Posible Timezone Discrepancy in Ground Truth Fault Timestamps [PPTX]
Hey! First of all, thanks for publishing this dataset , it's being super useful for my multimodal RCA model.
I wanted to flag something I ran into while working with the data: there's a timezone offset between the fault timestamps in some of the PPTX slides (where the groundtruth is if Im not wrong) and the actual metrics and log data.
The PPTX documentation uses JST (Japan Standard Time)
The metric and log files use UTC (Can I assume it?)
If Im not wrong, this creates a 9-hour offset
Example from scenario 20231221:
The PPTX slide explicitly states:
"Measurement period: 2023/12/19 12:02 (JST) ~ 2023/12/20 18:00 (JST)"
"Failure time: 2023/12/20 11:30 (JST)"
But if you use 11:30 JST directly as the fault timestamp, you won't find the anomaly in the data - because the metrics are in UTC.
I think the correct fault time in the data is actually 02:30 UTC (11:30 JST - 9 hours).
I spent a while debugging why my ground truth wasn't aligning with visible anomalies in the metrics until I figured this out 😅
Just wanted to share it, to double check with you guys and in case others run into the same issue.
Cheers and thanks again!
Hi! Thanks a lot for the detailed note — really appreciate you taking the time to dig into this and share the finding.
You’re absolutely right. The timestamps in the PPT slides are in JST, since the real testbed is located in Japan and the failure time was recorded locally. Meanwhile, all metric and log data are collected and stored in UTC. So the 9-hour offset you identified is expected, and your conversion (e.g., 11:30 JST → 02:30 UTC) is correct.
We’ll update the README to make this clearer for everyone.
Thanks again for pointing this out — this will definitely help others avoid confusion. Glad to hear the dataset has been useful for your multimodal RCA model!
Cheers!