| 摘要 |
A reliable and scalable system and method of broadcasting information to other computer nodes in a communication network requires only O(2) time steps. According to one aspect, after broadcasting data in O(1) steps to all nodes in the network, the system and method provides a distributed reliability protocol to ensure data delivery which only requires an additional O(1) steps. Therefore, unlike in prior art approaches where the root or co-root is responsible for the reliable data delivery, each node in the network takes on responsibility to deliver the message to a partner/neighborhood node. The broadcasting method and system of the can be used as building block for most collective/distributive operations, and provides a significant performance advantage in parallel computer systems that have multicast/broadcast capabilities. |
