Started with the implementation. Request for Feedback.
Goal For Next Week: Unsupervised Learning.
|2||Failure Emulation Update|
No Breakthrough Yet.
Still Trying with stress-ng (interrupts, system-calls, ...)
|3||FP Model Development Update|
models/failure_prediction/ *.ipynb, models/failure_prediction/static_htmls, models/failure_prediction/data_preprocessing.
Rohit presented his work as part of the 'Final-Presentation' of his Internship.
Beth: Were the results expected?
Steve: Out of 26 features, only 6 features are used. This choice is based on?
Girish: Going ahead, we will try with better data and improve these models.
Rohit: Continue to contribute to Thoth.
Target date for Whitepaper: October 20th.
|4||Data Extraction Tool Status||Stuck at Prometheus (tested), Elasticsearch(untested).|
|5||Synthetic Data Generation - GANs|
Girish L et. al.,
Generate Monitoring/logging data using GANs. This is a new project to create synthetic data for testing of operational ML algorithms. The idea is to create this data set to be used across the industry as a reference data set.
Hemashree, Vidyashree, Renuka
|6||Exploration: Openstack Log Analysis|
Existing Approaches: ELK - Kibana + Alarms.
Analysis: Common approach is using NLP.
Approach: Google's BERT model for Openstack Logs.
Problems: Anomaly Detection + Pattern Analysis
Deekshita and Rakshita - NLP, Try with Openstack Logs (??)
Use of NLP for Openstack Logs has already been tried: https://www.cs.utah.edu/~lifeifei/papers/deeplog.pdf . This work will be used as a reference.