智能表广播模型

INTELLIGENT LIST BROADCASTING MODEL

紧密耦合系统中的广播模型仍是现代网络系统中讨论的重点之一，现今经常使用的广播模型如传统广播模型及全局表模型都由于没有充分发挥软件的功能而具有较大的局限性．它们要么因存储多张表而不必要地耗费大量的系统资源，要么在一张表上呆板地使用一种预先规定好的路由选择策略而使得该模型的适应性较差并且难于设计．本文提出了一种新的广播模型——智能表广播模型ＩＬＢＭ．在该模型中每一结点都被赋予一个邻结点表与一个结点算法，由结点算法来灵活决定结点在广播过程中对各连接的转发顺序，从而以软件静态形式上的一致实现运行效果的不一致，进而统一邻居表；同时，ＩＬＢＭ中结点算法的引入还增强了网络广播操作的自适应性，能够灵活处理广播源点处于拓扑中不同位置的情况．在整体上，ＩＬＢＭ有效地改善了广播操作的时间性能、系统资源的消耗以及路由算法的设计复杂度．对该模型的实际应用情况，本文讨论了ＩＬＢＭ在一些紧耦合系统常用拓扑结构中的性质，如ｔｒｅ，ｒｉｎｇ，ｇｒｉｄ，ｔｏｒｕｓ，ｈｙｐｅｒｃｕｂｅ及ｄｅＢｒｕｉｊｎ．

Broadcasting communication model in tightly coupled systems is still an open issue. Usually, there are two models: classical and universal lists broadcasting models. But both of these models have some limitations in terms of complexity and performance. They either store many transmission lists in a node which uses substantial unnecessary local memory, or use prescribed sequential selection which make it lack adaptability and difficult to design. In this paper, the authors introduce a new broadcasting model ILBM. In this model, every node is assigned a single list whose entries represent the neighbors, and a node algorithm which is used to determine the order of retransmission. The introduction of node algorithm also enhances the adaptability to the source position. Generally, ILBM could ensure high performance broadcasting with small extra spatial cost, high adaptability and convenience in implementation. As for its application, we discuss some of its properties when applied in some typical tightly coupled topologies such as tree, ring, grid, torus, hypercube and de Bruijn.