00-setup-workspace.md

Setup the Azure Synapse Analytics (ASA) workspace

Task 1 - Create resources

1. Create a new resource group

2. In the resource group, create a regular blob storage account. Create in it two private containers named staging and models.

3. In the resource group, create an empty ASA workspace.

4. Create the following file systems in the primary storage account of the workspace: dev, staging, and wwi.

5. Create a linked service to the first pool of the workspace. Configure it to connect with username and password and use the workspace's SQL admin account credentials. It is recommended to name the linked service sqlpool01 to simplify importing datasets, data flows, and pipelines later.

6. Create a linked service to the primary storage account. Configure it to connect with the account key. It is recommended to name the linked service asadatalake01 to simplify importing datasets, data flows, and pipelines later.

7. Create a linked service to the blob storage account. Configure it to connect with the account key. It is recommended to name the linked service asastore01 to simplify importing datasets, data flows, and pipelines later.

8. For the remainder of this guide, the following terms will be used for various ASA-related resources (make sure you replace them with actual names and values):

   ASA resource	To be referred to as
   Workspace resource group	WorkspaceResourceGroup
   Workspace	Workspace
   Identity used to create Workspace	MasterUser
   Primary storage account	PrimaryStorage
   Blob storage account	BlobStorage
   First Spark pool	SparkPool01
   First SQL pool	SQLPool01
   SQL admin account	asa.sql.admin
   Linked service to first SQL pool	sqlpool01
   Linked service to primary storage account	asadatalake01
   Linked service to blob storage account	asastore01

9. Ensure the Workspace security principal (which has the same name as the Workspace) and the MasterUser (the one used to create the Workspace) are added with the Storage Blob Data Owner role to the PrimaryStorage.

Task 2 - Upload the data used in the lab

1. Create a folder named bronze in the dev file system of PrimaryStorage.

2. Upload the following data files to the bronze folder created above:

   File name	Size	Download from
   postalcodes.csv	1.8 MB	https://solliancepublicdata.blob.core.windows.net/wwi-01/postalcodes.csv
   wwi-dimcity.csv	17.5 MB	https://solliancepublicdata.blob.core.windows.net/wwi-01/wwi-dimcity.csv
   wwi-dimcustomer.csv	67.3 KB	https://solliancepublicdata.blob.core.windows.net/wwi-01/wwi-dimcustomer.csv
   wwi-dimdate.csv	146.3 KB	https://solliancepublicdata.blob.core.windows.net/wwi-01/wwi-dimdate.csv
   wwi-dimemployee.csv	18.7 KB	https://solliancepublicdata.blob.core.windows.net/wwi-01/wwi-dimemployee.csv
   wwi-dimpaymentmethod.csv	514 B	https://solliancepublicdata.blob.core.windows.net/wwi-01/wwi-dimpaymentmethod.csv
   wwi-dimstockitem.csv	114.6 KB	https://solliancepublicdata.blob.core.windows.net/wwi-01/wwi-dimstockitem.csv
   wwi-dimsupplier.csv	3.8 KB	https://solliancepublicdata.blob.core.windows.net/wwi-01/wwi-dimsupplier.csv
   wwi-dimtransactiontype.csv	1.3 KB	https://solliancepublicdata.blob.core.windows.net/wwi-01/wwi-dimtransactiontype.csv
   wwi-factmovement.csv	11.1 MB	https://solliancepublicdata.blob.core.windows.net/wwi-01/wwi-factmovement.csv
   wwi-factorder.csv	31.3 MB	https://solliancepublicdata.blob.core.windows.net/wwi-01/wwi-factorder.csv
   wwi-factpurchase.csv	461.1. KB	https://solliancepublicdata.blob.core.windows.net/wwi-01/wwi-factpurchase.csv
   wwi-factsale-big-1.csv	3.0 GB	https://solliancepublicdata.blob.core.windows.net/wwi-01/wwi-factsale-big-1.csv
   wwi-factsale-big-2.csv	3.0 GB	https://solliancepublicdata.blob.core.windows.net/wwi-01/wwi-factsale-big-2.csv
   wwi-factsale-big-3.csv	3.0 GB	https://solliancepublicdata.blob.core.windows.net/wwi-01/wwi-factsale-big-3.csv
   wwi-factsale-big-4.csv	3.0 GB	https://solliancepublicdata.blob.core.windows.net/wwi-01/wwi-factsale-big-4.csv
   wwi-factsale.csv	1.8 GB	https://solliancepublicdata.blob.core.windows.net/wwi-01/wwi-factsale.csv
   wwi-factstockholding.csv	8.9 KB	https://solliancepublicdata.blob.core.windows.net/wwi-01/wwi-factstockholding.csv
   wwi-facttransaction.csv	7.2 MB	https://solliancepublicdata.blob.core.windows.net/wwi-01/wwi-facttransaction.csv

Task 3 - Import datasets, data flows, and pipelines

Import datasets pointing to PrimaryStorage

Perform the following steps for each dataset to be imported:

1. Create a new, empty dataset with the same name as the one to be imported.

2. Switch to code view and replace the code with the content of the associated JSON file.

3. If the name used for the linked service to PrimaryStorage is not asadatalake01, replace the properties.linkedServiceName.referenceName value in JSON with the actual name of the linked service.

4. Save and publish the dataset. Optionally, you can publish all datasets at once, at the end of the import procedure.

The following datasets pointing to PrimaryStorage must be imported:

Dataset	Source code
external_postalcode_adls	external_postalcode_adls.json
staging_enrichedcustomer_adls	staging_enrichedcustomer_adls.json
wwi_dimcity_adls	wwi_dimcity_adls.json
wwi_dimcustomer_adls	wwi_dimcustomer_adls.json
wwi_dimdate_adls	wwi_dimdate_adls.json
wwi_dimemployee_adls	wwi_dimemployee_adls.json
wwi_dimpaymentmethod_adls	wwi_dimpaymentmethod_adls.json
wwi_dimstockitem_adls	wwi_dimstockitem_adls.json
wwi_dimsupplier_adls	wwi_dimsupplier_adls.json
wwi_dimtransactiontype_adls	wwi_dimtransactiontype_adls.json
wwi_factmovement_adls	wwi_factmovement_adls.json
wwi_factorder_adls	wwi_factorder_adls.json
wwi_factpurchase_adls	wwi_factpurchase_adls.json
wwi_factsale_adls	wwi_factsale_adls.json
wwi_factsale_big_1_adls	wwi_factsale_big_1_adls.json
wwi_factsale_big_2_adls	wwi_factsale_big_2_adls.json
wwi_factsale_big_3_adls	wwi_factsale_big_3_adls.json
wwi_factsale_big_4_adls	wwi_factsale_big_4_adls.json
wwi_factstockholding_adls	wwi_factstockholding_adls.json
wwi_facttransaction_adls	wwi_facttransaction_adls.json

Import datasets pointing to SQLPool1

Perform the following steps for each dataset to be imported:

1. Create a new, empty dataset with the same name as the one to be imported.

2. Switch to code view and replace the code with the content of the associated JSON file.

3. If the name used for the linked service to SQLPool01 is not sqlpool01, replace the properties.linkedServiceName.referenceName value in JSON with the actual name of the linked service.

4. Save and publish the dataset. Optionally, you can publish all datasets at once, at the end of the import procedure.

The following datasets pointing to SQLPool01 must be imported:

Dataset	Source code
wwi_dimcity_asa	wwi_dimcity_asa.json
wwi_dimcustomer_asa	wwi_dimcustomer_asa.json
wwi_dimdate_asa	wwi_dimdate_asa.json
wwi_dimemployee_asa	wwi_dimemployee_asa.json
wwi_dimpaymentmethod_asa	wwi_dimpaymentmethod_asa.json
wwi_dimstockitem_asa	wwi_dimstockitem_asa.json
wwi_dimsupplier_asa	wwi_dimsupplier_asa.json
wwi_dimtransactiontype_asa	wwi_dimtransactiontype_asa.json
wwi_factmovement_asa	wwi_factmovement_asa.json
wwi_factorder_asa	wwi_factorder_asa.json
wwi_factpurchase_asa	wwi_factpurchase_asa.json
wwi_factsale_asa	wwi_factsale_asa.json
wwi_factstockholding_asa	wwi_factstockholding_asa.json
wwi_facttransaction_asa	wwi_facttransaction_asa.json
wwi_perf_factsale_fast_asa	wwi_perf_factsale_fast_asa.json
wwi_perf_factsale_slow_asa	wwi_perf_factsale_slow_asa.json
wwi_staging_dimcustomer_asa	wwi_staging_dimcustomer_asa.json
wwi_staging_enrichedcustomer_asa	wwi_staging_enrichedcustomer_asa.json

Import data flows

Perform the following steps for each data flow to be imported:

1. Create a new, empty data flow with the same name as the one to be imported.

2. Switch to code view and replace the code with the content of the associated JSON file.

3. Save and publish the data flow. Optionally, you can publish all data flows at once, at the end of the import procedure.

The following data flows must be imported:

Data flow	Source code
EnrichCustomerData	EnrichCustomerData.json

Import pipelines

Perform the following steps for each pipeline to be imported:

1. Create a new, empty pipeline with the same name as the one to be imported.

2. Switch to code view and replace the code with the content of the associated JSON file.

3. Save and publish the pipeline. Optionally, you can publish all pipelines at once, at the end of the import procedure.

The following pipelines must be imported:

Pipeline	Source code
Exercise 2 - Enrich Data	Exercise 2 - Enrich Data.json
Import WWI Data	Import WWI Data.json
Import WWI Data - Fact Sale Full	Import WWI Data - Fact Sale Full.json
Import WWI Perf Data - Fact Sale Fast	Import WWI Perf Data - Fact Sale Fast.json
Import WWI Perf Data - Fact Sale Slow	Import WWI Perf Data - Fact Sale Slow.json

Task 4 - Populate PrimaryStorage with data

1. Import the Setup - Export Sales to Data Lake notebook.

2. Replace <primary_storage> with the actual data lake account name of PrimaryStorage in cells 1, 4, and 6.

3. Run the notebook to populate PrimaryStorage with data.

Task 5 - Configure the SQL on-demand pool

1. Create a SQL on-demand database running the following script on the master database of the SQL on-demand pool:

   CREATE DATABASE SQLOnDemand01

2. Ensure the SQL on-demand pool can query the storage account using the following script:

   CREATE CREDENTIAL [https://<primary_storage>.dfs.core.windows.net]
   WITH IDENTITY='User Identity';

   In the script above, replace <primary_storage> with the name of PrimaryStorage.

Task 6 - Configure SQLPool01

1. Connect with either the SQL Active Directory admin or the asa.sql.admin account to SQLPool01 using the tool of your choice.

2. Run the WWI-Reset-SQLPoolSchema SQL script to initialize the schema of the SQL pool.

3. Create the asa.sql.staging login in the master database of Workspace using the following script:

   CREATE LOGIN [asa.sql.staging]
   WITH PASSWORD = '<password>'
   GO

   In the script above, replace <password> with the actual password of the login.

4. Create the asa.sql.staging user in SQLPool01 using the following script:

   CREATE USER [asa.sql.staging]
       FOR LOGIN [asa.sql.staging]
       WITH DEFAULT_SCHEMA = dbo
   GO
   
   -- Add user to the required roles
   
   EXEC sp_addrolemember N'db_datareader', N'asa.sql.staging'
   GO
   
   EXEC sp_addrolemember N'db_datawriter', N'asa.sql.staging'
   GO
   
   EXEC sp_addrolemember N'db_ddladmin', N'asa.sql.staging'
   GO

5. Configure access control to workspace pipeline runs in SQLPool01 using the following script:

   --Create user in DB
   CREATE USER [<workspace>] FROM EXTERNAL PROVIDER;
   
   --Granting permission to the identity
   GRANT CONTROL ON DATABASE::<sqlpool> TO [<workspace>];

   In the script above, replace <workspace> with the actual name of Workspace and <sqlpool> with the actual name of SQLPool01.

6. Run the Import WWI Data pipeline to import all data except the sale facts into SQLPool01.

7. Run the Import WWI Data - Fact Sale Full pipeline to import the sale facts into SQLPool01.

8. Run the Import WWI Perf Data - Fact Sale Fast and Import WWI Perf Data - Fact Sale Slow pipelines to import the large-sized sale facts into SQLPool01.

Task 7 - Configure Power BI

1. Ensure the MasterUser has a Power BI Pro subscription assigned.

2. Sign in to the Power BI portal using the credentials of MasterUser and create a new Power BI workspace. In the remainder of this guide, this workspace will be referred to as PowerBIWorkspace.

3. Perform all the steps described in Exercise 3 - Task 1. In step 11, instead of using the suggested naming convention, name your dataset wwifactsales.

4. In the Power BI portal, edit the security settings of the wwifactsales dataset and configure it to authenticate to SQLPool01 using the credentials of the asa.sql.admin account. This allows the Direct Query option to work correctly for all participants in the lab.

Task 8 - Import all SQL scripts and Spark notebooks

Import the following SQL scripts into Workspace:

SQL script name	Source code	Replacements
Exercise 1 - Read with SQL on-demand	Exercise 1 - Read with SQL on-demand.sql	<primary_storage> with the actual name of PrimaryStorage
Exercise 4 - Analyze Transactions	Exercise 4 - Analyze Transactions.sql	None
Exercise 4 - Investigate query performance	Exercise 4 - Investigate query performance.sql	None
Exercise 5 - Create Sample Data for Predict	Exercise 5 - Create Sample Data for Predict.sql	None
Exercise 5 - Predict with model	Exercise 5 - Predict with model.sql	None
Exercise 5 - Register model	Exercise 5 - Register model.sql	<blob_storage_account_key> with the storage account key of BlobStorage; <blob_storage> with the storage account name of BlobStorage

Import the following Spark notebooks into Workspace:

Spark notebook name	Source code	Replacements
Exercise 1 - Read with Spark	Exercise 1 - Read with Spark.ipynb	<primary_storage> with the actual name of PrimaryStorage
Exercise 2 - Ingest Sales Data	Exercise 2 - Ingest Sales Data.ipynb	In cell 1 - <primary_storage> with the actual name of PrimaryStorage
Exercise 2 - Bonus Notebook with CSharp	Exercise 2 - Bonus Notebook with CSharp.ipynb	In cell 1 - <primary_storage> with the actual name of PrimaryStorage; In cell 3 - <sql_staging_password> with the password of asa.sql.staging created above in Task 4, step 3; In cell 3 - <workspace> with the name of the Workspace; In cell 3 - <sql_pool> with the name of SQLPool1

Task 9 - Prepare a machine learning model

Prepare the models container in BlobStorage by creating two folders: onnx and hex.

To prepare the machine learning model for Exercise 5, you have two options:

• Use the already trained and converted machine learning model (available as a starter artifact)
• Train and convert a new machine learning model

Import the already trained and converted machine learning model

1. Upload the model.onnx.hex file to the hex folder in the models container of BlobStorage.

2. Run the Exercise 5 - Create Sample Data for Predict SQL script to create sample data for machine learning predictions.

3. Run the Exercise 5 - Register model SQL script to register the model with the SQLPool01 SQL pool.

Train and convert a new machine learning model

1. Run the Exercise 5 - Model training Spark notebook to train the machine learning model and save it in ONNX format. The model will be saved as model.onnx in the onnx folder in the models container of BlobStorage.

2. Use the convertion PowerShell script to transform model.onnx into model.onnx.hex.

3. Perform steps 1, 2, and 3 described in the previous section.

Task 10 - Configure additional users to access the workspace

For each additional user that needs to have access to Workspace and run exercises 1 through 5, the following steps must be performed:

1. Assign the Reader role on the WorkspaceResourceGroup to the user.

2. Assign the Reader and Blob Data Contributor roles on the PrimaryStorage to the user.

3. Assign the Workspace admin role in the Workspace to the user.

4. Grant access to the SQLPool01 to the user using the script below. You must be signed in with the MasterUser credentials (use SQL Server Management Studio if the script fails in Synapse Studio).

   CREATE USER [<user_principal_name>] FROM EXTERNAL PROVIDER;
   EXEC sp_addrolemember 'db_owner', '<user_principal_name>';

   In the script above, replace <user_principal_name> with Azure Active Directory user principal name of the user.

5. Assign the Contributor role on the Power BI workspace of the MasterUser created in Task 5, step 2.

Task 11 - Create and configure an Azure Databricks workspace

1. In the resource group, create a new Azure Databricks workspace.

2. In the Azure Databricks workspace, create a new cluster with the following configuration:

3. On the newly created cluster, install the onnxmltools PyPi library:

4. Import the following Databricks notebooks into the Shared section of the Databricks workspace:

Spark notebook name	Source code	Replacements
Exercise 5 - Model Training	Exercise 5 - Model Training.dbc