Can Big Data and Machine Learning Really Improve Network Operations?
summary
A practical roadmap for adding analytics to operations: segment the problem (for example start with DNS logs or three DNS servers), collect passive data types like flows, pcaps and logs, store in Postgres or ELK or cloud, and prototype with Python and scikit-learn (avoid deep learning). Recommend batch processing first, verify predictions in a trust loop, expect staffing/training and modest costs, and observe reduced firefighting time though fully autonomous networks remain years away.
AI-written summary · about
chapters (12)
- 0:00Intro
- 0:12Interop DIGITAL
- 0:31What this talk is about
- 0:55Network Analytics to Automation
- 3:48Analytics/Machine Learning checklist
- 4:02Clearly defined usecases
- 6:00Collect the right data
- 8:31Storing and processing the data
- 11:14Machine Learning Framework/Approach/Algorithms
- 14:25What do you need to consider ?
- 19:52Intangibles ...
- 22:42To wrap up...
related talks
10mQuick Look: A Real-World Example of Using Network Behaviour to Detect Cyber-Compromise
25mAutomating Operations with ML
44mDevOps for Machine Learning - why is it different?
42mTales from the Field: The Future is Now
49mA Real-World Example of Using Network Behaviour to Detect Cyber-Compromise
26mFrom Data to Decision: Monitoring and Improving Water Networks using AI/ML
13mPredictive Insights for IT Operations #bbuzz
09mAI/ML Won't Save you