Sqoop, Hive and Impala for Data Analysts (Formerly CCA 159)

Sqoop, Hive and Impala for Data Analysts (Formerly CCA 159) Hands on Sqoop, Hive and Impala for Data Analysts.

As part of Sqoop, Hive, and Impala for Data Analysts (Formerly CCA 159), you will learn key skills such as Sqoop, Hive, and Impala.

This comprehensive course covers all aspects of the certification with real-world examples and data sets.

Overview of Big Data ecosystem

  • Overview Of Distributions and Management Tools
  • Properties and Properties Files – General Guidelines
  • Hadoop Distributed File System
  • YARN and Map Reduce2
  • Submitting Map ReduceJob
  • Determining Number of Mappers and Reducers
  • Understanding YARN and Map Reduce Configuration Properties
  • Review and Override Job Properties
  • Reviewing Map Reduce Job Logs
  • Map Reduce Job Counters
  • Overview of Hive
  • Databases and Query Engines
  • Overview of Data Ingestion in Big Data
  • Data Processing using Spark

HDFS Commands to manage files

  • Introduction to HDFS for Certification Exams
  • Overview of HDFS and PropertiesFiles
  • Overview of Hadoop CLI
  • Listing Files in HDFS
  • User Spaces or Home Directories in HDFS
  • Creating Directories in HDFS
  • Copying Files and Directories into HDFS
  • File and Directory Permissions Overview
  • Getting Files and Directories from HDFS
  • Previewing Text Files in HDFS
  • Copying or Moving Files and Directories within HDFS
  • Understanding Size of File System and Files
  • Overview of Block Size and ReplicationFactor
  • Getting File Metadata using hdfs fsck
  • Resources and Exercises

Getting Started with Hive

  • Overview of Hive Language Manual
  • Launching and using Hive CLI
  • Overview of Hive Properties
  • Hive CLI History and hiverc
  • Running HDFS Commands in Hive CLI
  • Understanding Warehouse Directory
  • Creating and Using Hive Databases
  • Creating and Describing Hive Tables
  • Retrieve Matadata of Tables using DESCRIBE
  • Role of Hive Metastore Database
  • Overview of beeline
  • Running Hive Commands and Queries using beeline

Creating Tables in Hive using Hive QL

  • Creating Tables in Hive – orders
  • Overview of Basic Data Types in Hive
  • Adding Comments to Columns and Tables
  • Loading Data into Hive Tables from Local File System
  • Loading Data into Hive Tables from HDFS
  • Loading Data – Overwrite vs Append
  • Creating External tables in Hive
  • Specifying Location for Hive Tables
  • Difference between Managed Table and External Table
  • Default Delimiters in Hive Tables using Text File
  • Overview of File Formats in Hive
  • Differences between Hive and RDBMS
  • Truncate and Drop tables in Hive
  • Resources and Exercises

Loading/Inserting data into Hive tables using Hive QL

  • Introduction to Partitioning and Bucketing
  • Creating Tables using Orc Format – order_items
  • Inserting Data into Tables using Stage Tables
  • Load vs. Insert in Hive
  • Creating Partitioned Tables in Hive
  • Adding Partitions to Tables in Hive
  • Loading into Partitions in Hive Tables
  • Inserting Data Into Partitions in Hive Tables
  • Insert Using Dynamic Partition Mode
  • Creating Bucketed Tables in Hive
  • Inserting Data into Bucketed Tables
  • Bucketing with Sorting
  • Overview of ACID Transactions
  • Create Tables for Transactions
  • Inserting Individual Records into Hive Tables
  • Update and Delete Data in Hive Tables

Overview of functions in Hive

  • Overview of Functions
  • Validating Functions
  • String Manipulation – Case Conversion and Length
  • String Manipulation – substr and split
  • String Manipulation – Trimming and Padding Functions
  • String Manipulation – Reverse and Concatenating Multiple Strings
  • Date Manipulation – Current Date and Timestamp
  • Date Manipulation – Date Arithmetic
  • Date Manipulation – trunc
  • Date Manipulation – Using date format
  • Date Manipulation – Extract Functions
  • Date Manipulation – Dealing with Unix Timestamp
  • Overview of Numeric Functions
  • Data Type Conversion Using Cast
  • Handling Null Values
  • Query Example – Get Word Count

Writing Basic Queries in Hive

  • Overview of SQL or Hive QL
  • Execution Life Cycle of Hive Query
  • Reviewing Logs of Hive Queries
  • Projecting Data using Select and Overview of From
  • Derive Conditional Values using CASE and WHEN
  • Projecting Distinct Values
  • Filtering Data using Where Clause
  • Boolean Operations in Where Clause
  • Boolean OR vs IN Operator
  • Filtering Data using LIKE Operator
  • Performing Basic Aggregations using Aggregate Functions
  • Performing Aggregations using GROUP BY
  • Filtering Aggregated Data Using HAVING
  • Global Sorting using ORDER BY
  • Overview of DISTRIBUTE BY
  • Sorting Data within Groups using SORT BY
  • Using CLUSTERED BY

Joining Data Sets and Set Operations in Hive

  • Overview of Nested Sub Queries
  • Nested Sub Queries – Using IN Operator
  • Nested Sub Queries – Using EXISTS Operator
  • Overview of Joins in Hive
  • Performing Inner Joins using Hive
  • Performing Outer Joins using Hive
  • Performing Full Outer Joins using Hive
  • Map Side Join and Reduce Side Join in Hive
  • Joining in Hive using Legacy Syntax
  • Cross Joins in Hive
  • Overview of Set Operations in Hive
  • Perform Set Union between two Hive Query Results
  • Set Operations – Intersect and Minus Not Supported

Windowing or Analytics Functions in Hive

  • Prepare HR Database in Hive with Employees Table
  • Overview of Analytics or Windowing Functions in Hive
  • Performing Aggregations using Hive Queries
  • Create Tables to Get Daily Revenue using CTAS in Hive
  • Getting Lead and Lag using Windowing Functions in Hive
  • Getting First and Last Values using Windowing Functions in Hive
  • Applying Rank using Windowing Functions in Hive
  • Applying Dense Rank using Windowing Functions in Hive
  • Applying Row Number using Windowing Functions in Hive
  • Difference Between rank, dense_rank, and row_number in Hive
  • Understanding the order of execution of Hive Queries
  • Overview of Nested Sub Queries in Hive
  • Filtering Data on Top of Window Functions in Hive
  • Getting Top 5 Products by Revenue for Each Day using Windowing Functions in Hive – Recap

Running Queries using Impala

  • Introduction to Impala
  • Role of Impala Daemons
  • Impala State Store and Catalog Server
  • Overview of Impala Shell
  • Relationship between Hive and Impala
  • Overview of Creating Databases and Tables using Impala
  • Loading and Inserting Data into Tables using Impala
  • Running Queries using Impala Shell
  • Reviewing Logs of Impala Queries
  • Synching Hive and Impala – Using Invalidate Metadata
  • Running Scripts using Impala Shell
  • Assignment – Using NYSE Data
  • Assignment – Solution

Getting Started with Sqoop

  • Introduction to Sqoop
  • Validate Source Database – MySQL
  • Review JDBC Jar to Connect to MySQL
  • Getting Help using Sqoop CLI
  • Overview of Sqoop User Guide
  • Validate Sqoop and MySQL Integration using Sqoop List Databases
  • Listing Tables in Database using Sqoop
  • Run Queries in MySQL using Sqoop Eval
  • Understanding Logs in Sqoop
  • Redirecting Sqoop Job Logs into Log Files

Importing data from MySQL to HDFS using Sqoop Import

  • Overview of Sqoop Import Command
  • Import Orders using target-dir
  • Import Order Items using warehouse-dir
  • Managing HDFS Directories
  • Sqoop Import Execution Flow
  • Reviewing Logs of Sqoop Import
  • Sqoop Import Specifying Number of Mappers
  • Review the Output Files generated by Sqoop Import
  • Sqoop Import Supported File Formats
  • Validating avro files using Avro Tools
  • Sqoop Import Using Compression

Apache Sqoop – Importing Data into HDFS – Customizing

  • Introduction to customizing Sqoop Import
  • Sqoop Import by Specifying Columns
  • Sqoop import Using Boundary Query
  • Sqoop import while filtering Unnecessary Data
  • Sqoop Import Using Split By to distribute import using non default column
  • Getting Query Results using Sqoop eval
  • Dealing with tables with Composite Keys while using Sqoop Import
  • Dealing with tables with Non Numeric Key Fields while using Sqoop Import
  • Dealing with tables with No Key Fields while using Sqoop Import
  • Using autoreset-to-one-mapper to use only one mapper while importing data using Sqoop from tables with no key fields
  • Default Delimiters used by Sqoop Import for Text File Format
  • Specifying Delimiters for Sqoop Import using Text File Format
  • Dealing with Null Values using Sqoop Import
  • Import Mulitple Tables from source database using Sqoop Import

Importing data from MySQL to Hive Tables using Sqoop Import

  • Quick Overview of Hive
  • Create Hive Database for Sqoop Import
  • Create Empty Hive Table for Sqoop Import
  • Import Data into Hive Table from source database table using Sqoop Import
  • Managing Hive Tables while importing data using Sqoop Import using Overwrite
  • Managing Hive Tables while importing data using Sqoop Import – Errors Out If Table Already Exists
  • Understanding Execution Flow of Sqoop Import into Hive tables
  • Review Files generated by Sqoop Import in Hive Tables
  • Sqoop Delimiters vs Hive Delimiters
  • Different File Formats supported by Sqoop Import while importing into Hive Tables
  • Sqoop Import all Tables into Hive from source database

Exporting Data from HDFS/Hive to MySQL using Sqoop Export

  • Introduction to Sqoop Export
  • Prepare Data for Sqoop Export
  • Create Table in MySQL for Sqoop Export
  • Perform Simple Sqoop Export from HDFS to MySQL table
  • Understanding Execution Flow of Sqoop Export
  • Specifying Number of Mappers for Sqoop Export
  • Troubleshooting the Issues related to Sqoop Export
  • Merging or Upserting Data using Sqoop Export – Overview
  • Quick Overview of MySQL – Upsert using Sqoop Export
  • Update Data using Update Key using Sqoop Export
  • Merging Data using allowInsert in Sqoop Export
  • Specifying Columns using Sqoop Export
  • Specifying Delimiters using Sqoop Export
  • Using Stage Table for Sqoop Export

Submitting Sqoop Jobs and Incremental Sqoop Imports

  • Introduction to Sqoop Jobs
  • Adding Password File for Sqoop Jobs
  • Creating Sqoop Job
  • Run Sqoop Job
  • Overview of Incremental Loads using Sqoop
  • Incremental Sqoop Import – Using Where
  • Incremental Sqoop Import – Using Append Mode
  • Incremental Sqoop Import – Create Table
  • Incremental Sqoop Import – Create Sqoop Job
  • Incremental Sqoop Import – Execute Job
  • Incremental Sqoop Import – Add Additional Data
  • Incremental Sqoop Import – Rerun Job
  • Incremental Sqoop Import – Using Last Modified

Here are the objectives for this course.

Provide Structure to the Data

Use Data Definition Language (DDL) statements to create or alter structures in the metastore for use by Hive and Impala.

  • Create tables using a variety of data types, delimiters, and file formats
  • Create new tables using existing tables to define the schema
  • Improve query performance by creating partitioned tables in the metastore
  • Alter tables to modify the existing schema
  • Create views in order to simplify queries

Data Analysis

Use Query Language (QL) statements in Hive and Impala to analyze data on the cluster.

  • Prepare reports using SELECT commands including unions and subqueries
  • Calculate aggregate statistics, such as sums and averages, during a query
  • Create queries against multiple data sources by using join commands
  • Transform the output format of queries by using built-in functions
  • Perform queries across a group of rows using windowing functions

Exercises will be provided to have enough practice to get better at Sqoop as well as writing queries using Hive and Impala.

All the demos are given on our state-of-the-art Big Data cluster. If you do not have multi-node cluster, you can sign up for our labs and practice on our multi-node cluster. You will be able to practice Sqoop and Hive on the cluster.


Free $24.99 Redeem Coupon
Online Tutorials
Show full profile

Online Tutorials

Online Tutorials is a website sharing online courses, and online tutorials for free on a daily basis. You can find the best free online courses and thousands of free online courses with certificates to take your knowledge to the next level with the free courses.

We will be happy to hear your thoughts

Leave a reply

Online College Courses
Logo
Register New Account