Talking to an OpenStack Cluster in Plain English
summary
Describes operating a 130-node OpenStack private cloud (including integration with Ceph) and a research project that automatically builds a knowledge graph from OpenStack system states and logs. Presents a natural-language interface that queries that graph to retrieve cross-module state information in plain English, and a simple anomaly-detection component that indicates likely causes of problems.
AI-written summary · about
chapters (17)
- 0:00Intro
- 1:47Our production cluster and Open Compute Lab
- 2:21We maintain the systems to make the the applications smarter
- 2:55Data center automation
- 5:32Trouble shooting for end users
- 6:27How many rules needs to know as a professional openstack operator?
- 8:53Key idea 2: automatically discover knowledge in systems using most basic rules
- 14:19Data source used to construct the graph
- 15:14How to construct the graph?
- 17:25System query as graph traversal
- 18:14Support queries in natural language!
- 20:09Building the natural language pipeline
- 20:21Model Training: labeling
- 21:16Detecting Slots in the questions
- 21:59Automatic anomaly detections: the general ideas
- 22:24A simple anomaly example: VM fails to start
- 23:05A more complicated case: matching of two properties
related talks
18mWhat is OpenStack?
30mOpenStack Works, so now what?
49mIntroduction to OpenStack
31mIntro: OpenStack SIG
23mOpen Matters: HP & OpenStack
45mBring OpenStack & Ceph to the Edge with Control Plane Services Managed by Kubernetes
31mIntro: OpenStack SIG
19m10 years of OpenStack at CERN (From 0 to 300k cores)
