SLATE-PTP: A Spine–Leaf Architecture Timestamps and Evaluation Dataset for IEEE 1588 Precision Time Protocol Benchmarking

SLATE-PTP (Spine–Leaf Architecture Timestamps and Evaluation Dataset) provides high-resolution IEEE 1588 Precision Time Protocol (PTP) telemetry gathered from an experimental physical 2-Spine 4-Leaf datacenter testbed operating with PTP-unaware switches. Dataset Highlights & Methodology: Oscillator Drift Decoupling: The capturing network interface card (Intel E810) was disciplined using an external GNSS PPS hardware reference, while the LinuxPTP (ptp4l) daemon executed in a free-running state (no clock adjustments). This isolates pure network dynamics, Packet Delay Variation (PDV), and link asymmetries from local clock drift. Realistic Traffic Injection: Synthetic background traffic was generated using TRex across four packet distributions (IMIX, MTU, Small Packets, Large Packets) and three load dynamics (Constant, Instantaneous Steps, Progressive Ramps). Structured Raw Telemetry: Includes nanosecond-granularity CSV files containing sequential IEEE 1588 timestamp records (t1, t2, t3, t4) and message metadata. Benchmarking Sandbox: Accompanied by ptp_pipeline.ipynb, an object-oriented Python processing pipeline allowing researchers to ingest data tracks and test custom offset estimators, Kalman filters, or ML delay classifiers without dedicated hardware. Licensing Notice: Telemetry CSV data files are released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license. Software, Python processing pipelines, and Jupyter Notebooks are released under the MIT License.