Run a genomics analysis in a JupyterLab notebook on Dataproc

---

This tutorial shows you how to run a single-cell genomics analysis using Dask, NVIDIA RAPIDS, and GPUs, which you can configure on Dataproc. You can configure Dataproc to run Dask either with its standalone scheduler or with YARN for resource management.

This tutorial configures Dataproc with a hosted JupyterLab instance to run a notebook featuring a single-cell genomics analysis. Using a Jupyter Notebook on Dataproc lets you combine the interactive capabilities of Jupyter with the workload scaling that Dataproc enables. With Dataproc, you can scale out your workloads from one to many machines, which you can configure with as many GPUs as you need.

This tutorial is intended for data scientists and researchers. It assumes that you are experienced with Python and have basic knowledge of the following:

• Dataproc
• Dask
• RAPIDS
• Jupyter Notebooks

Objectives

• Create a Dataproc instance which is configured with GPUs, JupyterLab, and open source components.
• Run a notebook on Dataproc.

Costs

In this document, you use the following billable components of Google Cloud:

Dataproc

Cloud Storage

GPUs

To generate a cost estimate based on your projected usage, use the pricing calculator. New Google Cloud users might be eligible for a free trial.

When you finish the tasks that are described in this document, you can avoid continued billing by deleting the resources that you created. For more information, see Clean up.

Before you begin

1. In the Google Cloud console, on the project selector page, select or create a Google Cloud project.

   Note: If you don't plan to keep the resources that you create in this procedure, create a project instead of selecting an existing project. After you finish these steps, you can delete the project, removing all resources associated with the project.

   Go to project selector

2. Make sure that billing is enabled for your Google Cloud project.

3. Enable the Dataproc API.

   Enable the API

Prepare your environment

1. Select a location for your resources.

   REGION=REGION
   

2. Create a Cloud Storage bucket.

   gcloud storage buckets create gs://BUCKET --location=REGION
   

3. Copy the following initialization actions to your bucket.

   SCRIPT_BUCKET=gs://goog-dataproc-initialization-actions-REGION
   gcloud storage cp ${SCRIPT_BUCKET}/gpu/install_gpu_driver.sh BUCKET/gpu/install_gpu_driver.sh
   gcloud storage cp ${SCRIPT_BUCKET}/dask/dask.sh BUCKET/dask/dask.sh
   gcloud storage cp ${SCRIPT_BUCKET}/rapids/rapids.sh BUCKET/rapids/rapids.sh
   gcloud storage cp ${SCRIPT_BUCKET}/python/pip-install.sh BUCKET/python/pip-install.sh
   

Create a Dataproc cluster with JupyterLab and open source components

1. Create a Dataproc cluster.

gcloud dataproc clusters create CLUSTER_NAME \
    --region REGION \
    --image-version 2.0-ubuntu18 \
    --master-machine-type n1-standard-32 \
    --master-accelerator type=nvidia-tesla-t4,count=4 \
    --initialization-actions
BUCKET/gpu/install_gpu_driver.sh,BUCKET/dask/dask.sh,BUCKET/rapids/rapids.sh,BUCKET/python/pip-install.sh
\
    --initialization-action-timeout=60m \
    --metadata
gpu-driver-provider=NVIDIA,dask-runtime=yarn,rapids-runtime=DASK,rapids-version=21.06,PIP_PACKAGES="scanpy==1.8.1,wget" \
    --optional-components JUPYTER \
    --enable-component-gateway \
    --single-node

The cluster has the following properties:

• --region: the region where your cluster is located.
• --image-version: 2.0-ubuntu18, the cluster image version
• --master-machine-type: n1-standard-32, the main machine type.
• --master-accelerator: the type and count of GPUs on the main node, four nvidia-tesla-t4 GPUs.
• --initialization-actions: the Cloud Storage paths to the installation scripts that install GPU drivers, Dask, RAPIDS, and extra dependencies.
• --initialization-action-timeout: the timeout for the initialization actions.
• --metadata: passed to the initialization actions to configure the cluster with NVIDIA GPU drivers, the standalone scheduler for Dask, and RAPIDS version 21.06.
• --optional-components: configures the cluster with the Jupyter optional component.
• --enable-component-gateway: allows access to web UIs on the cluster.
• --single-node: configures the cluster as a single node (no workers).

Access the Jupyter Notebook

1. Open the Clusters page in the Dataproc Google Cloud console.
   Open Clusters page
2. Click your cluster and click the Web Interfaces tab.
3. Click JupyterLab.
4. Open a new terminal in JupyterLab.

5. Clone the clara-parabricks/rapids-single-cell-examples repository and check out the dataproc/multi-gpu branch.

   git clone https://github.com/clara-parabricks/rapids-single-cell-examples.git
   git checkout dataproc/multi-gpu
   

6. In JupyterLab, navigate to the rapids-single-cell-examples/notebooks repository and open the 1M_brain_gpu_analysis_uvm.ipynb Jupyter Notebook.

7. To clear all the outputs in the notebook, select Edit > Clear All Outputs

8. Read the instructions in the cells of the notebook. The notebook uses Dask and RAPIDS on Dataproc to guide you through a single-cell RNA-seq workflow on 1 million cells, including processing and visualizing the data. To learn more, see Accelerating Single Cell Genomic Analysis using RAPIDS.

Clean up

To avoid incurring charges to your Google Cloud account for the resources used in this tutorial, either delete the project that contains the resources, or keep the project and delete the individual resources.

Delete the project

Caution: Deleting a project has the following effects:

• Everything in the project is deleted. If you used an existing project for the tasks in this document, when you delete it, you also delete any other work you've done in the project.
• Custom project IDs are lost. When you created this project, you might have created a custom project ID that you want to use in the future. To preserve the URLs that use the project ID, such as an appspot.com URL, delete selected resources inside the project instead of deleting the whole project.

If you plan to explore multiple architectures, tutorials, or quickstarts, reusing projects can help you avoid exceeding project quota limits.

1. In the Google Cloud console, go to the Manage resources page.

   Go to Manage resources

2. In the project list, select the project that you want to delete, and then click Delete.
3. In the dialog, type the project ID, and then click Shut down to delete the project.

Delete individual resources

1. Delete your Dataproc cluster.

   gcloud dataproc clusters delete cluster-name \
       --region=region
   

2. Delete the bucket:

   gcloud storage buckets delete BUCKET_NAME

   Important: Your bucket must be empty before you can delete it.

What's next

• Discover more about Dataproc.
• Explore reference architectures, diagrams, tutorials, and best practices. Take a look at our Cloud Architecture Center.