| 引用本文: | 宛考,江勇,徐恪.集中控制混合网络中基于流的资源分配算法.[J].国防科技大学学报,2018,40(6):89-104.[点击复制] |
| WAN Kao,JIANG Yong,XU Ke.Towards a flow based resource scheduling algorithm in center control hybrid network[J].Journal of National University of Defense Technology,2018,40(6):89-104[点击复制] |
|
| |
|
|
| 本文已被：浏览 6214次 下载 5819次 |
| 集中控制混合网络中基于流的资源分配算法 |
| 宛考, 江勇, 徐恪 |
|
(清华大学 计算机科学与技术系, 北京 100084)
|
| 摘要: |
| 集中控制混合网络中，异构化网络内部的流量具有一定的规律和特性，如果使用原有的单一的离散式最大极值和无状态的网络资源调度算法，忽略了异构网络规律，会造成网络利用率较低、易震荡、部分网络流延迟等问题。通过分析由集中控制网络和普通网络组成的混合网络的拓扑结构，对混合网络结构中常见问题如流闪现、不能估计的流、路径堵塞或连接震荡场景进行分析，并提出基于期望和状态的流量评价资源规划算法POS和POS-FME。算法考虑混合网络的运行状态，对系统可用资源进行评估，为系统中各种流匹配对应可用资源，并具有一定的预测作用，从而避免混合网络出现运行效率低下的场景。通过实验，POS算法和POS-FME算法相对传统算法，利用率提高了10%~30%，并降低了震荡和平均延迟。 |
| 关键词: 混合网络 资源分配 流分析 流调度 优化算法 |
| DOI：10.11887/j.cn.201806013 |
| 投稿日期：2017-09-27 |
| 基金项目:国家自然科学基金资助项目(61170292, 61140454)；国家科技重大专项基金资助项目(2012ZX03005001，2014ZX03002004)；国家重点基础研究发展计划资助项目(2012CB315803)；国家863计划资助项目(2013AA013302) |
|
| Towards a flow based resource scheduling algorithm in center control hybrid network |
| WAN Kao, JIANG Yong, XU Ke |
|
(Department of Computer Science and Technology, Tsinghua University, Beijing 100084, China)
|
| Abstract: |
| In the hybrid center control networks, the flows of isomerization networks have its own regularity and characteristic. While ignoring the regularity of the flows in isomerization networks, the original single and stateless network resource scheduling algorithms which discretely maximize value can cause the problems such as the low utilization, easier to concuss, some delayed network traffic. The topological architecture of the hybrid network which combines normal network with centered intelligent controllable network was analyzed, specially aimed to the common problems such as the flow burst，unpredictable flow，path jam/connection shocking scene. Then the resource scheduling algorithm named POS and POS-FME based on expectation and status were proposed. The algorithms consider the running status of hybrid network, evaluate the available resources of the system, and match available resources for various flows in the system, and have a certain prediction effect. The algorithms avoid the scenes of low running efficiency in the hybrid network, improve the utilization rate by about 10%~30% im comparison with the traditional algorithm, and also reduce the concussion and average delay through experiments. |
| Keywords: hybrid network resource scheduling flow analyses flow scheduling optimization algorithm |
|
|
|
|
|
