问题
I am trying to develop a network safe decentralized replication based distributed system. I am looking for a java library with following requirements:
Library should be able to initialize the n nodes in a decentralized fashion (no master or slave). It should be able to recover from network failure at start itself. For example: I try to initiate a network with 5 nodes, but only 3 get started.
Once initialized it should be able to detect the node loss, notify the user so user can take some remedial steps on application front and recover from it. I am not concerned about any new node or failed node joining the cluster back again. But if it supports that too, it is good.
It should allow P2P communication. If it can support efficient P2P and multicast both that is very good.
Allow sending Runnable message and serializable objects between nodes as in Aleph. Alepha is good one, it does not support the node failure/recovery.
Basically I will be creating the dynamic quorums of nodes based on list of active nodes and replicating the objects on different quorums. My frame-work will allow the users to talk these quorums and access the objects. In case of node failure I need to rebuild quorum with new list of active nodes. I want to concentrate on quorum algorithm and save energy on networking capability. Please suggest some suitable library for this purpose. If you know any similar quorum solution too, kindly refer that too.
回答1:
As no one answered, I am just updating the solution I used: JGroups. It is great for cluster based multi casting.
来源：https://stackoverflow.com/questions/8782922/decentralized-clustering-library-for-java