• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文学位 >Consistency based fault-tolerant time synchronization in a distributed system.
【24h】

Consistency based fault-tolerant time synchronization in a distributed system.

机译：分布式系统中基于一致性的容错时间同步。

获取原文
获取原文并翻译 | 示例
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

The problem addressed is fault tolerant time synchronization of clocks in a distributed system. A distributed system is a set of nodes that communicate by sending messages. Each node contains a physical clock and executes processes to exchange messages and to time synchronize the clocks. The local time of a node is calculated by reading the physical clock and adding a correction. Time synchronization consists of determining the corrections to the clocks so that the clock offsets (clock differences) are as small as possible.;Based on the bound on the frequency offsets of the non-faulty clocks, error bounds are calculated for the estimated clock offsets and round-trip delays. It is shown how these error bounds are used to determine if a clock is consistent.;This dissertation presents an algorithm, called the group method (GM), that, when possible, detects and corrects faulty estimated clock offsets. The clock offset between any two clocks is measured by exchanging the time via messages over links in a network. The clock information that is measured are the clock offsets, and the round-trip delays. Once the clock information is measured between all pairs of clocks, all of the clock information is distributed to each node for processing. Finally, GM is used to determine which estimated clock offsets are faulty by separating clocks that consistently report their time from those clocks that do not.;GM requires that: (1) the frequency offsets of the non-faulty clocks are bounded, (2) that there are more than 3m clocks, where m is the number of faulty clocks, and (3) that all the nodes are connected. It is not required that the clocks be initially time synchronized. But, if the clocks are time synchronized, then a bound on the magnitude of the clock offsets can be calculated.;To achieve fault tolerant time synchronization means that the clock offsets between all of the non-faulty clocks are bounded, and that the frequency offsets of the corrected clocks are bounded. The fault detection and correction are based on the consistency of the estimated clock offsets. A clock is consistent if the estimated clock offsets agree with the fact that the frequency offsets of the physical clocks are bounded.;After the faulty estimated clock offsets are removed or are corrected, the estimated clock offsets are then used to adjust the clock corrections. It is shown that, if the error bounds are non-zero, then the clock offsets can be bounded, but not eliminated, after the adjustments to the corrections of the clocks.
机译：解决的问题是分布式系统中时钟的容错时间同步。分布式系统是一组通过发送消息进行通信的节点。每个节点都包含一个物理时钟，并执行过程以交换消息并进行时钟时间同步。节点的本地时间是通过读取物理时钟并添加校正来计算的。时间同步包括确定对时钟的校正，以使时钟偏移（时钟差）尽可能小。；基于无故障时钟的频率偏移的边界，计算估计时钟偏移的误差边界和往返延迟。展示了如何使用这些误差范围来确定时钟是否一致。本文提出了一种称为组方法（GM）的算法，该算法在可能的情况下可以检测并纠正错误的估计时钟偏移。任何两个时钟之间的时钟偏移是通过网络中的链接通过消息交换时间来测量的。测量的时钟信息是时钟偏移和往返延迟。一旦在所有时钟对之间测量了时钟信息，就将所有时钟信息分配给每个节点进行处理。最后，GM通过将一致报告其时间的时钟与那些没有报告时间的时钟分开来确定哪些估计时钟偏移是有缺陷的; GM要求：（1）非故障时钟的频率偏移是有界的，（2 ）的时钟数超过3m，其中m是故障时钟数，并且（3）所有节点都已连接。不需要时钟最初是时间同步的。但是，如果时钟是时间同步的，则可以计算出时钟偏移量的界限。;要实现容错时间同步，则意味着所有非故障时钟之间的时钟偏移量是有界的，并且频率校正后的时钟的偏移量是有界的。故障检测和纠正基于估计的时钟偏移的一致性。如果估算的时钟偏移量与物理时钟的频率偏移量有界这一事实相符，则时钟是一致的。;在错误的估算时钟偏移量被消除或校正后，估算的时钟偏移量将用于调整时钟校正。结果表明，如果误差范围不为零，则在对时钟的校正进行调整之后，可以限制时钟偏移量，但是不能消除时钟偏移量。

著录项

  • 作者

    Monington, Kenneth William.;

  • 作者单位

    University of Delaware.;

  • 授予单位 University of Delaware.;
  • 学科 Electrical engineering.
  • 学位 Ph.D.
  • 年度 1995
  • 页码 89 p.
  • 总页数 89
  • 原文格式 PDF
  • 正文语种 eng
  • 中图分类
  • 关键词

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号