This dataset contains low level packet and packet timing data, collected from Network Time Protocol (NTP) request-response pairs sent across the public Internet between a timing testbed located within FEIT at UTS, and 479 timeservers (all but 6 public) spread globally.

The servers targeted were Stratum-1, meaning they are synchronized locally to reference hardware such as an atomic clock or GPS. They are therefore of particular interest as they anchor the NTP timing system. We believe that this set covers a high proportion of the global set of public facing Stratum-1 timeservers at the time of collection.

The data collection covered a 2 month period bridging the Leap Second event occurring at the last second of 2016. For each server a record was collected periodically over the period, typically once per second. The dataset contains a total of some 5.5 billion records, and will be made available both on a per-server, and complete, basis. Each record consists of 6 fields. In column order these are:

LI value	[ in [0-3] ]
Incoming server timestamp	[ UTC ]
Outgoing server timestamp	[ UTC ]
Stratum value	[ in [0-16] ]
Local incoming timestamp	[ GPS timescale ]
Local outgoing timestamp	[ GPS timescale ]

An extensive description of the dataset and its collection is provided in [1], including how to convert between the UTC and GPS timescales.

In addition to the data itself, this collection makes available additional metadata, as well as results of studies performed so far based on the data. These add value to the data collection itself as results are given per-server. Details are provided in the 'readme' sheet of the spreadsheet file

TimeServer_Dataset2016-17_perserver_results.xlsx

that contains the metadata, accessible via the provided URL. The studies all relate to analyzing the performance of the servers, toward the goal of promoting a discipline of Server Health Monitoring (SHM).

Furthermore, the Matlab based `VATtools' (Visualization, Analysis and Tagging) that were used to analyze the data are also included, as well as the detailed server anomaly `tagging' files, which contain detailed information about each anomaly discovered within the data, and enable it to be visualized using the tool. Documentation is provided in VATtool_readme.txt .

[1] Yi, C. & Veitch, D. 2018, 'Network Timing, Weathering the 2016 Leap Second', Proceedings of the The 37th Annual IEEE International Conference on Computer Communications, IEEE INFOCOM 2018.