Performance evaluation of NTPv4 when SPIK state cannot step frequency

Abstract

Network time protocol version 4 (NTPv4) is widely used to keep the time of NTPv4 client machine to be synchronized with the NTPv4 time server. NTPv4 does not adjust only the time but it also adjusts the frequency of the clock on client machine correspondingly to the NTPv4 server. In the NTPv4 specification, when time difference on NTPv4 client and server is greater than a threshold for a period of time, the NTPv4 is able to largely adjust the frequency at once. This action is called frequency stepping. However, in the real NTPv4 implementation, the frequency stepping was removed under this condition. This paper studies the effects of this missing action focusing on the time that NTPv4 client spends to adjust its clock frequency. The experiment was done under the scenario that NTPv4 client had to readjust its frequency. The results show that without frequency stepping, the NTPv4 spends time noticeably long to readjust its frequency. This ensures that NTPv4 should still use the frequency stepping for frequency readjustment in the real implementation.