A technique to estimate a system's asymptotic delay and throughput
A technique to estimate a system's asymptotic delay and throughput
复制标题
估计系统渐近延迟和吞吐量的技术
DOI：
10.1145/3199524.3199549
复制
发表时间：
2018
期刊：
影响因子：
--
通讯作者：
Y. Tay
中科院分区：
文献类型：
--
作者：
Y. Tay
计算机系统的性能通常有两个渐近参数：当工作负载较低时的最小延迟R0和当工作负载较高时的最大吞吐量X1。这些参数很重要，因为它们可以用于网络连接中的拥塞控制，或者在视频游戏的图形引擎之间进行选择。R0和X1的估计并不简单：硬件可能处于黑盒中，软件可能以复杂的方式进行交互，并且估计取决于工作负载。这篇短文提出了一种使用统计回归的技术，以及一个来自统计分析的方程，来估计R0和X1。
The performance of a computer system usually has two asymptotic parameters: its minimum latency R0 when workload is low, and its maximum throughput X1 when workload is high. These parameters are important because they can be used for, say, congestion control in a network connection, or choosing between graphics engines for a video game. The estimation of R0 and X1 is not straightforward: the hardware may be in a blackbox, the software may interact in complicated ways, and the estimates depend on the workload. This short paper proposes a technique for using statistical regression, and an equation from queueing analysis, to estimate R0 and X1.