Hugging Face's logo

Datasets:
logfit-project
/
HDFS_v1

Tasks:
Text Classification
Modalities:
Tabular
Text
Formats:
parquet
Languages:
English
Size:
10M - 100M
Libraries:
Datasets
Dask
License:
Dataset card Data Studio Files
xet
Community
1
HDFS_v1 / README.md
calmodovar's picture
calmodovar
Update dataset card
da85a01 verified
preview code
|
raw
history blame contribute delete
2.31 kB
metadata
pretty_name: HDFS_v1
dataset_name: logfit-project/HDFS_v1
task_categories:
  - text-classification
language:
  - en
size_categories:
  - 10M<n<20M
annotations_creators:
  - logfit-project
license: other

Dataset Card for logfit-project/HDFS_v1

Dataset Summary

The HDFS v1 log dataset captures Hadoop Distributed File System (HDFS) console logs that were collected from a private cloud deployment while benchmark workloads were executed. Each log line can be associated with one or more block identifiers; block-level anomaly labels were generated by manually crafted rules. This script preserves the raw line structure while attaching a binary anomaly flag for downstream anomaly detection research.

Supported Tasks and Leaderboards

  • anomaly-detection: binary classification of logs.

Dataset Structure

  • date: Six-digit date stamp from the original console output (YYMMDD).
  • time: Six-digit time stamp (HHMMSS).
  • pid: Process identifier extracted from the log line.
  • level: Log level (e.g., INFO).
  • component: Java/daemon component emitting the log entry.
  • content: Verbose message content for the event.
  • block_id: Space-separated block identifiers discovered in the log line (empty if none present).
  • anomaly: Binary indicator derived from block-level labels (1 = anomalous block).

Source Data

Dataset Creation

The raw logs were parsed in a streaming fashion using a deterministic regular expression so that large-scale HDFS deployments can be transformed without exhausting memory. Block-level labels are joined on the fly by searching for block identifiers in each line.

Uses

Suitable for supervised and semi-supervised anomaly detection across distributed system logs, log template mining, and benchmarking log representation learning.

Citation

  • Wei Xu, Ling Huang, Armando Fox, David Patterson, Michael Jordan. "Detecting Large-Scale System Problems by Mining Console Logs", SOSP 2009.
  • Jieming Zhu, Shilin He, Pinjia He, Jinyang Liu, Michael R. Lyu. "Loghub: A Large Collection of System Log Datasets for AI-driven Log Analytics", ISSRE 2023.

Dataset Statistics

  • Number of log lines: 11175629