Study on distributed internal synchronization in a LAN network

Abstract

To meet rising demands of synchronization precision in distributed nodes in LAN network, a specialized scheme is proposed based on the intermediate-level modeling for initial synchronization and periodical resynchronizations with basic concurrency control and tight bounds. Several representative threads are designed and well scheduled with definite and thorough functional partitioning inside to create a fast and efficient circumstance for communication between nodes. An experimental realization based on RT Linux kernel and corresponding measurements are accomplished. The results indicate that in a small-scale distributed environment, the nodes are able to be well synchronized during in the long run with hundreds of resynchronizations and the deviations are significantly limited in a reasonable range. The proposed scheme effectively promotes the performance of clock synchronization in the LAN network.