This experiment is designed for the evaluation of various ML frameworks on the client. For this purpose, we use a single client, and we change its backend.
kubectl frisbee submit test fedbed- examples/federated-learning/1.ml-backend.yml ./charts/system/ charts/federated-learning/fedbedThis experiment is designed for the evaluation of resource heterogeneity. For this purpose, we use multiple clients and assign the total resources to clients according to a distribution.
kubectl frisbee submit test fedbed- examples/federated-learning/2.resource-distribution.yml ./charts/system/ charts/federated-learning/fedbedThis experiment is designed for the evaluation of client placement across nodes. In this case, we only use the placement primitives, without any kind of resource throttling
kubectl frisbee submit test fedbed- examples/federated-learning/3.client-placement.yml ./charts/system/ charts/federated-learning/fedbedThis experiment is designed for the evaluation of dataset heterogeneity. For this purpose, we use multiple clients and split the dataset to clients according to a distribution.
kubectl frisbee submit test fedbed- examples/federated_learning/4.dataset-distribution.yml ./charts/system/ charts/federated-learning/fedbed
# Parallel Workflows
This experiment is designed for the evaluation of parallel workflows on same clients.
For this purpose, we run two workflows with controllable interference
```shell
kubectl frisbee submit test fedbed- examples/federated_learning/6.parallel-workflows.yml ./charts/system/ charts/federated-learning/fedbed