http:^^www.cs.washington.edu^research^projects^lis^chaos^www^presentation.html
来自「This data set contains WWW-pages collect」· HTML 代码 · 共 69 行
HTML
69 行
Date: Tue, 10 Dec 1996 03:21:29 GMTServer: NCSA/1.4.2Content-type: text/html<html><head> <title>Presentation of Simulation Results </title></head><body><!WA0><img src=http://www.cs.washington.edu/research/projects/lis/chaos/www/chaos.xbm> <p><h1> <a name="top">Presentation of Simulation Results</a><br></h1><p><hr> <p>Because the format of presentation of simulation (or experimental) results innetwork routing studies varies greatly, we present these pages as asuggestion of standard presentation formats. The formats were inspired by discussion at <!WA1><a href=http://www.cs.washington.edu/research/projects/lis/chaos/www/pcrcw.html> PCRCW '94 </a>.There were two main issues discussed:<ul><li> Units used to label the X-axis <li> General form of the presented graph(s)</ul><p><hr><p><h1> Graph Format </h1>Two formats were suggested for graphing results. Paired throughput vs. applied load and latencyvs. applied load comprise <!WA2><a href=http://www.cs.washington.edu/research/projects/lis/chaos/www/cnf.html> Chaos Normal Form.</a> A single latency vs. achieved throughput graph makes <!WA3><a href=http://www.cs.washington.edu/research/projects/lis/chaos/www/bnf.html> Burton Normal Form.</a>Use of only latency (including source queueing) vs. applied load is discouraged because it is impossible to gain any data about performance above saturation using such graphs.<p><hr><p><h1> Units for Bandwidth/Load </h1>The literature describing results typically includes graphs of latency or throughputvs. applied load. Many units have been used to express this, including flits injected/cycle,flits/ns, bits/cycle, normalized bandwidth, and many others. Because the many formsrequire conversions to compare with other results, we propose using a single form: <em> normalizedbandwidth. </em><p>Normalized bandwidth simply expresses the load/throughput as a fraction of the bisection-bandwidthlimited maximum bandwidth of the network for uniform random traffic. Essentially, this constraint,corresponding to normalized throughput = 1.0, is derived by considering that 50% of uniform randomtraffic crosses the bisection of the network. Thus, if a network has bisection bandwidth B bits/sec, each node in an N-node network can inject 2B/N bits/sec at the maximum load. An optimal routing algorithm could handle such a load without before saturating.<p>Non-uniform or non-random traffic patterns may saturate at different loads than uniform random traffic.However, please normalize to the uniform random traffic limit.<hr><!WA4><a href="http://www.cs.washington.edu/research/projects/lis/chaos/www/chaos.html"><!WA5><img src=http://www.cs.washington.edu/research/projects/lis/chaos/www/chaos.xbm></a> Back to the Chaotic Routing Home Page <p><p><address>kwb@cs.washington.edu</address></html>⌨️ 快捷键说明
复制代码Ctrl + C
搜索代码Ctrl + F
全屏模式F11
增大字号Ctrl + =
减小字号Ctrl + -
显示快捷键?