Lab 3: Explore Big Data using Azure Databricks

In this lab you will use Azure Databricks to explore the New York Taxi data files you saved in your data lake in Lab 2. Using a Databricks notebook you will connect to the data lake and query taxi ride details.

The estimated time to complete this lab is: 45 minutes.

Microsoft Learn & Technical Documentation

The following Azure services will be used in this lab. If you need further training resources or access to technical documentation please find in the table below links to Microsoft Learn and to each service's Technical Documentation.

Azure Service	Microsoft Learn	Technical Documentation
Azure Databricks	Perform data engineering with Azure Databricks	Azure Databricks Technical Documentation
Azure Data Lake Gen2	Large Scale Data Processing with Azure Data Lake Storage Gen2	Azure Data Lake Gen2 Technical Documentation
Azure Synapse Analytics	Implement a Data Warehouse with Azure Synapse Analytics	Azure Synapse Analytics Technical Documentation

Lab Architecture

Step	Description
	Build an Azure Databricks notebook to explore the data files you saved in your data lake in the previous exercise. You will use Python and SQL commands to open a connection to your data lake and query data from data files.
	Integrate datasets from Azure Synapse Analytics data warehouse to your big data processing pipeline. Databricks becomes the bridge between your relational and non-relational data stores.

IMPORTANT: Some of the Azure services provisioned require globally unique name and a “-suffix” has been added to their names to ensure this uniqueness. Please take note of the suffix generated as you will need it for the following resources in this lab:

Name	Type
synapsedatalakesuffix	Storage Account
ADPDatabricks-suffix	Databricks Workspace
synapsesql-suffix	SQL server

IMPORTANT: The code snippets below illustrate the simplest and quickest way to establish connections between Databricks and other Azure services. They ARE NOT considered best practices as they expose secrets and passwords in plain text. For a secure implementation following the security best practices, please consider the use of Azure Key Vault in conjuntion with Databricks Secret Scopes (https://docs.azuredatabricks.net/user-guide/secrets/secret-scopes.html).

Create the NYCTaxiData-Curated Container on Azure Data Lake Storage Gen2

In this section you will create a container in your SynapseDataLake that will be used as a repository for curated Parquet files for the Spark table that will be generated by your Databricks notebook. These files contain cleansed and validated data ready to be used for reporting.

IMPORTANT
Execute these steps on your host computer

In the Azure Portal, go to the lab resource group and locate the Azure Storage account synapsedatalakesuffix.
On the Overview panel, click Containers.
On the synapsedatalakesuffix – Countainers blade, click + Container. On the New container blade, enter the following details:
- Name: nyctaxidata-curated
- Public access level: Private (no anonymous access)
Click OK to create the new container.

Create Azure Databricks Cluster

In this section you are going to create an Azure Databricks cluster that will be used to execute notebooks.

IMPORTANT
Execute these steps on your host computer

In the Azure Portal, navigate to the lab resource group and locate the Azure Databricks resource ADPDatabricks-suffix.
On the ADPDatabricks-suffix blade, click the Launch Workspace button. The Azure Databricks portal will open on a new browser tab.
On the Azure Databricks portal, click the Clusters button on the left-hand side menu.
On the Clusters blade, click + Create Cluster.
On the Create Cluster blade, enter the following connection details:
- Cluster Name: ADPDatabricksCluster
- Terminate after: 30 minutes
- Min Workers: 1
- Max Workers: 2

Leave all other fields with their default values.
Click Create Cluster. It should take around 5 to 7 minutes for the cluster to be fully operational.

Create an Azure Databricks Notebook

In this section you are going to create an Azure Databricks notebook that will be used to explore the taxi data files you copied to your data lake in the Lab 2.

IMPORTANT
Execute these steps on your host computer

On the Azure Databricks portal, click the Home button on the left-hand side menu.
On the Workspace blade, click the down arrow next to your user name and then click Create > Notebook.
On the Create Notebook pop-up window type “NYCTaxiData” in the Name field.
Ensure you have the Language field set to Python and the Cluster field is set to ADPDatabricksCluster.
Click Create.
On the Cmd 1 cell, you will invoke the Spark API to establish a connection to your SynapseDataLake storage account. For this you will need to retrieve the name and the key of your SynapseDataLake storage account from the Azure Portal.
Use the Python code below and replace synapsedatalake[suffix] with your own suffix on line 22 and to replace [your synapsedatalake account key] with the data lake account key on line 23.

#Setup connection to SynapseDataLake storage account

def MountDataLakeContainer(datalakeAccountName, datalakeAccountKey, containerName, mountPointName):
    dataLakeAccountDNS = datalakeAccountName + ".blob.core.windows.net"
    
    dataLakeAuthTypeConfig = "fs.azure.account.auth.type." + dataLakeAccountDNS
    dataLakeKeyConfig = "fs.azure.account.key." + dataLakeAccountDNS
    
    dataLakeExtraConfig = {dataLakeAuthTypeConfig:"SharedKey"
                           , dataLakeKeyConfig:datalakeAccountKey}
    
    containerUri = "wasbs://" + containerName + "@" + dataLakeAccountDNS
    mountPointFolder = "/mnt/" + mountPointName

    try:
        dbutils.fs.mount(source = containerUri, mount_point = mountPointFolder, extra_configs = dataLakeExtraConfig)
    except Exception as e:
        if "Directory already mounted" in str(e):
            pass # Ignore error if already mounted.
        else:
            raise e
    return containerUri

#Set Account Name and Key Variable Values
dataLakeAccountName = 'synapsedatalake[suffix]' #<-- Your synapsedatalake account name goes here. Remember to replace <suffix> with your own suffix.
dataLakeAccountKey = '[your synapsedatalake account key]' #<-- Your synapsedatalake account key goes here.

#Mount NYCTaxiData-Raw and NYCTaxiData-Curated Containers
MountDataLakeContainer(dataLakeAccountName, dataLakeAccountKey, 'nyctaxidata-raw', 'raw')
MountDataLakeContainer(dataLakeAccountName, dataLakeAccountKey, 'nyctaxidata-curated', 'curated')

Press Shift + Enter to execute and create a new notebook cell.
In the Cmd 2 cell, define a new StructType object that will contain the definition of the data frame schema.
Using the schema defined in the previous step, initialise a new data frame by invoking the Spark API to read the contents of the nyctaxidata-raw container in the SynapseDataLake storage account. Use the Python code below:

#Define NYCTaxiData schema and load data into a Data Frame

from pyspark.sql.types import *

nycTaxiDataSchema = StructType([
  StructField("VendorID",IntegerType(),True)
  , StructField("tpep_pickup_datetime",DateType(),True)
  , StructField("tpep_dropoff_datetime",DateType(),True)
  , StructField("passenger_count",IntegerType(),True)
  , StructField("trip_distance",DoubleType(),True)
  , StructField("RatecodeID",IntegerType(),True)
  , StructField("store_and_fwd_flag",StringType(),True)
  , StructField("PULocationID",IntegerType(),True)
  , StructField("DOLocationID",IntegerType(),True)
  , StructField("payment_type",IntegerType(),True)
  , StructField("fare_amount",DoubleType(),True)
  , StructField("extra",DoubleType(),True)
  , StructField("mta_tax",DoubleType(),True)
  , StructField("tip_amount",DoubleType(),True)
  , StructField("tolls_amount",DoubleType(),True)
  , StructField("improvement_surcharge",DoubleType(),True)
  , StructField("total_amount",DoubleType(),True)])
  
dfNYCTaxiData = spark.read.format('csv').options(header='true').schema(nycTaxiDataSchema).load('/mnt/raw/*.csv')
dfNYCTaxiData.cache()

Hit Shift + Enter to execute the command and create a new cell.
In the Cmd 3 cell, call the display function to show the contents of the data frame dfNYCTaxiData. Use the Python code below:

#Display Data Frame Content

display(dfNYCTaxiData)

Hit Shift + Enter to execute the command and create a new cell. You will see a data grid showing the top 1000 records from the dataframe:
In the Cmd 4 cell, call the select() method of the data frame object to select the columns "tpep_pickup_datetime", "passenger_count" and "total_amount". Then use the filter() method to filter rows where "passenger_count > 6" and "total_amount > 50.0". Use the Python code below:

# Use Data Frame API Operations to Filter Data

display(dfNYCTaxiData.select("tpep_pickup_datetime", "passenger_count", "total_amount").filter("passenger_count > 6 and total_amount > 50.0"))

Hit Shift + Enter to execute the command and create a new cell.
In the Cmd 5 cell, call the createOrReplaceTempView method of the data frame object to create a temporary view of the data in memory. Use the Python code below:

# Create Local Temp View

dfNYCTaxiData.createOrReplaceTempView('NYCTaxiDataTable')

Hit Shift + Enter to execute the command and create a new cell.
In the Cmd 6 cell, change the cell language to SQL using the %sql command.
Write a SQL query to retrieve the total number of records in the NYCTaxiDataTable view. Use the command below:

%sql
--Use SQL to count NYC Taxi Data records

select count(*) from NYCTaxiDataTable

Hit Shift + Enter to execute the command and create a new cell. You will see the total number of records in the data frame at the bottom of the cell.
In the Cmd 7 cell, write a SQL query to filter taxi rides that happened on the Apr, 7th 2019 that had more than 5 passengers. Use the command below:

%sql
-- Use SQL to filter NYC Taxi Data records

select cast(tpep_pickup_datetime as date) as pickup_date
  , tpep_dropoff_datetime
  , passenger_count
  , total_amount
from NYCTaxiDataTable
where cast(tpep_pickup_datetime as date) = '2019-04-07'
  and passenger_count > 5

Hit Shift + Enter to execute the command and create a new cell. You will see a grid showing the filtered result set.
In the Cmd 8 cell, write a SQL query to aggregate records and return total number of rides by payment type. Use the command below:

%sql
-- Use SQL to aggregate NYC Taxi Data records and visualize data
select case payment_type
            when 1 then 'Credit card'
            when 2 then 'Cash'
            when 3 then 'No charge'
            when 4 then 'Dispute'
            when 5 then 'Unknown'
            when 6 then 'Voided trip'
        end as PaymentType
  , count(*) as TotalRideCount
from NYCTaxiDataTable
group by payment_type
order by TotalRideCount desc

Hit Shift + Enter to execute the command and create a new cell. Results will be displayed in a grid in the cell.
Click the Bar chart button to see results as a bar chart.
Using Python, open a JDBC connection to your Azure Synapse Analytics and load Taxi location lookup data from the Staging.NYCTaxiLocationLookup table into a new data frame called dfLocationLookup.

IMPORTANT: Don't forget to replace the 'synapsesql-suffix' with your specific Azure Synapse Analytics server name.

In the same cell, create a temporary view called "NYCTaxiLocation" and display the contents of the data frame. Use the Python code below:

# Load Taxi Location Data from Azure Synapse Analytics

jdbcUrl = "jdbc:sqlserver://synapsesql-suffix.database.windows.net:1433;database=SynapseDW" #Replace "suffix" with your own  
connectionProperties = {
  "user" : "adpadmin",
  "password" : "P@ssw0rd123!",
  "driver" : "com.microsoft.sqlserver.jdbc.SQLServerDriver"
}

pushdown_query = '(select * from NYC.TaxiLocationLookup) as t'
dfLookupLocation = spark.read.jdbc(url=jdbcUrl, table=pushdown_query, properties=connectionProperties)

dfLookupLocation.createOrReplaceTempView('NYCTaxiLocation')

display(dfLookupLocation)

Hit Shift + Enter to execute the command and create a new cell. Results will be displayed in a grid in the cell.
In the Cmd 10 cell, write a SQL query to cleanse data and apply standard column definitions. You should joining the two dataframes using their view names and filter out any invalid data. The resulting dataset should be saved in a Spark table using Parquet files sitting in the NYCTaxiData-Curated container in your SynapseDataLake storage account. Use the SQL command below:

%sql
--Create Spark table with cleansed, validated dataset combining raw CSV files and reference data from you Synapse data warehouse
--Data will be saved into the NYCTaxiData-Curated container in your SynapseDataLake account

create table if not exists NYCTaxiData_Curated
using parquet partitioned by (PickUpYearMonth) --Use Parquet format partitioned by YYYY-MM
options ('compression'='snappy') --Parquet compression options
location '/mnt/curated/' --Data to be saved in the NYCTaxiData-Curated container in your SynapseDataLake storage account.
as
select 
    VendorID
    , cast(tpep_pickup_datetime as date) as PickUpDate
    , concat(year(tpep_pickup_datetime), '-', format_string('%02d',month(tpep_pickup_datetime),'##')) as PickUpYearMonth --Partition Key
    , cast(tpep_pickup_datetime as timestamp) as PickUpDateTime
    , cast(tpep_dropoff_datetime as date) as DropOffDate
    , cast(tpep_dropoff_datetime as timestamp) as DropOffDateTime
    , passenger_count as PassengerCount
    , trip_distance as TripDistance
    , cast(PULocationID as int) as PickUpLocationID
    , pu.Zone as PickUpLocationZone
    , pu.Borough as PickUpLocationBorough
    , cast(DOLocationID as int) as DropOffLocationID
    , do.Zone as DropOffLocationZone
    , do.Borough as DropOffLocationBorough
    , cast(payment_type as int) as PaymentTypeID
    , case payment_type
            when 1 then 'Credit card'
            when 2 then 'Cash'
            when 3 then 'No charge'
            when 4 then 'Dispute'
            when 5 then 'Unknown'
            when 6 then 'Voided trip'
        end as PaymentTypeDescription
    , cast(case when fare_amount < 0 then 0.00 else fare_amount end as decimal(8,2)) as FareAmount --Cleanse invalid data
    , cast(case when extra < 0 then 0.00 else extra end as decimal(8,2)) as ExtraAmount --Cleanse invalid data
    , cast(case when mta_tax < 0 then 0.00 else mta_tax end as decimal(8,2)) as MTATaxAmount --Cleanse invalid data
    , cast(case when tip_amount < 0 then 0.00 else tip_amount end as decimal(8,2)) as TipAmount --Cleanse invalid data
    , cast(case when tolls_amount < 0 then 0.00 else tolls_amount end as decimal(8,2)) as TollsAmount --Cleanse invalid data
    , cast(case when improvement_surcharge < 0 then 0.00 else improvement_surcharge end as decimal(8,2)) as ImprovementSurchargeAmount --Cleanse invalid data
    , cast(case when total_amount < 0 then 0.00 else total_amount end as decimal(8,2)) as TotalRideAmount --Cleanse invalid data
from NYCTaxiDataTable as rides
  join NYCTaxiLocation as pu
    on rides.PULocationID = pu.LocationID
  join NYCTaxiLocation as do
    on rides.DOLocationID = do.LocationID
where passenger_count > 0 --Data Cleanup Rules
  and year(tpep_pickup_datetime) = 2019

Hit Shift + Enter to execute the command. Results will be displayed in a grid in the cell.
In the Cmd 11 cell, write a SQL query to retrieve the first 100 rows from the Spark table NYCTaxiData_Curated you just created in the previous step. Use the SQL command below.

%sql

--Retrieve curated dataset sample

select *
from NYCTaxiData_Curated
limit 100

Review the NYCTaxiData-Curated Container Content

In this section you are going to review the content of the NYCTaxiData-Curated container in your SynapseDataLake storage account after you created the Spark table using the Databricks notebook in the previous exercise. You should see a list of folders created, one for each partition and inside them the collection of Parquet files containing cleansed and validated taxi data ready to be used for reporting.

IMPORTANT
Execute these steps on your host computer

In the Azure Portal, go to the lab resource group and locate the Azure Storage account synapsedatalakesuffix.
On the Overview panel, click Containers.
Open the NYCTaxiData-Curated container to see its contents. You should see a folder structure similar to this one:
Inside each folder you will find the collection of individual Parquet files that make up the NYCTaxiData_Curated Spark table.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Lab3.md

Lab3.md

Lab 3: Explore Big Data using Azure Databricks

Microsoft Learn & Technical Documentation

Lab Architecture

Create the NYCTaxiData-Curated Container on Azure Data Lake Storage Gen2

Create Azure Databricks Cluster

Create an Azure Databricks Notebook

Review the NYCTaxiData-Curated Container Content

Files

Lab3.md

Latest commit

History

Lab3.md

File metadata and controls

Lab 3: Explore Big Data using Azure Databricks

Microsoft Learn & Technical Documentation

Lab Architecture

Create the NYCTaxiData-Curated Container on Azure Data Lake Storage Gen2

Create Azure Databricks Cluster

Create an Azure Databricks Notebook

Review the NYCTaxiData-Curated Container Content