A throughput-centric view of the performance of datacenter topologies
A throughput-centric view of the performance of datacenter topologies
复制标题
DOI：
10.1145/3452296.3472913
复制
发表时间：
2021-08
期刊：
影响因子：
--
通讯作者：
Pooria Namyar;Sucha Supittayapornpong;Mingyang Zhang;Minlan Yu;R. Govindan
中科院分区：
文献类型：
--
作者：
Pooria Namyar;Sucha Supittayapornpong;Mingyang Zhang;Minlan Yu;R. Govindan
虽然先前的工作已经探索了许多建议的数据中心设计，但只有两种设计，即基于Clos和基于扩展器的设计，通常被认为是实用的，因为它们可以使用商品交换芯片进行扩展。先前的工作使用了两个不同的指标，二分带宽和吞吐量，在规模上评估这些拓扑结构。理论上或实际上，这些指标如何相互关联，我们知之甚少。利用这些拓扑结构的特点，我们证明了他们的吞吐量上界，然后表明，这个上界更好地估计最坏情况下的吞吐量比所有以前提出的吞吐量估计和规模比他们中的大多数。使用这个上限，我们表明，扩展器为基础的拓扑结构，不像克洛斯，超过一定规模的网络，没有拓扑结构可以有完整的吞吐量，即使它有完整的平分带宽;事实上，即使是相对较小的扩展器为基础的拓扑结构无法实现完整的吞吐量。我们的结论表明，使用吞吐量来评估数据中心的性能，而不是平分带宽可以改变在以前的工作中的结论，数据中心的成本，可靠性和可靠性。
While prior work has explored many proposed datacenter designs, only two designs, Clos-based and expander-based, are generally considered practical because they can scale using commodity switching chips. Prior work has used two different metrics, bisection bandwidth and throughput, for evaluating these topologies at scale. Little is known, theoretically or practically, how these metrics relate to each other. Exploiting characteristics of these topologies, we prove an upper bound on their throughput, then show that this upper bound better estimates worst-case throughput than all previously proposed throughput estimators and scales better than most of them. Using this upper bound, we show that for expander-based topologies, unlike Clos, beyond a certain size of the network, no topology can have full throughput, even if it has full bisection bandwidth; in fact, even relatively small expander-based topologies fail to achieve full throughput. We conclude by showing that using throughput to evaluate datacenter performance instead of bisection bandwidth can alter conclusions in prior work about datacenter cost, manageability, and reliability.