feat: Add instructions in the README

terraform/cross-device/README.md

@@ -1,13 +1,53 @@
 # Cross-device Federated Learning
 
-This module is an example of an end to end demo for cross-device Federated Learning
+This module is an example of an end to end demo for cross-device Federated Learning. This example deploys 6 different workloads:
+- `aggregator`: this is an offline job that reads device gradients and calculates aggregated result with DP
+- `collector`: this is an offline job that runs periodically to query active task and encrypted gradients, resulting in deciding when to kick off aggregating
+- `modelupdater`: this is an offline job that listens to events and publishes results so that device can download
+- `task-assignment`: this is a front end service that distributes training tasks to devices
+- `task-management`: this is an offline job that manages tasks
+- `task-scheduler`: this is an offline job that either runs periodically or is triggered by some events
+
+This example builds on top of the infrastructure that the
+[blueprint provides](../../../../README.md), and follows the best practices the
+blueprint establishes.
+
+## Prerequisites
+
+- A POSIX-compliant shell
+- Git (tested with version 2.41)
+- Docker (tested with version 20.10.21)
 
 ## Infrastructure
+
 It creates:
 - A spanner instance for storing the status of training
 - Pubsub topics that act as buses for messages between microservices
 - Buckets for storing the trained models
 
-To deploy this solution and ensure end-to-end confidentiality, you need to enable confidential nodes.
+To deploy this solution, just set the `cross-device` flag to `true`.
+
+To ensure end-to-end confidentiality, you need to enable confidential nodes.
+
 However, it is also necessary to use VM families that support this feature, such as **N2D** or **C2D**.
-When using confidential nodes, set `enable_confidential_nodes` to `true` and `cluster_tenant_pool_machine_type` to `n2d-standard-8`. In addition, in order to have the minimum number of replicas required during deployment, you need at least 4 nodes and set `cluster_tenant_pool_min_nodes` to `4`.
+When using confidential nodes, set `enable_confidential_nodes` to `true` and `cluster_tenant_pool_machine_type` to `n2d-standard-8`. In addition, in order to have the minimum number of replicas required during deployment, you need at least 4 nodes.
+
+You will then deploy the cross-device workloads in a namespace. You will need to set the `tenant_namespace` variable with the name of the namespace in which you want to deploy the workloads.
+
+### Containers running in different namespaces, in the same GKE cluster
+
+1. Provision infrastructure by following the instructions in the [main README](../../../../README.md).
+1. From Cloud Shell, change the working directory to the `terraform` directory.
+1. Initialize the following Terraform variables:
+
+    ```hcl
+        enable_confidential_nodes         = true
+        cluster_tenant_pool_machine_type  = "n2d-standard-4"
+        cluster_default_pool_machine_type = "n2d-standard-4"
+        cross-device                      = true
+        tenant_namespace                  = "main"
+    ```
+
+1. Run `terraform apply`, and wait for Terraform to complete the provisioning process.
+1. Open the [GKE Workloads Dashboard](https://cloud.google.com/kubernetes-engine/docs/concepts/dashboards#workloads)
+    and wait for the workers Deployments and Services to be ready.