assume, that you know the IP of others. This isn't the case in my example. – Dominik Brandstetter Jul 28 '16 at 13:32
  • In Jade you have the AMS and tbe DF agents. You must use their services to find other agents and lookup their IP adresses. This is the answer that Jade platform provides for your question. – darlinton Jul 30 '16 at 11:50

0 Answers0