论文部分内容阅读
在人们通常强调防止流溢出而同时又必须对效率给予更多注意的时候,流控制目标的方案是不明确的。在层次体系结构中,流控制分成两类:端到端(end-to-end)控制和拥挤(congestion)控制。象通常在传输协议中所碰到的问题(丢失、重复、顺序错…等)一样,本文讨论节流(throttling)的工具(停/发、信用、速度、类、名字空间等)。我们将讨论在现有网络中所用到的端到端控制和拥挤流控制的方法。读者将会看到,使用简单的机构便能把数据流控制住。在有些网络中,也使用与虚拟线路有关的步进式(step-wise)流控制,但这并没有显出多大的优越性,而是弄得更加复杂。真正的问题是为好多用户共享的分布式资源的最优化。所提出的方法是要详细说明所保证的服务类,对这些服务类;资源将按优先权进行分配。例如,低传输时延,有保证的吞吐量,有保证的传输。此外,为了完全免除干扰,用户必须了解服务保证的种种限制。
At a time when people generally emphasize prevention of run-off and at the same time they must pay more attention to efficiency, the goal of flow control is ambiguous. In a hierarchical architecture, flow control falls into two categories: end-to-end control and congestion control. Like the problems that are commonly encountered in transport protocols (lost, duplicated, sequential wrong ... etc), this article discusses throttling tools (stop / wait, credit, speed, class, namespace, etc.). We will discuss the end-to-end control and congestion flow control methods used in existing networks. The reader will see that using a simple mechanism can control the data flow. In some networks, step-wise flow control related to virtual wires is also used, but this does not show much superiority, but more complex. The real problem is optimizing the distributed resources that many users share. The proposed method is to specify the guaranteed service classes for these service classes; the resources will be allocated on a priority basis. For example, low transmission delay, guaranteed throughput, guaranteed transmission. In addition, users must understand the limitations of service assurance in order to be completely immune to interference.