# Kubernetes Meetup Report

Kubernetes is a tool to automate the management of containerized applications like Cassandra. The structure of the

[Overview of k8ssandra](https://medium.com/building-the-open-data-stack/integrating-apache-cassandra-and-kubernetes-through-k8ssandra-33efdf2692ef)

Operators make complex databases work on kubernetes.
Database replicas are asymmetric, databases need a more sophisticated replica model.
Operators need adaptation to handle dbms failures

Patrick McFadin
[e-mail]

Kubernetes scheduler
- reconciles the state, basically assigns resources based on the yaml file
- alternative schedulers :
yunikorn, designed for spark
You can run multiple schedulers in kubernetes, specify which to use in the yaml file.
Volcano, designed for high performance computing, recommended for non spark projects

Dask, Ray, feast (feature store),

jasleen

operators make complex databases work on kubernetes
creating using kubernetes stateful set databseshave to come up in order to fix so you don't have to go to the file system

attached local storage vs network only with local cache
similar to a family of starge which increases elasticyt

## Running AI/ML Workloads in Kubernetes
### Patrick McFadin, DataStax @PatrickMcFaden
wrote oriely book - Cloud Native Data on Kubernetes
- working on moving form cassandra to k8ssandra
- yunikorn : another scheduler that's not the default, good scheduler for spark
- fairness algorithm - fixes noisy neightbor issue
- can run multiple schedulers (specify in yaml)
- volcano scheduler - use this for anything other than spark

dask - tool to connect to kubernetes and run tests