NXG05-6: Minimum Delay Scheduling in Scalable Hybrid Electronic/Optical Packet Switches
NXG05-6: Minimum Delay Scheduling in Scalable Hybrid Electronic/Optical Packet Switches
复制标题
NXG05-6：可扩展混合电子/光分组交换机中的最小延迟调度
DOI：
--
复制
发表时间：
2006
期刊：
影响因子：
--
通讯作者：
K. Yeung
中科院分区：
文献类型：
--
作者：
Bin Wu;K. Yeung
一种混合电子/光分组交换机由通过光交换结构互连的电子缓冲线路卡组成。它为下一代高速路由器提供了一种可扩展的交换机架构。由于不可忽略的交换机重新配置开销，许多分组调度算法被发明以确保性能保证的交换（即，100%的吞吐量与有限的分组延迟），以加速为代价。特别地，如果算法总是可以为任何输入流量矩阵找到不超过N个配置的调度，则可以实现最小延迟性能，其中N是交换机大小。提出了各种最小延迟调度算法（MIN、alphai-SCALE和QLEF）。其中，QLEF需要最低的加速比界。在本文中，我们表明，现有的加速约束QLEF是不够紧。一个新的界限，这是10%，低于现有的推导。
A hybrid electronic/optical packet switch consists of electronically buffered line-cards interconnected by an optical switch fabric. It provides a scalable switch architecture for next generation high-speed routers. Due to the non-negligible switch reconfiguration overhead, many packet scheduling algorithms are invented to ensure performance guaranteed switching (i.e. 100% throughput with bounded packet delay), at the cost of speedup. In particular, minimum delay performance can be achieved if an algorithm can always find a schedule of no more than N configurations for any input traffic matrix, where N is the switch size. Various minimum delay scheduling algorithms (MIN, alphai-SCALE and QLEF) are proposed. Among them, QLEF requires the lowest speedup bound. In this paper, we show that the existing speedup bound for QLEF is not tight enough. A new bound which is 10% lower than the existing one is derived.