{"title":"Fault-Tolerant Clock Synchronization with High Precision","authors":"Attila Kinali, F. Huemer, C. Lenzen","doi":"10.1109/ISVLSI.2016.88","DOIUrl":null,"url":null,"abstract":"We present the first FPGA implementation of a distributed clock synchronization algorithm with sub-nanosecond skews that can tolerate arbitrary faults of individual components. Each of n nodes is equipped with its own quartz oscillator and the nodes broadcast their clock pulses to enable synchronization. The algorithm provably maintains synchronization even if fewer than n/3 nodes exhibit arbitrary faulty behavior. Moreover, aslong as more than 2n/3 nodes remain synchronized, nodes will recover and resynchronize after transient faults. Using 4 boards with Cyclone IV FPGAs, our implementation achieves precision better than 300 ps. This is in accordance with the worst-case precision of 870 ps predicted by theory. Furthermore, our experiments demonstrate that nodes recover from transient faults as described above. Finally, frequency stability of the overall system improved by an order of magnitude.","PeriodicalId":140647,"journal":{"name":"2016 IEEE Computer Society Annual Symposium on VLSI (ISVLSI)","volume":"67 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2016-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"8","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2016 IEEE Computer Society Annual Symposium on VLSI (ISVLSI)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ISVLSI.2016.88","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 8
Abstract
We present the first FPGA implementation of a distributed clock synchronization algorithm with sub-nanosecond skews that can tolerate arbitrary faults of individual components. Each of n nodes is equipped with its own quartz oscillator and the nodes broadcast their clock pulses to enable synchronization. The algorithm provably maintains synchronization even if fewer than n/3 nodes exhibit arbitrary faulty behavior. Moreover, aslong as more than 2n/3 nodes remain synchronized, nodes will recover and resynchronize after transient faults. Using 4 boards with Cyclone IV FPGAs, our implementation achieves precision better than 300 ps. This is in accordance with the worst-case precision of 870 ps predicted by theory. Furthermore, our experiments demonstrate that nodes recover from transient faults as described above. Finally, frequency stability of the overall system improved by an order of magnitude.
我们提出了分布式时钟同步算法的第一个FPGA实现,该算法具有亚纳秒偏差,可以容忍单个组件的任意故障。n个节点中的每一个都配备了自己的石英振荡器,节点广播它们的时钟脉冲以实现同步。该算法可以证明,即使出现任意错误行为的节点少于n/3个,该算法也能保持同步。只要保持同步状态的节点数大于2n/3,节点会在短暂故障后恢复并重新同步。使用4块电路板和Cyclone IV fpga,我们的实现精度优于300ps,这符合理论预测的最坏情况精度870ps。此外,我们的实验表明,节点从上述瞬态故障中恢复。最后,整个系统的频率稳定性提高了一个数量级。