In this two-part paper, we consider SDL constructions of optical queues witha limited number of recirculations through the optical switches and the fiberdelay lines. We show that the constructions of certain types of optical queues,including linear compressors, linear decompressors, and 2-to-1 FIFOmultiplexers, under a simple packet routing scheme and under the constraint ofa limited number of recirculations can be transformed into equivalent integerrepresentation problems under a corresponding constraint. Given $M$ and $k$,the problem of finding an \emph{optimal} construction, in the sense ofmaximizing the maximum delay (resp., buffer size), among our constructions oflinear compressors/decompressors (resp., 2-to-1 FIFO multiplexers) isequivalent to the problem of finding an optimal sequence ${\dbf^*}_1^M$ in$\Acal_M$ (resp., $\Bcal_M$) such that $B({\dbf^*}_1^M;k)=\max_{\dbf_1^M\in\Acal_M}B(\dbf_1^M;k)$ (resp., $B({\dbf^*}_1^M;k)=\max_{\dbf_1^M\in\Bcal_M}B(\dbf_1^M;k)$), where $\Acal_M$ (resp., $\Bcal_M$) is the set of allsequences of fiber delays allowed in our constructions of linearcompressors/decompressors (resp., 2-to-1 FIFO multiplexers). In Part I, wepropose a class of \emph{greedy} constructions of linearcompressors/decompressors and 2-to-1 FIFO multiplexers by specifying a class$\Gcal_{M,k}$ of sequences such that $\Gcal_{M,k}\subseteq \Bcal_M\subseteq\Acal_M$ and each sequence in $\Gcal_{M,k}$ is obtained recursively in a greedymanner. We then show that every optimal construction must be a greedyconstruction. In Part II, we further show that there are at most two optimalconstructions and give a simple algorithm to obtain the optimalconstruction(s).
展开▼
机译:在这个由两部分组成的论文中,我们考虑通过光开关和光纤延迟线进行有限数量的再循环的光队列的SDL构造。我们表明,在简单的分组路由方案和有限的循环次数约束下,某些类型的光学队列的构造(包括线性压缩器,线性解压缩器和2对1 FIFO多路复用器)可以转化为等效的整数表示问题。相应的约束。在给定$ M $和$ k $的情况下,在我们的线性压缩器/解压缩器的构造(resp。,2-to-to)中,在最大化最大延迟(resp。,缓冲区大小)的意义上,找到\ emph {最优}构造的问题。 -1 FIFO多路复用器)等效于以下问题:找到一个最佳序列$ {\ dbf ^ *} _ 1 ^ M $ in $ \ Acal_M $(分别为$ \ Bcal_M $),使得$ B({\ dbf ^ *} _1 ^ M; k)= \ max _ {\ dbf_1 ^ M \ in \ Acal_M} B(\ dbf_1 ^ M; k)$(resp。,$ B({\ dbf ^ *} _ 1 ^ M; k)= \ max _ {\ dbf_1 ^ M \ in \ Bcal_M} B(\ dbf_1 ^ M; k)$),其中$ \ Acal_M $(分别为$ \ Bcal_M $)是在我们的结构中允许的所有光纤延迟序列集线性压缩器/解压缩器(分别为2比1 FIFO多路复用器)。在第一部分中,我们通过指定类$ \ Gcal_ {M,k} $的序列,例如$ \ Gcal_ {M,k ,,提出了线性压缩器/解压缩器和2比1 FIFO多路复用器的\ emph {greedy}结构。 } \ subseteq \ Bcal_M \ subseteq \ Acal_M $,而$ \ Gcal_ {M,k} $中的每个序列都是通过greedymanner递归获得的。然后,我们证明每个最优构造都必须是贪心构造。在第二部分中,我们进一步证明了最多有两个最佳构造,并给出了一个简单的算法来获得最佳构造。
展开▼