Exam DP-203: Data Engineering on

Microsoft Azure Master Cheat Sheet

Various modules and percentage involved in DP-203.

Skills measured

 Design and implement data storage (40-45%)

 Design and develop data processing (25-30%)
 Design and implement data security (10-15%)
 Monitor and optimize data storage and data processing (10-15%)

Data Storage:

Type of Data

Azure Storage
4 configurations options available includes

pg. 1
 Demysttify.Tech

1. Azure Blob

o Massive storage for Text and binary

2. Azure Files

o Mange files or share for cloud or on premise deployment

3. Azure Queues

o Messaging store for reliable messaging between application components

4. Azure Tables

o A NoSQL stores for schema less storage of structured data

Performance:

 Standard allows you to have any data service (Blob, File, Queue, and Table) and
uses magnetic disk drives.
 Premium limits you to one specific type of blob called a page blob and uses
solid-state drives (SSD) for storage.

Access tier:

 Hot
o When the frequent operation is data retrieved.
 Cold
o When the data is not often accessed.

Note:

 Data Lake Storage (ADLS) Gen2 can be enabled in the Azure Storage. Hierarchical
Namespace:
o The ADLS Gen2 hierarchical namespace accelerates big data analytics
workloads and enables file-level access control lists (ACLs)
 Account kind: StorageV2 (general purpose v2)
o The current offering that supports all storage types and all of the latest
features
 A storage account is a container that groups a set of Azure Storage services
together.

pg. 2
 Demysttify.Tech

Azure Blob Usage

 When we don’t have to query on the data stored

 Less cost
 Works well with images and unstructured format

What service to use for Data?

Architecture and usage of different Azure services

pg. 3
 Demysttify.Tech

Azure data bricks

 Apache Spark-based analytics platform

o Simplifies the provisioning and collaboration of Apache Spark-based

analytical solutions

 Enterprise Security

o Utilizes the security capabilities of Azure

 Integration with other Cloud Services

o Can integrate with variety of Azure data platform services and Power BI

Azure HD-Insight
 Deploy cluster of Hadoop or Storm or Spark

Azure Active Directory

 To guarantee security and manage person.
 Role and user permission to data bricks and data lake.

pg. 4
 Demysttify.Tech

Reading Data in Azure Databricks

SQL DataFrame

SELECT col_1 FROM myTable df.select(col("col_1"))

DESCRIBE myTable df.printSchema()

SELECT * FROM myTable WHERE col_1 > 0 df.filter(col("col_1") > 0)

..GROUP BY col_2 ..groupBy(col("col_2"))

..ORDER BY col_2 ..orderBy(col("col_2"))

..WHERE year(col_3) > 1990 ..filter(year(col("col_3"))) > 1990)

SELECT * FROM myTable LIMIT 10 df.limit(10)

display(myTable)(text format) df.show()

display(myTable)(html format) display(df)

Performing ETL to populate a data model

pg. 5
 Demysttify.Tech

Transformations usually performed on a dataset

 Basic Transformations

o Normalizing values
o Missing/Null data
o De-duplication
o Pivoting Data frames

 Advanced Transformations

o User Defined functions

o Joins and lookup tables
o Multiple databases

COSMOS-DB
Can Build Globally Distributed Databases with Cosmos DB, it can handle

 Document databases
 Key value stores
 Column family stores
 Graph databases

pg. 6
 Demysttify.Tech

Azure Cosmos DB indexes every field by default

Azure Cosmos DB (NoSQL)

 Scalability
 Performance
 Availability
 Programming Models

Request Units in Cosmos-DB

Request Unit (RU) for a DB

 A single RU is equivalent to 1 KB of Get request

 Creation, deletion and insertion require additional processing costing more RU.
 RU can be changed at any point of time
 Value of RU can be set via Capacity Planner
o Upload the sample JSON doc
o Define no of documents
o Minimum RU = 400
o Maximum RU = 215 thousand (If we require more throughput then a ticket
needs to raised in the Azure portal for it)

pg. 7
 Demysttify.Tech

Choosing Partition-Key
 Enable quick lookup of data
 Enable it to Auto scale when needed
 Selection of right partition key is important during development process
 Partition key is the value used to organise your data into Logical divisions.
o e.g.: In a Retail scenario
 ProductID and UserID value as a partition key is a good choice.

Note: A physical node can have 10 GB of information that means each Unique partition
Key can have 10 GB of unique values.

Creating a Cosmos-DB

1. Click on resources and create it

2. Click on Data Explorer to create a Database name and the table
3. Use New Item tab to add the values to the table
4. UDF can also be created as Stored procedures in JavaScript.

We can also create the same using Azure CLI

az account list —output table // Lists the set of Azure subscriptions that we
have

Az account set —subscription “<subscription name>”

az group list —out table // List of resource groups

export NAME=“<Azure Cosmos DB account name>”

export RESOURCE_GROUP=“<rgn>[sandbox resource group name]</rgn>”

Export LOCATION=“<location>” // Data centre location

Export DB_NAME=“Products”

Az group create —name <name> —location <location>

Az cosmosdb create —name $NAME —kind GlobalDocumentDB —resource-group $RESOURCE_GROUP

Az cosmosdb database create —name $NAME —db-name $DB_NAME —resource-group

$RESOURCE_GROUP

pg. 8
 Demysttify.Tech

Az cosmosdb collection create —collection-name “Clothing“ —partition-key-path

“/productId” —throughput 1000 - name $NAME —db-name $DB_NAME —resource-group
$RESOURCE_GROUP

After creating a COSMOSDB

 Navigate to Data Explorer

 Click on New container and Database
 A container can have multiple Databases

Cosmos DB fail over management

Cosmos DB Consistency Levels

Consistency
Guarantees
Level

Linearizability. Reads are guaranteed to return the most recent

Strong
version of an item

pg. 9
 Demysttify.Tech

Consistency
Guarantees
Level

Bounded Consistent Prefix. Reads lag behind writes by at most k prefixes

Staleness or t interval.

Consistent Prefix. Monotonic reads, monotonic writes, read-

Session
your-writes, write-follows-reads.

Updates returned are some prefix of all the updates, with no

Consistent Prefix
gaps.

Eventual Out of order reads.

 Eventual consistency provide the weakest read consistency but offer lowest
latency of both reads and writes. ‼️ 🚩

Question related to setting up latency ‼️ 🚩

What is the Latency I will have to use in order to provide the lower latency of reads and
writes ‼️ 🚩 - Eventual Consistency

COSMOS-DB takes care of consistency of data when replicated ‼️ 🚩

AZURE SQL DATABASE CONFIGURATION

 DTUs (Database Transaction Unit)
o Combined measure of Compute, storage, and IO resources
 VCores
o Enables you to configure resources independently
o Greater control over compute and storage resources
 SQL Elastic Pools ‼️ 🚩
o Relate to eDTUs.
o Enable you to buy set of compute and storage resources that are shared
among all the databases in the pool.
o Each database can use the resources they need.
 SQL Managed Instances

pg. 10
 Demysttify.Tech

o Creates a database with near 100% compatibility with the latest SQL
server.
o Useful for SQL Server customers who would like to migrate on-premises
servers instance in a “lift and shift” manner.

shell.azure.com to start Azure shell

To connect to Database
jay@Azure:~$ az configure --defaults group=ms-dp-200 sql-server=jaysql01

jay@Azure:~$ az sql db list

O/P:

jay@Azure:~$ az sql db list | jq '[.[] | {name: .name}]'

O/P:
[
{
"name": "master"
},
{
"name": "sqldbjay01"
}
]

jay@Azure:~$ az sql db show --name sqldbjay01

az sql db show-connection-string --client sqlcmd --name sqldbjay01

O/P:
"sqlcmd -S tcp:<servername>.database.windows.net,1433 -d sqldbjay01 -U
<username> -P <password> -N -l 30"

"sqlcmd -S tcp:sqldbjay01.database.windows.net,1433 -d sqldbjay01 -U jay -P “******”

-N -l 30"

SELECT name FEOM sys.tables; GO

SQL-DB does not take care of consistency of data when replicated, it needs to be done
manually. ‼️ 🚩

AZURE SQL-DW
3 types

 Enterprise DW
o Centralized data store that provides analytics and decision support
 Data Marts
o Designed for the needs of a single Team or business unit such as sales

pg. 11
 Demysttify.Tech

 Operational Data Stores

o Used as interim store to integrate real-time data from multiple sources for
additional operations on the data.

2 Architectural way of building a DW

 Bottom-Up Architecture

o Approach based on the notion of connected Data Marts

o Depends on Star Schema
o Benefit
 Start departmental Data Mart

 Top-down Architecture

o Creating one single integrated Normalized Warehouse

o Internal relational constructs follow the rules of normalization

Azure SQL-DW Advantage

 Elastic scale & performance

o Scales to petabytes of data

o Massively Parallel Processing
o Instant-on compute scales in seconds
o Query Relational / Non-Relational

 Powered by the Cloud

o Starts in minutes
o Integrated with AzureML, PowerBI & ADF
o Enterprise Ready

Azure-DW GEN-2

 Introduced Cache and tempDB to pull data from remote datasets

 Max DWU is 30Kc
 120 connections and 128 queries
 MPP

pg. 12
 Demysttify.Tech

Creation of Azure DW

 Create New resource

 DB
 SQL Data Warehouse

Using PolyBase to Load Data in Azure SQL Data Warehouse ‼️ 🚩

How PolyBase works ‼️ 🚩

The MPP engine’s integration method with PolyBase

 Azure SQLDW is a relational datawarehose store which use MPP architecture

which takes advantage of the on demand Elastic scale of Azure compute and
storage to load and process Petabytes of data
 Transfers data between SQLDW and external resource providing the fast
performance
 Faster way to access Data Nodes

PolyBase ETL for DW are

 Extract the source data into Text file

 Load the data into Azure Blob Storage / Hadoop DataLake store

pg. 13
 Demysttify.Tech

 Import the data into SQLDW staging table using PolyBase

 Transform the data (optional state)
 Insert the data into Partition tables

Create a Storage Account

- Go to Resource
Blobs

 REST-based object storage for Unstructured data.

Import the Blob file into SQL-DW

CREATE MASTER KEY;

CREATE DATABASE SCOPED CREDENTIAL AzureStorageCredential
WITH
IDENTITY = ‘jayDW’,
SECRET = ‘THE-VALUE-OF-THE-ACCESS-KEY’ -- put key1’s value here
;

CREATE EXTERNAL DATA SOURCE AzureStorage

WITH (
TYPE = HADOOP ,
LOCATION = ‘wasbs://[email protected]‘,
CREDENTIAL = AzureStorageCredential
);

CREATE EXTERNAL FILE FORMAT TextFile

WITH (
FORMAT_TYPE = DelimiteddText,
FORMAT_OPTIONS (FIELD_TERMINATOR = ‘,’)
);

— Load the data from Azure Blob storage to SQL Data Warehouse

CREATE TABLE [dbo].[StageDate]

WITH (
CLUSTERED COLUMNSTORE INDEX,
DISTRIBUTION = ROUND_ROBIN
)
AS
SELECT * FOM [dbo].[Temp];

— Create statistics on the new data

CREATE STATISTICS [DataKey] on [StageDate] ([DateKey]);

CREATE STATISTICS [Quarter] on [StageDate] ([DateKey]);
CREATE STATISTICS [Month] on [StageDate] ([Month]);

pg. 14
 Demysttify.Tech

Import the Blob file into SQL-DW (Alternative)

pg. 15
 Demysttify.Tech

Check Ingest Polybase in Data warehouse ‼️ 🚩

Data Streams

pg. 16
 Demysttify.Tech

ORCHESTRATING DATA MOVEMENT WITH ADF AND

SECURING AZURE DATA PLATFORMS

Azure Event Hubs:

 Is a highly scalable publish-subscribe service that can invest millions of events per
second and stream them into multiple applications

pg. 17
 Demysttify.Tech

 A Event hub is a cloud-based event service capable of receiving and assessing

millions of events per second.
 An Event is a small packet of information, a datagram that contain a notification.
 Events can be published individually or in batch.
 Single Publication or batch count can exceed 256KB.

Create Event Hub

 Navigate to Entities
 Event Hub
 Shared Access policies
o Policy will generate Primary key and Secondary key and the connection
string

Configure Application to use Event Hubs

Azure Stream Analytics Workflow

pg. 18
 Demysttify.Tech

Azure Data Factory - ADF

 Creates, orchestrates, and automates the movement, transformation and/or
analysis of data through in the cloud.

The Data Factory Process

 Connect & collect

 Transform & Enrich
 Publish
 Monitor

Azure Data Factory Components

pg. 19
 Demysttify.Tech

Azure Data Factory Contributor Role

 Create, edit, and delete factories and child resources including datasets, linked
services, pipelines, triggers, and integration runtimes.
 Deploy Resource Manager Templates. Resource Manager Deployment is the
deployment method used by Data Factory in the Azure portal.
 Manage App Insights alerts for a data factory
 At the resource group level or above, lets users deploy Resource Manager
Template.
 Create support tickets.

Linked Services

Linked services are much like connection strings, which define the connection
information needed for Data Factory to connect to external resources.

pg. 20
 Demysttify.Tech

Linked Service Example

Data Sets

pg. 21
 Demysttify.Tech

Time Slicing Data

Data Factory Activities

pg. 22
 Demysttify.Tech

Activities within ADF defines the actions that will be performed on the data and there
are three categories including:

 Data movement activities

o Simply move data from one data store to another.
o A common example of this is in using Copy Activity.
 Data transformation activities
o Use compute resource to change or enhance data through transformation,
or it can call a compute resource to perform an analysis of the data
 Control Activities
o Orchestrate pipeline activities that includes chaining activities in a
sequence, branching, defining parameters at the pipeline level, and
passing arguments while invoking the pipeline on-demand or from a
trigger

Pipelines

 Pipeline is a grouping of logically related activities.

 Pipeline can be scheduled so the activities within it get executed.
 Pipeline can be managed and monitored.

Working with documents programmatically

 Create Storage Account

 Create ADF
 Create data workflow pipeline
 Add data bricks workbook to pipeline
 Perform analysis on the data

Network Security
Securing your network from attacks and unauthorized access is an important part of any
architecture.

pg. 23
 Demysttify.Tech

Identity and Access (Azure Active Directory (AD))

Encryption

pg. 24
 Demysttify.Tech

Azure Key-Vaults (2 Ques) ‼️ 🚩

 It is a centralised cloud service for storing your application secrets

 Provides secure access capability
 Key management can be done

Different Keys available

 RSA
 EC

Managing Encryption
Databases stores information that is sensitive, such as physical address, email address,
and phone numbers. The following is used to protect this data:

pg. 25
 Demysttify.Tech

Azure Data Lake Storage Gen2 Security Features

 Role Based Access Control

 Posix Complaint ACL
 AAD Oauth 2.0 Token
 Azure Services Integration

MONITORING, TROUBLESHOOTING DATA STORAGE

AND OPTIMIZING DATA PLATFORMS

Azure Monitor
Azure Monitor provides a holistic monitoring approach by collecting, analysing, and
acting on telemetry from both cloud and on-premises environments

Metric Data

 Provides quantifiable information about a system over time that enables you to
observe the behaviour of a system.

Log Data

 Logs can be queried and even analysed using Log Analytics. In addition, this
information is typically presented in the overview page of an Azure Resource in
the Azure portal.

pg. 26
 Demysttify.Tech

Alerts

 Alerts notify you of critical conditions and potentially take corrective automated
actions based on triggers from metrics or logs.

Monitoring the network

Log Analytics within Azure monitor has the capability to monitor and measure network
activity.

Network Performance Monitor

 Measures the performance and reachability of the networks that you have
configured.

Application Gateway Analytics

 Contains rich, out-of-the box views you can get insights into key scenarios,
including:
o Monitor client and server errors.
o Check requests per hour

Connectivity Issues

pg. 27
 Demysttify.Tech

Performance Issues (To speed up query performance)

 Data Lake Storage

o Ensure hierarchical Namespace is enabled

 SQL Database

o Install the latest Document-DB SDK

o Use direct mode as your connection mode when configuring your
connection policy.
o Increase the no of thread or tasks to decrease the wait time while fulfilling
the requests.
o Identify and add missing indexes.

 Cosmos DB

o Avoid full scans on the collection, so query part of the collections

o All UDF's and built-in function will scan across all the documents within
the query
o Use direct mode as your connection mode when configuring your
connection policy.
o Tune the page size for querying and read feeds for better performance
using the x-ms-max-itime.count.header
o For any partisient collections query in parallel to increase performance and
leverage more throughput
o Use direct https connectivity mode for best performance

 Colocation of Resources

o Try increasing the RU between your collection

 SQL Data Warehouse

o Ensure the statistics are up-to-date

o Query optimizer

Storage Issues ‼️ 🚩

 Consistency

pg. 28
 Demysttify.Tech

 Corruption

Troubleshoot Streaming data

When using Stream Analytics, a Job encapsulates the stream Analytic work and is made
up of three components:

Troubleshoot batch data loads

When trying to resolve data load issues, it is first pragmatic to make the holistic checks
on Azure, as well as the network checks and diagnose and solve the issue check. After
that, then check:

Data redundancy

pg. 29
 Demysttify.Tech

Data redundancy is the process of storing data in multiple locations to ensure that it is
highly available.

Disaster Recovery
There should be process that are involved in backing up or providing failover for
databases in an Azure data platform technology. Depending on circumstances, there are
numerous approaches that can be adopted.

Scenarios
1. Recommended service: Azure Cosmos-DB

 Semi-structured: because of the need to extend or modify the schema for new
product
 Azure Cosmos DB indexes every field by default

pg. 30
 Demysttify.Tech

 ACID-compliant and faster while querying compared to other services

Advantages:

 Latency & throughput: High throughput and low latency

 Transactional support: Required
 Customers require a high number of read operations, with the ability to query on
many fields within the database.
 The business requires a high number of write operations to track the constantly
changing inventory.

2. Recommended service: Azure Blob storage

 Unstructured: Product catalog data

 Only need to be retrieved by ID.

 Customers require a high number of read operations with low latency.

 Creates and updates will be somewhat infrequent and can have higher latency
than read operations.

 Latency & throughput: Retrievals by ID need to support low latency and high
throughput. Creates and updates can have higher latency than read operations.

 Transactional support: Not required

3. Recommended service: Azure SQL Database

 Structured: Business data

 Operations: Read-only, complex analytical queries across multiple databases
 Latency & throughput: Some latency in the results is expected based on the
complex nature of the queries.
 Transactional support: Required

Tips to remember, A day prior to the

Exam.

pg. 31
 Demysttify.Tech

Azure Service Purpose

Synchronization of data between Azure Sql & On-

Azure SQL Data Sync
premises SQL with bi-directional

Azure SQL DB Elastic

Depend on eDTUs or Vcore's and max data size
pool's

Azure Data Lake Storage Azure Storage with Hierarchical nature

Data Migration between On-premise & Cloud with

almost 100% compatibility e.g.: From on-premises or
Azure SQL Database
IaaS, self-built, or ISV provided environment to fully
Managed Instance
managed PaaS cloud environment, with as low
migration effort as possible.

Used when same operation needs to be performed

Azure Resource Manager
frequently or daily basis with minimal effort eg:
Templates
Clusters

Synchronize data from on-premises Microsoft SQL

Server database to Azure SQL Database and to
Data Migration Assistant
determine whether data will move without
compatibility issues

Azure Data Warehouse Used frequently for Analytical data store

Azure Data Factory Orchestrate and manage the data lifecycle

In memory processing (or) support for usage of Scala,

Azure Data bricks (Spark) Java, Python, R languages (or) Cluster scale up or
scale down

Data load between any of

99% of the cases we use CTAS(Create Table As Select)
the two services SQl <=>
and not other operations such as Insert into, so on..
Blob <=> Data-warehouse

pg. 32
 Demysttify.Tech

Azure Service Purpose

A fully managed service designed to enable seamless

Azure Database Migration migrations from multiple database sources to Azure
Service (DMS) data platforms with minimal downtime (online
migrations).

Helps you evaluate a targeted version of SQL Server

for a specific workload. Customers upgrading from
Database Experimentation
earlier versions of SQL Server (starting with 2005) to
Assistant (DEA)
more recent versions of SQL Server can use the
analysis metrics that the tool provides.

A tool designed to automate database migration to

SQL Server Migration
SQL Server from Microsoft Access, DB2, MySQL,
Assistant (SSMA)
Oracle, and SAP ASE.

Azure Data Warehouse | Synapse Analytics

Azure Data Data

Reason Fit For
Warehouse distribution

Data size star schema with less

Small Dimension
Replicated usually less than 2 GB of storage
Table
than 2 GB after compression

Data size No obvious joining key

Temporary/Staging
Round Robin usually less or good candidate
Table
than 5 GB column

Data Size is
Hash
Fact Table huge more Large dimension tables
Distributed
than 100 GB

pg. 33
 Demysttify.Tech

Azure Data Warehouse | Synapse Analytics Selection of Table

Index

Type Fit For

Staging or temporary table, Small tables with small

Heap
lookups

Tables with up to 100 million rows, Large tables (more

Clustered index than 100 million rows) with only 1-2 columns heavily
used

Clustered column store

Large tables (more than 100 million rows)
index (CCI) (default)

Note:

 Preferred Index type is usually Clustered Colum-Store.

o e.g.: Similar to Parquet file.

Data bricks - Cluster Configurations

STANDARD HIGH CONCURRENCY

Recommended for... Single User Multiple Users

Language Support SQL, Python, R, and Scala SQL, Python, and R (not Scala)

Notebook Isolation No Yes

Azure Data Factory Triggers

pg. 34
 Demysttify.Tech

TYPE DESCRIPTION

Schedule Runs on a wall-clock schedule (e.g. every X mins/h/d/w/m's).

Tumbling A series of fixed-sized, non-overlapping, and contiguous time

Window intervals.

Runs pipelines in response to an event (e.g. Blob

Event-based
Created/Deleted).

Azure Data Factory Integration Runtime (IR) Usage

Cosmos DB Consistency level

Cosmos DB Entities by API

pg. 35
 Demysttify.Tech

ENTITY SQL CASSANDRA MONGODB GREMLIN TABLE

Container Container Table Collection Graph Table

Node or
Item Document Row Document Item
Edge

Database Database Key space Database Database N/A

Azure SQL DB Dynamic Data Masking

Data Masking
Masking value Example
Attribute

Zero value for numeric data types

Default value (0, xxxx,
Default (or) Mask 4 or less string data type
01-01-1900)
characters

Mask everything except characters at Custom string (prefix

Custom
beginning and at the end [padding] suffix)

Random Random number range

Returns a random number
Number e.g.: 0 to 100

Email Mask first letter and domain [email protected]

Exposes the last four digits of the

Credit card XXXX-XXXX-XXXX-1234
designated fields

Performance Parameters

pg. 36
 Demysttify.Tech

Azure Service Parameters

Azure Stream
Depends on Streaming Unit
Analytics

Azure SQL DB Depends on DTU's

Azure Data Depends on Cache used. Unit measured is Data warehouse

warehouse Units (DWU)

Azure Cosmos DB Depends on Data Integration Unit (or) Request Units (RU)

Pillars of Azure Architecture

pg. 37
 Demysttify.Tech

Design for Performance and Scalability

 Scaling
o Compute resources can be scaled in two different directions:
 Scaling up is the action of adding more resources to a single
instance.
 Scaling out is the addition of instances.
 Performance When optimizing for performance, you'lll look at network and
storage to ensure performance is acceptable. Both can impact the response time
of your application and databases.
 Patterns and Practices
o Partitioning
 In many large-scale solutions, data is divided into seperate
partitions that can be managed and accessed seperatly.
o Scaling
 Is the process of allocating scale units to match performance
requiremnets. This can be done either automatically or manually
o Caching
 Is a mechanism to store frequently used data or assests (web pages,
images) for faster retrieval.

Design for Availability and Recoverability

 Availabilty
o Focus on maintaining uptime through small-scale incidents and temporary
conditions like partial network outages.
 Recoverability
o Focus on recovery from data loss and from large scale disasters.
o Recovery Ponit Objective
 The maximum duration of acceptable data loss.
o Recovery Time Objective
 The maximum duration of acceptable downtime.

Design Azure Data Storage Solutions

Azure Storage

pg. 38
 Demysttify.Tech

BLOB

 It is also a backbone for creating a storage account that can be used as a Data
Lake storage

CosmosDB

 Globally distributed and elastically scalable database.

i) Core (SQL) API

 Default API for Azure Cosmos DB

 Can query hierarchical JSON documents with a SQL-like language
 Uses Javascript's type system, expression evaluation, and function invocation.

ii) MongoDB API

 Allows existing MongoDB client SDKs, drivers, and tools to interact with the data
transparently, as if they are running against an actual MongoDB database.
 Data is stored in document format, similar to Core (SQL)

iii) Cassandra API

pg. 39
 Demysttify.Tech

 Using Cassandra Query language (CQL), the data will appear to be a partitioned
row store.

iv) Table API

 The original table API only allows for indexing on the partition and row keys;
there are no secondary indexes.
 Storing table data in Cosmos DB automatically indexes all the properties, requires
no index management.
 Querying is accomplished by using OData and LINQ queries in code, and the
original REST API for GET operations.

v) Gremlin API

 Provides a graph based view over the data. Remember that at the lowest level, all
data in any Azure Cosmos DB is stored in an ARS format.
 Use a traversal language to query a graph database, and Azure Cosmos DB
supports Apache Tinkepop's Gremlin language.

Analyze the storage decision criteria

Scenario's to choose different CosmosDB API's

pg. 40
 Demysttify.Tech

pg. 41
 Demysttify.Tech

- We are not looking into any relationships so Gremlin is not the right choice
- Other CosmosDB API's are not used since the existing queries are MongoDB native and
there MongoDB is the best fit

pg. 42
 Demysttify.Tech

pg. 43
 Demysttify.Tech

Request Unit Considerations for CosmosDB

 Item size
 Item indexing
 Item property count
 Indexed properties
 Data consistency
 Query patterns
 Script usage

CosmosDB Partition Design

 Items are placed into logical partitions by partition key

 Partition keys should generally be based on unique values
 Ideally the partition key should be part of a query to prevent "fan out"
 Logical partitions are mapped to physical partitions
 A physical partition always contains at least one logical partition
 Physical partitions are capped at 10GB
 As physical partitions fill-up they will seamlessly split
 Logical partitions cannot be split

pg. 44
 Demysttify.Tech

Cosmos DB Change Feed feature

 Enables you to build efficient and scalable solutions for each of the patterns
shown below

pg. 45
 Demysttify.Tech

Azure SQL Database hosting options

pg. 46
 Demysttify.Tech

SQL Security

pg. 47
 Demysttify.Tech

Scenario to choose Azure SQL

Scenario to choose Azure Synapse

pg. 48
 Demysttify.Tech

Scenario to choose Azure Data Lake Storage Gen2

Azure Stream Analytics

Scenario to choose Stream Analytics

pg. 49
 Demysttify.Tech

3.Design Data Processing Solutions

Components of Big Data architecture

Lambda Architecture

pg. 50
 Demysttify.Tech

 When working with very large data sets, it can take a long time to run the sort of
queries that clients need.
 Often require algorithms such as Spark/ Map reduce that operate in parallel
across the entire data set.
 The results are then stored separately from the raw data and used for querying.
 Drawback to this approach is that it introduces latency

The lambda architecture, addresses this problem by creating two paths for data flow:

 Batch layer (cold path)

 Speed layer (hot path)

Kappa Architecture

pg. 51
 Demysttify.Tech

 A drawback to the lambda architecture is its complexity.

 Processing logic appears in two different places - the cold and hot paths - using
different frameworks
 This leads to duplicate computation logic and the complexity of managing the
architecture for both paths.
 The kappa architecture was proposed by Jay Kreps as an alternative to the
lambda architecture.
 All data flows through a single path, using a stream processing system.

IOT

Azure IOT Reference Architecture

pg. 52
 Demysttify.Tech

IOT Edge devices: Devices cannot be constantly connected to the cloud in this case IOT
edge devices contian some processing analysis logic within it. So that there is no
constant dependency for the cloud.

 eg: Shipment containers

IOT devices: Are constantly connected to the cloud which provides capability tp
perform data processing and analysis

Cloud Gateway (IOT Hub): Provides a cloud for a device to connect securely to the
cloud and send data. It acts a message broker between the devices and the other azure
services.

pg. 53
 Demysttify.Tech

Batch Processing

Scenario's to use Batch Processing

 From simple data transformations to a more complete ETL (extract-transform-

load) pipeline

pg. 54
 Demysttify.Tech

 In a big data context, batch processing may operate over very large data sets,
where the computation takes significant time.
 One example of batch processing is transforming a large set of flat, semi-
structured CSV or JSON files into a schematized and structured format that is
ready for further querying.

Design considerations for Batch processing

 Data format and encoding

o When files use an unexpected format or encoding

 Example is text fields that contain tabs, spaces, or commas that are
interpreted as delimiters
o Data loading and parsing logic must be flexible enough to detect and
handle these issues.

 Orchestrating time slices

o Often source data is placed in a folder hierarchy that reflects processing

windows, organized by year, month, day, hour, and so on.
o Can the downstream processing logic handle out-of-order records?

Batch processing Logical components

Batch processing Technology choices

pg. 55
 Demysttify.Tech

Batch processing with Azure Databricks

 Fast cluster start times, auto termination, autoscaling

 Built-in integration with Azure Blob Storage, ADLS, Azure Synapse, and other
services.
 User authentication with Azure Active Directory.
 Web-based notebooks for collaboration and data exploration.
 Supports GPU-enabled clusters.

Usage of Azure Databricks

pg. 56
 Demysttify.Tech

 To read data from multiple data sources such as Azure Blob Storage, ADLS, Azure
Cosmos DB, or SQL DW and turn it into breakthrough insights using spark.

Azure Machine Learning

Machine Learning Typical E2E process

pg. 57
 Demysttify.Tech

Devops Loop for Data Science

Azure Databricks + Azure ML

 Log experiments and models in a central place

pg. 58
 Demysttify.Tech

 Maintain audit trails centrally

 Deploy models seamlessly in Azure ML
 Manage your models in Azure ML

Standardizing the ML lifecycle on Azure Databricks

Realtime Processing

pg. 59
 Demysttify.Tech

Challenges

 One of the big challenges of real-time processing solutions is to ingest, process,

and store messages in real time, especially at high volumes.
 Processing must be done in such a way that it does not block the ingestion
pipeline.
 The data store must support high-volume writes.
 Another challenge is to act on data quickly such as generating alerts in real time
or presenting the data in a real-time (or near real-time) dashboard.

Real Time Processing Architecture - Logical COmponents

pg. 60
 Demysttify.Tech

Real Time Processing Technology choices

Azure Data Factory (ADF)

A cloud-based data integration service that allows you to orchestrate and automate
data movement and data transformation.

 Connect & collect

 Transform and enrich

ADF Components

pg. 61
 Demysttify.Tech

Data Transformation in Azure

Scenario to use ADF

pg. 62
 Demysttify.Tech

Real Time Analytics

Complexities in Stream Processing

 Complex Data

o Diverse data formats (json, avro, binary, ...)

o Data can be dirty, latem out of order

pg. 63
 Demysttify.Tech

 Complex Workloads

o Combining Streaming with interactive queries

o Machine learning

Design for Data Security and Compliance

Design for security

Identity Management
Identifying users that access your resources in an important part of security design.

 Identity as a security Layer

 Single sign-on
o With SSO, users only need to remember one ID and one password. Access
across database systems or applications is granted to a single identity tied
to a user. -SSO with Azure Active Directory

pg. 64
 Demysttify.Tech

o Azure AD is a cloud based identity service. It has built-in support for

synchronizing with your existing on-premises AD or can be used stand-
alone, This means that all your applications, whether on premise, in the
cloud (including Office 365) or even mobile can share the credentials.

Infrastructure Protection

Role Based Access Control

Roles are defined as collections of access permissions. Security principals are mapped to
roles directly or through group membership.

Role and Management groups:

 Roles are sets of permissions that users can be granted to. Management groups
add the ability to group subscriptions together and apply policy at an even
higher level.

Privileged Identity Management:

 Azure AD Privileged Identity Management (PIM) is an additional paid-for offering

that provides oversight of role assignments, self-service, and just-in-time role
activation.

Providing identities to services

An Azure service can be assigned an identity to ease the management of service access
to other Azure resources.

Service Principals:

 It is an identity that is used by a service or application. Like other identities it can

be assigned roles.

Managed identities:

 When you create a managed identity for a service, you create an account on the
Azure AD tenant. Azure infrastructure will automatically take care of
authentication.

Securing Azure Storage

pg. 65
 Demysttify.Tech

 Azure services such as Blob storage, Files share, Table storage, and Data Lake
Store all build on Azure Storage.
 High-level security benefits for the data in the cloud:
o Protect the data at rest
 That is encrypt the data before persisting it into the storage and
decrypt while retrieving. eg: Blob, Queue
o Protect the data in transit
o Support browser cross-domain access
o Control who can access data
o Audit storage class

Encryption at rest - Azure Storage Service Encryption (SSE)

 All storage data encrypted at rest - protected from physical breach

o By default, one master key per account, managed by Microsoft

o Optionally, protect the master key with your own key in Azure Key Vault
o Each write encrypted with a unique derived key

 All data writtern to storage is encrypted with SSE i.e, 256 bit advanced standard
AES cipher. SSE automatically encrypts data on writting to Azure storage. This
feature cannot be disabled

 For VM's Azure lets you encrypt virtual hard-disks by using Azure disk encryption.
This encryption uses bit locker for windows images and uses DEM encrypt for
Linux.

 Azure key Vault stores the keys automatically to help you control and manage
disk encryption, keys and secret automatically.

pg. 66
 Demysttify.Tech

Encryption at rest models

pg. 67
 Demysttify.Tech

Azure Key-Vault

 Safeguard cryptographic keys and other secrets used by cloud apps and services.

Encryption in transit

 Keep your data secure by enabling transport-level security between Azure and
the client.
 Always use HTTPS to secure communication over the public internet.

pg. 68
 Demysttify.Tech

 When you call the REST APIs to access objects in storage accounts, you can
enforce the use of HTTPS by requiring secure transfer for the storage account.

Cross Origin Resource Sharing (CORS) support

 Azure Storage supports cross-domain access through cross-origin resource

sharing (CORS)
 It is an optional flag that can be applied on storage accounts. The flag adds
appropriate headers when you use http requests to retrieve resources from
storage account.
 It uses HTTP headers so that a web application at one domain can access
resources from a server at a different domain.
 By using CORS, web apps ensure that they load only authorizes content from
authorized sources.

Identity-Based Access Control for Azure Blob Storage

Grant access to user and service identities from Azure Active Directory

 Federate with enterprise identity systems

 Leverage powerful AAD capabilities including 2-factor and biometric
authentication, conditional access, identity protection and more.

Control access with role-based access control (RBAC)

 Grant access to storage scopes ranging from entire enterprise down to one blob
container
 Define custom roles that match your security model
 Leverage Privileged Identity Management to reduce standing administrative
access.

AAD Authentication and RBAC currently support AAD, OAuth and RBAC on Storage
Resource Provider via ARM.

Managed identities for Azure resources

pg. 69
 Demysttify.Tech

 Auto-managed identity in Azure AD for Azure resource.

 Use the MSI endpoint to get access tokens from Azure AD (no secretes required).
 Direct authentication with services, or retrieve creds from Azure key vault
 No additional charge for MSI.

Managed Shared Access Policies and Signatures

Storage Explorer provides the ability to manage access policies for containers.

 A shared access signature (SAS) provides you with a way to grant limited access
to other clients, without exposing your account key.
 Provides delegated access to resources in your storage account.

Types of Shared Access Signature (SAS)

 Service level

o Service level SAS are defined on a resource under a particular service.

o Used to allow access to specific resources in a storage account.

pg. 70
 Demysttify.Tech

o For example, to allow an app tp retrieve a list of files in a file system or to

download a file.

 Account level

o Targets the storage account and can apply to multiple services and
resources
o For example, you can use an account-level SAS to allow the ability to
create file systems.

Immutability Policies

 Support for time-based retention

o container level configuration

o RBAC support and policy auditing
o BLobs cannot be modified or deleted for N days

 Support for legal holds with tags

o Container level configuration

o Blobs cannot be modified or deleted when legal hold is set.

 Support for all Blob tiers

o Applies to hot, cool and cold data

o Policies retained when data is tiered

 SEC 17a-4(f) complaint

Firewall rules

pg. 71
 Demysttify.Tech

 Azure SQL DB has a built-in firewall that is used to allow and deny network access
to both the db server itself, as well as individual db.
 Server-level firewall rules
o Allow access to Azure services
o IP address rules
o Virtual network rules
 Database-level firewall rules
o IP address rules

Network Security
Network security is protecting the communication of resources within and outside of
your network. The goal is to limit exposure at the network layer across your services and
systems

pg. 72
 Demysttify.Tech

Internet protection:

 By assessing the resources that are internet-facing, and only allow inbound and
outbound communication when necessary. Ensure that they are restricted to only
ports/protocols required.

VIrtual network security:

 To isolate Azure services to only allow communication from virtual networks, use
VNet service endpoints. With service endpoints, Azure service resources can be
secured to your virtual network.

Network Integration:

 VPN connections are a common way of establishing secure communication

channels between networks, and this is no different when working with virtual
networking on Azure. Connection between Azure VNets and an on-premises VPN
device is a great way to provide secure communication.

Firewalls and VNET access

pg. 73
 Demysttify.Tech

 Storage Firewall
o Block internet access to data
o Grant access to clients in specific vnet
o Grant access to clients from on-premise networks via public peering
network gateway

Private endpoints

 Azure private endpoint is a fundamental building block for private link in Azure. It
enables service like Azure VM to communicate privately with private link
resources.
 It is a network interwork interface that connects you privately and securely to
service powered by Azure Private link.
 A private endpoint assigns a private IP address from your Azure Virtual Network
(VNET) to the storage account.

pg. 74
 Demysttify.Tech

 private endpoint enables communication from the same VNet, regionally peered
VNets, globally peered VNets, and on-premises using VPN or Express Route, and
services powered by private link.
 It secures all traffic between your VNet and the storage account over a private
link.

Advanced threat protection for Azure Storage

 An additional layer of security intelligence that detects unusual and potentially

harmful attempts to access or exploit storage accounts
 These security alerts are integrated with Azure Security Center.

Secure Azure Cosmos DB Data

 Using Firewall settings

 Add inbound and Outbound networks

Azure SQL database - Secure your data in transit, at rest and on display

 TLS network encryption

o Azure SQL DB enforces Transport Layer Security (TLS) encryption at all
times for all connections, which ensures all data is encrypted "in transit"
between the database and the client.
 Transparent Data Encryption (TDE)
o Protects your data at rest using TDE.
o TDE performs real-time encryption and decryption of the DB, associated
backups, and transaction log files at rest without requiring changes to the
application.
 Dynamic data masking
o By using the, we can limit the data that is displayed to the user.
o Policy-based security feature that hides the sensitive data in the result set
of a query over designated DB fields, while the data in the DB is not
changed e.g: phone numbers, credit card numbers.

Enable Database Auditing

 For SQL Server you can create audits that contain specifications for server-level
events and specifications for database-level events.
 Audited events can be written to the event logs or to audit files

pg. 75
 Demysttify.Tech

 There are several levels of auditing for SQL Server, depending on government or
standards requirements for your installation.
 Azure SQL DB and Azure Synapse Analytics auditing tracks database events and
writes them to an audit log in your Azure storage account.
 Enable Threat detection to know any malicious activities on SQL DB or potential
security threats.

Use an Azure SQL Database managed instance securely with public endpoints

 A SQL DB managed instance provides a private endpoint to allow connectivity

from inside its VNET.

Scenarios where you need to provide public endpoint connection

 The managed instance must integrate with multi-tenant-only PaaS offerings. -You
need higher throughput of data exchange than is possible when you're using
VPN.
 Company policies prohibit PaaS inside corporate networks.

Managed Instance - Lock down inbound and outbound connectivity

pg. 76
 Demysttify.Tech

 A managed instance has a dedicated public endpoint address.

 In the client-side outbound firewall and in the NSG rules, set this public endpoint
IP address to limit outbound connectivity.
 Use a NSG to limit access to the managed instance public endpoint on port 3342.

Azure SQL Database and Azure Synapse Analytics data discovery & classification

 Discovery & recommendations

o The classification engine scans your DB and identifies columns containing
potentially sensitive data. It then provides you an easy way to review and
apply the appropriate classification recommendations via the Azure portal.
 Labelling
o Sensitivity classification labels can be persistently tagged on columns
using new classification metadata attributes introduced into the SQL
Engine. This metadata can then be utilized for advanced sensitivity-based
auditing and protection scenarios.
 Query result set sensitivity
o The sensitivity of query result set is calculated in real time for auditing
purposes.
 Visibility
o The DB classification state can be viewed in a detailed dashboard in the
portal.

Discover, classify & label sensitive columns

The classification includes two metadata attributes:

 Labels
o The main classification attributes used to define the sensitivity level of the
data stored in the column.
 Information Types
o Provide additional granularity into the type of data stored in the column.

Architecture of Azure Data Factory

pg. 77
 Demysttify.Tech

Security aspects that are part of the above architecture

 AAD access control

o SQLDB, ADLS Gen2 and Azure function only allow the Managed Identity
(MI) of ADFv2 to access the data. This means that no keys need to be
stored in ADFv2 or Key vaults.
o To secure ADLS Gen2 account:
 Add RBAC rule that only MI of ADFv2 can access ADLS Gen2
 Add firewall rule that only VNET of Self Hosted Integrated Runtime
(SHIR) can access ADLS Gen2 container.
 Firewall rules
o SQLDB, ADLS Gen2 and Azure function all have firewall rules in which only
the VNET of the SHIR is allowed as inbound network.
o To secure SQLDB:
 Add Database rule that only MI of ADFv2 can access SQLDB
 Add firewall rule that only VNET of SHIR can access SQLDB

Miscellaneous

Serverless Computing
Containers

pg. 78
 Demysttify.Tech

 A container is a method running applications in a virtualized environment. The

virtualization is done at the OS level, making it possible to run multiple identical
application instances within the same OS.

Azure Kubernetes Service (AKS)

 Azure Kubernetes Service allows you to set up virtual machines to act as your
nodes. Azure hosts the Kubernetes management plane and only bills for the
running worker nodes that host your containers.

Azure Container Instance (ACI)

 It is a serverless approach that lets you create and execute containers on

demand. You're charged only for the execution time per second.

Performance Bottlenecks
Azure Monitor

 A single management point for infrastructure-level logs and monitoring for most
of your Azure services.

Log Analytics

 You can query and aggregate data across logs. This cross-source correlation can
help you identity issues or performance problems that may not be evident when
looking at logs or metrics individually.

Application performance management

 Telemetry can include individual page request times, exceptions within your
application, and even custom metrics to track business logic. This telemetry can
provide a wealth of insight into apps.

Tips to remember, A day prior to the

Exam.

pg. 79
 Demysttify.Tech

Azure Service Type of File

Azure CosmosDB Graph Databases

Azure Hbase and HDInsight Column Family in-memory key-value store

Azure Service Usage

Azure Synapse Data analytics

Azure Search Search engine databases

Azure Timeseries Insights Time series databases

Azure Blob Object store

Azure FileStorage Shared files

Azure Cosmos DB Usage:

 For Real-Time Customer Experiences

 Telemetry stores for IOT
 Migrate NoSQL apps

Azure Cosmos DB Authentication

 It uses two types of keys to authenticate users and provide access to its data and
resources.

Key Type Resources

Used for administrative resources: database accounts, databases,

Master Keys
users, and permissions

pg. 80
 Demysttify.Tech

Key Type Resources

Resource Used for application resources: containers, documents,

tokens attachments, stored procedures, triggers, and UDFs

Azure Cosmos DB SLA for Read/Write operation

Azure Storage Service

pg. 81
 Demysttify.Tech

Azure Storage Availability

Azure Data Factory

pg. 82
 Demysttify.Tech

 Does not store any data except for linked service credentials for cloud data
stores, which are encrypted by using certificates.

Azure Databricks (2 types of clusters)

 Interactive clusters are used to analyze data collaboratively with interactive

notebooks.
 Job clusters are used to run fast and robust automated workloads using the UI or
API.

Azure SQL Database - Security Overview

LAYER TYPE DESCRIPTION

Grant access to databases based on the

Network IP Firewall Rules
originating IP address of each request.

Only accept communications that are

Virtual Network
Network sent from selected subnets inside a
Firewall Rules
virtual network.

Access Authentication of a user when using a

SQL Authentication
Management username and password.

Access Azure AD Leverage centrally managed identities

Management Authentication in Azure Active Directory (Azure AD).

Control access to rows in a table based

Authorization Row-level Security
on the characteristics of the user/query.

Tracks database activities by recording

Threat
Auditing events to an audit log in an Azure
Protection
storage account.

pg. 83
 Demysttify.Tech

LAYER TYPE DESCRIPTION

Analyzing SQL Server logs to detect

Threat Advanced Threat
unusual and potentially harmful
Protection Protection
behavior.

Information Transport Layer Encryption-in-transit between client

Protection Security (TLS) and server.

Information Transparent Data Encryption-at-rest using AES (Azure

Protection Encryption (TDE) SQL DB encrypted by default).

Encryption-in-use (Column-level
Information
Always Encrypted granularity; Decrypted only for
Protection
processing by client).

Information Dynamic Data Limits sensitive data exposure by

Protection Masking masking it to non-privileged users.

Security Vulnerability Discover, track, and help remediate

Management Assessment potential database vulnerabilities.

Discovering, classifying, labeling, and

Security Data Discovery &
protecting the sensitive data in your
Management Classification
databases.

Security Been certified against a number of

Compliance
Management compliance standards.

Azure SQL - Network Access Controls

pg. 84
 Demysttify.Tech

CONTROL DESCRIPTION

When set to ON, other resources within the Azure

Allow Azure Services
boundary can access the SQL resource.

Use this feature to explicitly allow connections from a

IP firewall rules
specific IP address.

Virtual Network Use this feature to allow traffic from a specific Virtual
firewall rules Network within the Azure boundary.

Encryption on Azure

Technique or service
Type Enables encryption of
used

Azure Storage, VM Disks, Disk

Raw Encryption -
Encryption

Database Transparent Data

Databases and SQL DW
Encryption Encryption

Encypting Secrets Azure Key Vault Storing application secrets

HDInsight cluster types to run Apache Hive queries

Cluster Type Usage

Interactive Query To optimize for ad hoc, interactive queries

Apache Hadoop To optimize for hive queries used as batch process

Spark & HBase Run hive queries

pg. 85
 Demysttify.Tech

SQLDW or Synapse
To achieve fastest loading speed for moving data into a DW table

 load data into a staging table. Define the staging table as a heap and use round-
robin for the distribution option.

Criteria to select a Distribution column

 Has many Unique values

 Does not have Nulls, or has only a few Nulls
 Is not a date column
 Use a Column from Group BY, not from where clause

Distribution Type

 Round Robin for small Fact tables

 Hash distributed for large Fact tables

Data corruption checks

 We create a user-defined restore point before data is uploaded. Delete the

restore point after data corruption checks complete.

Simple yet very powerful Hacks:

(You can use these hacks when uncertain about any question or scenario to make
quick decisions)

 IOT HUb, Event Hub, Blob are the three ways to bring data into Stream Analytics
 Anything related to RBAC Identities majority cases answer would be Service
Srincipal
 In MySQL sharding is the best way to partition the data.
o Criteria to select a column for sharding
 Unique (data should be well distributed)
 Cosmos DB Partition keys should generally be based on unique values.
 For a Database using a nonclustered columnstore index will improve performance
on analytics and not clustered columnstore index.

pg. 86
 Demysttify.Tech

 You can use Azure Event Hubs, IoT hub, and Azure Blob storage for streaming
data input.
 Azure Stream Analytics Supports Azure SQL DB or Azure Blob storage
for reference data input.
 Primary key and secondary key grant access to remotely administer the Storage
account.
 Event Hubs Capture creates files in Avro format.
 If notebooks are involved with scheduling or autoscale of clusters it is databricks.
 RBAC support for databricks via Premium clusters.
 If there is a question based on IOT Hub or Event Hub probability of the answer
being Stream Analytics for processing is maximum.
 If you see the term "Relationship" or nodes and vertices in CosmosDB question
by default the option in Gremlin API.
 If hierarchical or Big Data related storage involved then ADLS Gen2.
 If flat file related storage then blob.

pg. 87