网站大量收购独家精品文档,联系QQ:2885784924

Convergence Speed of Binary Interval Consensus二元区间一致性的收敛速度.ppt

Convergence Speed of Binary Interval Consensus二元区间一致性的收敛速度.ppt

  1. 1、本文档共25页,可阅读全部内容。
  2. 2、原创力文档(book118)网站文档一经付费(服务费),不意味着购买了该文档的版权,仅供个人/单位学习、研究之用,不得用于商业用途,未经授权,严禁复制、发行、汇编、翻译或者网络传播等,侵权必究。
  3. 3、本站所有内容均由合作方或网友上传,本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺!文档内容仅供研究参考,付费前请自行鉴别。如您付费,意味着您自己接受本站规则且自行承担风险,本站不退款、不进行额外附加服务;查看《如何避免下载的几个坑》。如果您已付费下载过本站文档,您可以点击 这里二次下载
  4. 4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等,请点击“版权申诉”(推荐),也可以打举报电话:400-050-0827(电话支持时间:9:00-18:30)。
查看更多
Convergence Speed of Binary Interval Consensus二元区间一致性的收敛速度.ppt

An obvious disadvantage for any real inplementation of this formula is the infinity term in the summation. Here, we use a batch of K = arrival rate times C requests, the C is set to a contant 2000. Also we substitute the infinity term with only the first 10% of the batch size. We show how I value is calculated for this simple trace and successfully detects the traffic surges. The x-axis is time in seconds, y-axis is the arrival rate. In the first 84 seconds, the arrival rate is relative small and stable. The I value measured during this period ranges from 3 to 15. After that a sudden increase in arrival causes a later high I value report at the end of next window. After this sudden change, the arrivals keep at a high level, so the I value decrease to 1~107 range. Finally, the burst endding behavior is also reported because of a high I value about 602. * In the previous example, we show that in both burst starts and ends, the I value changes significantly. So the algorithm that detects burstiness need to be coupled with arrival rate infromation to determine whether it is a beginning of a burst or ending of burst. We show our burst detection algorithm here: calculate I value for current batch Measure the arrival rate for current time Update total average arrival rate and SCV If we find significant I value change and (current_rate total_rate) then burst starts If we find significant I value change and (current_rate total_rate) then burst ends With the help of this algorithm, we could easily say that here is the burst beginning and here is the burst ending. * By using the same example, we plot the I values for successive batches as a function of elapsed time. The upper figures are instantaneous arrival rates of the two cases. The bottom shows the corresponding I value changes. The I value in the bottom left figure is relative modest (the blue line), so the bursty detection algorithm reports no burst (the green line). However, the bottom right figure shows a dra

文档评论(0)

cai + 关注
实名认证
内容提供者

该用户很懒,什么也没介绍

1亿VIP精品文档

相关文档