Introduction
This tutorial describes how to refine data for a Trucking IoT Data Discovery (aka IoT Discovery) use case using the Hortonworks Data Platform. The IoT Discovery use cases involves vehicles, devices and people moving across a map or similar surface. Your analysis is interested in tying together location information with your analytic data.
Hello World is often used by developers to familiarize themselves with new concepts by building a simple program. This tutorial aims to achieve a similar purpose by getting practitioners started with Hadoop and HDP. We will use an Internet of Things (IoT) use case to build your first HDP application.
For our tutorial we are looking at a use case where we have a truck fleet. Each truck has been equipped to log location and event data. These events are streamed back to a datacenter where we will be processing the data. The company wants to use this data to better understand risk.
Here is the video of Analyzing Geolocation Data to show you what you’ll be doing in this tutorial.
Prerequisites:
* Hortonworks Sandbox 2.3 (installed and running)
- Data Set Used: Geolocation.zip
- Optional: Hortonworks ODBC driver installed and configured – see the tutorial on installing the ODBC driver for Windows or OS X. Refer to
- Installing and Configuring the Hortonworks ODBC driver on Windows 7
- Installing and Configuring the Hortonworks ODBC driver on Mac OS X
- Microsoft Excel 2013 Professional Plus is required for the Windows 7 or later installation to be able to construct the maps.
NOTE**
Tutorial Overview
In this tutorial we will be providing the collected geolocation and truck data. We will import this data into HDFS and build derived tables in Hive. Then we will process the data using Pig, Hive and Spark. The processed data is then imported into Microsoft Excel where it can be visualized.
To refine and analyze Geolocation data, we will:
- Review some Hadoop Fundamentals
- Download and extract the Geolocation data files.
- Load the captured data into the Hortonworks Sandbox.
- Run Hive,Pig and Spark scripts that compute truck mileage and driver risk factor.
- Access the refined sensor data with Microsoft Excel.
- Visualize the sensor data using Excel Power View.
Goals of the Tutorial
** **The goal of this tutorial is that you get familiar with the basics of following:
- Hadoop and HDP
- Ambari File User Views and HDFS
- Ambari Hive User Views and Apache Hive
- Ambari Pig User Views and Apache Pig
- Apache Spark
- Data Visualization with Excel (Optional)
- Introduction
- Prerequisites 1. Data Set Used: Geolocation.zip 2. HDP version: 2.3
- Tutorial Overview
- Goals of the Tutorial (outcomes)
- Hadoop Data Platform Concepts (New to Hadoop or HDP- Refer following) 1. Apache Hadoop and HDP(5 Pillars) 2. Apache Hadoop Distributed File System(HDFS) 3. Apache MapReduce and Apache YARN 4. Apache Hive and Apache Pig
- Get Started with HDP Labs
1. Step 0: Start Sandbox VM and Open Ambari
2. Lab 1: Loading Sensor Data into HDFS
3. Lab 2: Data Manipulation with Hive (AMBARI HIVE USER VIEWS)
4. Lab 3: Use Pig to compute Driver Risk Factor
5. Lab 4: Use Apache Spark to compute Driver Risk Factor
6. Lab 5: Optional Visualization and Reporting
- Configuring ODBC driver (Mac and Windows)
- Using Microsoft Excel
- Using Zeppelin (Coming Soon…)
- Next Step/Try These
1. Practitioner Journey- As a Hadoop Practitioner you can adopt following learning paths
- Hadoop Developer- Click Here!
- Hadoop Administrator-Click Here!
- Data Scientist- Click Here! 2. Case Studies– Learn how Hadoop is being used by various industries.
Concepts: Hadoop & HDP
In this module you will learn about Apache Hadoop and what makes it scale to large data sets. We will also talk about various components of Hadoop ecosystem that make Apache Hadoop enterprise ready in form of Hortonworks Data Platform(HDP) distribution. The module discusses Apache Hadoop, its capabilities as a data platform and how the core of Hadoop and its surrounding ecosystem solution vendors provides the enterprise requirements to integrate alongside the Data Warehouse and other enterprise data systems as part of a modern data architecture, and as a step on the journey toward delivering an enterprise ‘Data Lake’
- Understanding Hadoop.
- Understanding five pillars of HDP
- Understanding HDP components and their purpose.
Apache Hadoop® is an open source framework for distributed storage and processing of large sets of data on commodity hardware. Hadoop enables businesses to quickly gain insight from massive amounts of structured and unstructured data. Numerous Apache Software Foundation projects make up the services required by an enterprise to deploy, integrate and work with Hadoop.
NOTE
******
The base Apache Hadoop framework is composed of the following modules:
- Hadoop CommonLibraries and utilities needed by other Hadoop modules.
- **Hadoop Distributed File System (HDFS)**A distributed file-system that stores data on commodity machines, providing very high aggregate bandwidth across the cluster.
- Hadoop YARNA resource-management platform responsible for managing computing resources in clusters and using them for scheduling of users’ applications.
- Hadoop MapReduceA programming model for large scale data processing.
Each project has been developed to deliver an explicit function and each has its own community of developers and individual release cycles. There are five pillars to Hadoop that make it enterprise ready:
Store and process vast quantities of data in a storage layer that scales linearly. Hadoop Distributed File System (HDFS) is the core technology for the efficient scale out storage layer, and is designed to run across low-cost commodity hardware. Apache Hadoop YARN is the pre-requisite for Enterprise Hadoop as it provides the resource management and pluggable architecture for enabling a wide variety of data access methods to operate on data stored in Hadoop with predictable performance and service levels.
- **Apache Hadoop YARN**Part of the core Hadoop project, YARN is a next-generation framework for Hadoop data processing extending MapReduce capabilities by supporting non-MapReduce workloads associated with other programming models.
- **HDFS**Hadoop Distributed File System (HDFS) is a Java-based file system that provides scalable and reliable data storage that is designed to span large clusters of commodity servers.
Interact with your data in a wide variety of ways – from batch to real-time. Apache Hive is the most widely adopted data access technology, though there are many specialized engines. For instance, Apache Pig provides scripting capabilities, Apache Storm offers real-time processing, Apache HBase offers columnar NoSQL storage and Apache Accumulo offers cell-level access control. All of these engines can work across one set of data and resources thanks to YARN and intermediate engines such as Apache Tez for interactive access and Apache Slider for long-running applications. YARN also provides flexibility for new and emerging data access methods, such as Apache Solr for search and programming frameworks such as Cascading.
- **Apache Hive**Built on the MapReduce framework, Hive is a data warehouse that enables easy data summarization and ad-hoc queries via an SQL-like interface for large datasets stored in HDFS.
- **Apache Pig**A platform for processing and analyzing large data sets. Pig consists of a high-level language (Pig Latin) for expressing data analysis programs paired with the MapReduce framework for processing these programs.
- **MapReduce**MapReduce is a framework for writing applications that process large amounts of structured and unstructured data in parallel across a cluster of thousands of machines, in a reliable and fault-tolerant manner.
- **Apache SparkSpark is ideal for in-memory data processing. It allows data scientists to implement fast, iterative algorithms for advanced analytics such as clustering and classification of datasets. **
- **Apache Storm**Storm is a distributed real-time computation system for processing fast, large streams of data adding reliable real-time data processing capabilities to Apache Hadoop® 2.x
- **Apache HBase**A column-oriented NoSQL data storage system that provides random real-time read/write access to big data for user applications.
- **Apache Tez**Tez generalizes the MapReduce paradigm to a more powerful framework for executing a complex DAG (directed acyclic graph) of tasks for near real-time big data processing.
- **Apache Kafka**Kafka is a fast and scalable publish-subscribe messaging system that is often used in place of traditional message brokers because of its higher throughput, replication, and fault tolerance.
- **Apache HCatalog**A table and metadata management service that provides a centralized way for data processing systems to understand the structure and location of the data stored within Apache Hadoop.
- **Apache Slider**A framework for deployment of long-running data access applications in Hadoop. Slider leverages YARN’s resource management capabilities to deploy those applications, to manage their lifecycles and scale them up or down.
- **Apache Solr**Solr is the open source platform for searches of data stored in Hadoop. Solr enables powerful full-text search and near real-time indexing on many of the world’s largest Internet sites.
- **Apache Mahout**Mahout provides scalable machine learning algorithms for Hadoop which aids with data science for clustering, classification and batch based collaborative filtering.
- **Apache Accumulo**Accumulo is a high performance data storage and retrieval system with cell-level access control. It is a scalable implementation of Google’s Big Table design that works on top of Apache Hadoop and Apache ZooKeeper.
Quickly and easily load data, and manage according to policy.Apache Falcon provides policy-based workflows for data governance, while Apache Flume and Sqoop enable easy data ingestion, as do the NFS and WebHDFS interfaces to HDFS.
- **Apache Falcon**Falcon is a data management framework for simplifying data lifecycle management and processing pipelines on Apache Hadoop®. It enables users to orchestrate data motion, pipeline processing,disaster recovery, and data retention workflows.
- **Apache Flume **Flume allows you to efficiently aggregate and move large amounts of log data from many different sources to Hadoop.
- **Apache Sqoop **Sqoop is a tool that speeds and eases movement of data in and out of Hadoop. It provides a reliable parallel load for various, popular enterprise data sources.
Address requirements of Authentication, Authorization, Accounting and Data Protection. Security is provided at every layer of the Hadoop stack from HDFS and YARN to Hive and the other Data Access components on up through the entire perimeter of the cluster via Apache Knox.
- **Apache Knox**The Knox Gateway (“Knox”) provides a single point of authentication and access for Apache Hadoop services in a cluster. The goal of the project is to simplify Hadoop security for users who access the cluster data and execute jobs, and for operators who control access to the cluster.
- **Apache Ranger**Apache Ranger delivers a comprehensive approach to security for a Hadoop cluster. It provides central security policy administration across the core enterprise security requirements of authorization, accounting and data protection.
Provision, manage, monitor and operate Hadoop clusters at scale.
- **Apache Ambari**An open source installation lifecycle management, administration and monitoring system for Apache Hadoop clusters.
- **Apache Oozie**Oozie Java Web application used to schedule Apache Hadoop jobs. Oozie combines multiple jobs sequentially into one logical unit of work.
- **Apache ZooKeeper**A highly available system for coordinating distributed processes. Distributed applications use ZooKeeper to store and mediate updates to important configuration information.
Apache Hadoop can be useful across a range of use cases spanning virtually every vertical industry. It is becoming popular anywhere that you need to store, process, and analyze large volumes of data. Examples include digital marketing automation, fraud detection and prevention, social network and relationship analysis, predictive modeling for new drugs, retail in-store behavior analysis, and mobile device location-based marketing. To learn more about Apache Hadoop, watch the following introduction:
Hortonworks Data Platform is a packaged software hadoop distribution that aim to ease deployment and management of Hadoop clusters compared with simply downloading the various Apache code bases and trying to run them together a system. Architected, developed, and built completely in the open, Hortonworks Data Platform (HDP) provides an enterprise ready data platform that enables organizations to adopt a Modern Data Architecture.
With YARN as its architectural center it provides a data platform for multi-workload data processing across an array of processing methods – from batch through interactive to real-time, supported by key capabilities required of an enterprise data platform — spanning Governance, Security and Operations.
The Hortonworks Sandbox is a single node implementation of the Hortonworks Data Platform (HDP). It is packaged as a virtual machine to make evaluation and experimentation with HDP fast and easy. The tutorials and features in the Sandbox are oriented towards exploring how HDP can help you solve your business big data problems. The Sandbox tutorials will walk you through bringing some sample data into HDP and manipulating it using the tools built into HDP. The idea is to show you how you can get started and show you how to accomplish tasks in HDP. HDP is free to download and use in your enterprise and you can download it here: HDP on Sandbox
NOTE**
- Hadoop Common
- Hadoop YARN
- Hadoop MapReduce.
To learn more about HDFS watch the following HDFS introduction video.
To learn more about YARN watch the following YARN introduction video.
Hadoop 2.0 Blogs:
Hadoop 2.7.0 Blog
Understanding Hadoop 2.0
Concepts: HDFS
A single physical machine gets saturated with its storage capacity as the data grows. Thereby comes impending need to partition your data across separate machines. This type of File system that manages storage of data across a network of machines is called Distributed File Systems. HDFS is a core component of Apache Hadoop and is designed to store large files with streaming data access patterns, running on clusters of commodity hardware. With Hortonworks Data Platform HDP 2.2, HDFS is now expanded to support heterogeneous storage media within the HDFS cluster.
Goals of this module
- Understanding HDFS architecture
- Understanding Hortonworks Sandbox Amabri File User View
Hadoop Distributed File System
HDFS is a distributed file system that is designed for storing large data files. HDFS is a Java-based file system that provides scalable and reliable data storage, and it was designed to span large clusters of commodity servers. HDFS has demonstrated production scalability of up to 200 PB of storage and a single cluster of 4500 servers, supporting close to a billion files and blocks. HDFS is a scalable, fault-tolerant, distributed storage system that works closely with a wide variety of concurrent data access applications, coordinated by YARN. HDFS will “just work” under a variety of physical and systemic circumstances. By distributing storage and computation across many servers, the combined storage resource can grow linearly with demand while remaining economical at every amount of storage.
An HDFS cluster is comprised of a NameNode, which manages the cluster metadata, and DataNodes that store the data. Files and directories are represented on the NameNode by inodes. Inodes record attributes like permissions, modification and access times, or namespace and disk space quotas.
The file content is split into large blocks (typically 128 megabytes), and each block of the file is independently replicated at multiple DataNodes. The blocks are stored on the local file system on the DataNodes.
The Namenode actively monitors the number of replicas of a block. When a replica of a block is lost due to a DataNode failure or disk failure, the NameNode creates another replica of the block. The NameNode maintains the namespace tree and the mapping of blocks to DataNodes, holding the entire namespace image in RAM.
The NameNode does not directly send requests to DataNodes. It sends instructions to the DataNodes by replying to heartbeats sent by those DataNodes. The instructions include commands to:
- replicate blocks to other nodes,
- remove local block replicas,
- re-register and send an immediate block report, or
- shut down the node.
For more details on HDFS: http://hortonworks.com/hadoop/hdfs/
With next generation HDFS data architecture that comes with HDP 2.0, HDFS has evolved to provide automated failure with a hot standby, with full stack resiliency. Please spare some time to go through this video for more clarity on HDFS.
Ambari Files User View on Hortonworks Sandbox
Ambari Files User View
Ambari Files User View provides a user friendly interface to upload, store and move data. Underlying all components in Hadoop is the Hadoop Distributed File System(HDFS™). This is the foundation of the Hadoop cluster. The HDFS file system manages how the datasets are stored in the Hadoop cluster. It is responsible for distributing the data across the datanodes, managing replication for redundancy and administrative tasks like adding, removing and recovery of data nodes.
NOTE**
- Hadoop Common
- Hadoop YARN
- Hadoop MapReduce.
To learn more about HDFS watch the following HDFS introduction video.
To learn more about YARN watch the following YARN introduction video.
HDFS Blogs:
Heterogeneous Storage Policies in HDP 2.2
HDFS Metadata Directories Explained
Heterogeneous Storages in HDFS
HDFS 2.0 Next Generation Architecture
NameNode High Availability in HDP 2.0
Introducing… Tez: Accelerating processing of data stored in HDFS
Concepts: MapReduce & YARN
Cluster computing faces several challenges such as how to store data persistently and keep it available if nodes fail or how to deal with node failures during a long running computation. Then there is network bottleneck which delays the time of processing data. MapReduce offers a solution by bring computation close to data thereby minimizing data movement. It is a simple programming model designed for process large volumes of data in parallel by dividing the job into a set of independent tasks.
The biggest limitation with MapReduce programming is that map and reduce jobs are not stateless. This means that Reduce jobs have to wait for map jobs to be completed first. This limits maximum parallelism and therefore YARN was born as a generic resource management and distributed application framework.
Goals of the Module
- Understanding Map and Reduce jobs.
- Understanding YARN
MapReduce is the key algorithm that the Hadoop data processing engine uses to distribute work around a cluster. A MapReduce job splits a large data set into independent chunks and organizes them into key, value pairs for parallel processing. This parallel processing improves the speed and reliability of the cluster, returning solutions more quickly and with greater reliability.
The Map function divides the input into ranges by the InputFormat and creates a map task for each range in the input. The JobTracker distributes those tasks to the worker nodes. The output of each map task is partitioned into a group of key-value pairs for each reduce.
- map(key1,value) -> list<key2,value2>
The Reduce function then collects the various results and combines them to answer the larger problem that the master node needs to solve. Each reduce pulls the relevant partition from the machines where the maps executed, then writes its output back into HDFS. Thus, the reduce is able to collect the data from all of the maps for the keys and combine them to solve the problem.
- reduce(key2, list) -> list
The current Apache Hadoop MapReduce System is composed of the JobTracker, which is the master, and the per-node slaves called TaskTrackers. The JobTracker is responsible for _resource management _(managing the worker nodes i.e. TaskTrackers), tracking resource consumption/availability and also job life-cycle management (scheduling individual tasks of the job, tracking progress, providing fault-tolerance for tasks etc).
The TaskTracker has simple responsibilities – launch/teardown tasks on orders from the JobTracker and provide task-status information to the JobTracker periodically.
The Apache Hadoop projects provide a series of tools designed to solve big data problems. The Hadoop cluster implements a parallel computing cluster using inexpensive commodity hardware. The cluster is partitioned across many servers to provide a near linear scalability. The philosophy of the cluster design is to bring the computing to the data. So each datanode will hold part of the overall data and be able to process the data that it holds. The overall framework for the processing software is called MapReduce. Here’s a short video introduction to MapReduce:
Apache YARN** (Yet Another Resource Negotiator):**
Hadoop HDFS is the data storage layer for Hadoop and MapReduce was the data-processing layer in Hadoop 1x. However, the MapReduce algorithm, by itself, isn’t sufficient for the very wide variety of use-cases we see Hadoop being employed to solve. Hadoop 2.0 presents YARN, as a generic resource-management and distributed application framework, whereby, one can implement multiple data processing applications customized for the task at hand. The fundamental idea of YARN is to split up the two major responsibilities of the JobTracker i.e. resource management and job scheduling/monitoring, into separate daemons: a global ResourceManager and per-application ApplicationMaster (AM).
The ResourceManager and per-node slave, the NodeManager (NM), form the new, and generic, systemfor managing applications in a distributed manner.
The ResourceManager is the ultimate authority that arbitrates resources among all the applications in the system. The per-application ApplicationMaster is, in effect, a _framework specific _entity and is tasked with negotiating resources from the ResourceManager and working with the NodeManager(s) to execute and monitor the component tasks.
ResourceManager has a pluggable Scheduler, which is responsible for allocating resources to the various running applications subject to familiar constraints of capacities, queues etc. The Scheduler is a _pure scheduler _in the sense that it performs no monitoring or tracking of status for the application, offering no guarantees on restarting failed tasks either due to application failure or hardware failures. The Scheduler performs its scheduling function based on the _resource requirements _of the applications; it does so based on the abstract notion of a **_Resource Container _**which incorporates resource elements such as memory, cpu, disk, network etc.
NodeManager is the per-machine slave, which is responsible for launching the applications’ containers, monitoring their resource usage (cpu, memory, disk, network) and reporting the same to the ResourceManager.
The per-application ApplicationMaster has the responsibility of negotiating appropriate resource containers from the Scheduler, tracking their status and monitoring for progress. From the system perspective, the ApplicationMaster itself runs as a normal container.
Here is an architectural view of YARN:
One of the crucial implementation details for MapReduce within the new YARN system that I’d like to point out is that we have reused the existing MapReduce framework without any major surgery. This was very important to ensure compatibility for existing MapReduce applications and users. Here is a short video introduction for YARN
NOTE**
- Hadoop Common
- Hadoop YARN
- Hadoop MapReduce.
To learn more about HDFS watch the following HDFS introduction video.
To learn more about YARN watch the following YARN introduction video.
Hadoop 2.0 Blogs:
Hadoop 2.7.0 Blog
Understanding Hadoop 2.0
YARN Blogs:
YARN series-1
YARN series-2
Slider Blogs:
Announcing Apache Slider 0.60.0
Onboarding Long Running Services to Apache Hadoop YARN Using Apache Slider
Build YARN Apps on Hadoop with Apache Slider: Technical Preview Now Available
Capacity Scheduler Blogs:
Understanding Apache Hadoop’s Capacity Scheduler
Configuring YARN Capacity Scheduler with Ambari
Multi-Tenancy in HDP 2.0: Capacity Scheduler and YARN
Better SLAs via Resource-preemption in YARN’s Capacity Scheduler
Concepts: Hive and Pig
Hive is a SQL like query language that enables analysts familiar with SQL to run queries on large volumes of data. Hive has three main functions: data summarization, query and analysis. Hive provides tools that enable easy data extraction, transformation and loading (ETL).
Goals of the module
- Understanding Apache Hive
- Understanding Apache Tez
- Understanding Ambari Hiver User Views on Hortonworks Sandbox
Data analysts use Hive to explore, structure and analyze that data, then turn it into business insight. Hive implements a dialect of SQL (Hive QL) that focuses on analytics and presents a rich set of SQL semantics including OLAP functions, sub-queries, common table expressions and more. Hive allows SQL
developers or users with SQL tools to easily query, analyze and process data stored in Hadoop.Hive also allows programmers familiar with the MapReduce framework to plug in their custom mappers and reducers to perform more sophisticated analysis that may not be supported by the built-in capabilities of the language.
Hive users have a choice of 3 runtimes when executing SQL queries. Users can choose between Apache Hadoop MapReduce, Apache Tez or
Apache Spark frameworks as their execution backend
Here are some advantageous characteristics of Hive for enterprise SQL in Hadoop:
Feature Description
Familiar Query data with a SQL-based language
Fast Interactive response times, even over huge datasets
Scalable and Extensible As data variety and volume grows, more commodity machines can be added, without a corresponding reduction in performance
How Hive Works
The tables in Hive are similar to tables in a relational database, and data units are organized in a taxonomy from larger to more granular units. Databases are comprised of tables, which are made up of partitions. Data can be accessed via a simple query language and Hive supports overwriting or appending data.
Within a particular database, data in the tables is serialized and each table has a corresponding Hadoop Distributed File System (HDFS) directory. Each table can be sub-divided into partitions that determine how data is distributed within sub-directories of the table directory. Data within partitions can be further broken down into buckets.
Components of Hive
- HCatalog is a component of Hive. It is a table and storage management layer for Hadoop that enables users with different data processing tools — including Pig and MapReduce — to more easily read and write data on the grid. HCatalog holds a set of files paths and metadata about data in a Hadoop cluster. This allows scripts, MapReduce and Tez, jobs to be decoupled from data location and metadata like the schema. Additionally, since HCatalog also supports tools like Hive and Pig, the location and metadata can be shared between tools. Using the open APIs of HCatalog external tools that want to integrate, such as Teradata Aster, can also use leverage file path location and metadata in HCatalog.
NOTE**
- WebHCat provides a service that you can use to run Hadoop MapReduce (or YARN), Pig, Hive jobs or perform Hive metadata operations using an HTTP (REST style) interface.
Here is a short video introduction on Hive:
Apache™ Tez is an extensible framework for building high performance batch and interactive data processing applications, coordinated by YARN in Apache Hadoop. Tez improves the MapReduce paradigm by dramatically improving its speed, while maintaining MapReduce’s ability to scale to petabytes of data. Important Hadoop ecosystem projects like Apache Hive and Apache Pig use Apache Tez, as do a growing number of third party data access applications developed for the broader Hadoop ecosystem.
Apache Tez provides a developer API and framework to write native YARN applications that bridge the spectrum of interactive and batch workloads. It allows those data access applications to work with petabytes of data over thousands nodes. The Apache Tez component library allows developers to create Hadoop applications that integrate natively with Apache Hadoop YARN and perform well within mixed workload clusters.
Since Tez is extensible and embeddable, it provides the fit-to-purpose freedom to express highly optimized data processing applications, giving them an advantage over end-user-facing engines such as MapReduce and Apache Spark. Tez also offers a customizable execution architecture that allows users to express complex computations as dataflow graphs, permitting dynamic performance optimizations based on real information about the data and the resources required to process it.
Here is a short video introduction on Tez.
The Stinger Initiative was started to enable Hive to support an even broader range of use cases at truly Big Data scale: bringing it beyond its Batch roots to support interactive queries – all with a common SQL access layer.
Stinger.next is a continuation of this initiative focused on even further enhancing the speed, scale and breadth of SQL support to enable truly real-time access in Hive while also bringing support for transactional capabilities. And just as the original Stinger initiative did, this will be addressed through a familiar three-phase delivery schedule and developed completely in the open Apache Hive community.
Ambari Hive User Views on Hortonworks Sandbox
To make it easy to interact with Hive we use a tool in the Hortonworks Sandbox called the Ambari Hive User View. Ambari Hive User View provides an interactive interface to Hive. We can create, edit, save and run queries, and have Hive evaluate them for us using a series of MapReduce jobs or Tez jobs.
Let’s now open the Ambari Hive User View and get introduced to the environment, go to the Ambari User VIew icon and select Hive :
Ambari Hive User View
Now let’s take a closer look at the SQL editing capabilities in the User View:
- There are four tabs to interact with SQL: 1. Query: This is the interface shown above and the primary interface to write, edit and execute new SQL statements 2. Saved Queries: You can save your favorite queries and quickly have access to them to rerun or edit. 3. History: This allows you to look at past queries or currently running queries to view, edit and rerun. It also allows you to see all SQL queries you have authority to view. For example, if you are an operator and an analyst needs help with a query, then the Hadoop operator can use the History feature to see the query that was sent from the reporting tool. 4. UDFs: Allows you to define UDF interfaces and associated classes so you can access them from the SQL editor.
- Database Explorer: The Database Explorer helps you navigate your database objects. You can either search for a database object in the Search tables dialog box, or you can navigate through Database -> Table -> Columns in the navigation pane.
- The principle pane to write and edit SQL statements. This editor includes content assist via CTRL + Space to help you build queries. Content assist helps you with SQL syntax and table objects.
- Once you have created your SQL statement you have 3 options: 1. Execute: This runs the SQL statement. 2. Explain: This provides you a visual plan, from the Hive optimizer, of how the SQL statement will be executed. 3. Save as: Allows you to persist your queries into your list of saved queries.
- When the query is executed you can see the Logs or the actual query results. 1. Logs: When the query is executed you can see the logs associated with the query execution. If your query fails this is a good place to get additional information for troubleshooting. 2. Results: You can view results in sets of 50 by default.
- There are four sliding views on the right hand side with the following capabilities, which are in context of the tab you are in: 1. Query: This is the default operation,which allows you to write and edit SQL. 2. Settings: This allows you to set properties globally or associated with an individual query. 3. Visual Explain: This will generate an explain for the query. This will also show the progress of the query. 4. TEZ: If you use TEZ as the query execution engine then you can view the DAG associated with the query. This integrates the TEZ User View so you can check for correctness and helps with performance tuning by visualizing the TEZ jobs associated with a SQL query. 5. Notifications: This is how to get feedback on query execution.
The Apache Hive project provides a data warehouse view of the data in HDFS. Using a SQL dialect, HiveQL (HQL), Hive lets you create summarizations of your data and perform ad-hoc queries and analysis of large datasets in the Hadoop cluster. The overall approach with Hive is to project a table structure on the dataset and then manipulate it with SQL. The notion of projecting a table structure on a file is often referred to as Schema-On-Read. Since you are using data in HDFS, your operations can be scaled across all the datanodes and you can manipulate huge datasets.
MapReduce allows allows you to specify map and reduce functions, but working out how to fit your data processing into this pattern may sometimes require you to write multiple MapReduce stages. With Pig, data structures are much richer and the transformations you can apply to data are much more powerful.
Goals of this Module
- Understanding Apache Pig
- Understanding Apache Pig on Tez
- Understanding Ambari Pig User Views on Hortonworks Sandbox
Apache Pig allows Apache Hadoop users to write complex MapReduce transformations using a simple scripting language called Pig Latin. Pig translates the Pig Latin script into MapReduce so that it can be executed within YARN for access to a single dataset stored in the Hadoop Distributed File System (HDFS).
Pig was designed for performing a long series of data operations, making it ideal for three categories of Big Data jobs:
- Extract-transform-load (ETL) data pipelines,
- Research on raw data, and
- Iterative data processing.
Whatever the use case, Pig will be:
Characteristic Benefit
Extensible Pig users can create custom functions to meet their particular processing requirements
Easily programmed Complex tasks involving interrelated data transformations can be simplified and encoded as data flow sequences. Pig programs accomplish huge tasks, but they are easy to write and maintain.
Self-optimizing Because the system automatically optimizes execution of Pig jobs, the user can focus on semantics.
Please refer the following video on Pig for more clarity:
How Pig Works
Pig runs on Apache Hadoop YARN and makes use of MapReduce and the Hadoop Distributed File System (HDFS). The language for the platform is called Pig Latin, which abstracts from the Java MapReduce idiom into a form similar to SQL. While SQL is designed to query the data, Pig Latin allows you to write a data flow that describes how your data will be transformed (such as aggregate, join and sort).
Since Pig Latin scripts can be graphs (instead of requiring a single output) it is possible to build complex data flows involving multiple inputs, transforms, and outputs. Users can extend Pig Latin by writing their own functions, using Java, Python, Ruby, or other scripting languages. Pig Latin is sometimes extended using UDFs (User Defined Functions), which the user can write in any of those languages and then call directly from the Pig Latin.
The user can run Pig in two modes, using either the “pig” command or the “java” command:
- MapReduce Mode. This is the default mode, which requires access to a Hadoop cluster. The cluster may be a pseudo- or fully distributed one.
- Local Mode. With access to a single machine, all files are installed and run using a local host and file system
Ambari Pig User Views on Hortonworks Sandbox**:**
To get to the Ambari Pig User View on Sandbox, click on the User Views icon at top right and select Pig:
This will bring up the Ambari Pig User View interface. Your Pig View does not have any scripts to display, so it will look like the following:
On the left is a list of your scripts, and on the right is a composition box for writing scripts. A special feature of the interface is the Pig helper at the bottom. The Pig helper will provide us with templates for the statements, functions, I/O statements, HCatLoader() and Python user defined functions. At the very bottom are status areas that will show the results of our script and log files.
The following screenshot shows and describes the various components and features of the Pig User View:
Suggested Readings
NOTE**
Cost-Based Optimizer Makes Apache Hive 0.14 More Than 2.5X Faster
Discover HDP 2.2: Even Faster SQL Queries with Apache Hive and Stinger.next
Announcing Apache Hive 1.2
HIVE 0.14 Cost Based Optimizer (CBO) Technical Overview
5 Ways to Make Your Hive Queries Run Faster
Secure JDBC and ODBC Clients’ Access to HiveServer2
Speed, Scale and SQL: The Stinger Initiative, Apache Hive 12 & Apache Tez
Hive/HCatalog – Data Geeks & Big Data Glue** Tez Blogs**:
Apache Tez: A New Chapter in Hadoop Data Processing
Data Processing API in Apache TezORC Blogs:
Apache ORC Launches as a Top-Level Project
ORCFile in HDP 2: Better Compression, Better Performance
Lab 0: Set-up
Start the HDP Sandbox following the Sandbox Install Guide to start the VM:
Once you have installed the Sandbox VM, it resolves to the host on your environment, the address of which varies depending upon the Virtual Machine you are using(Vmware, VirtualBox etc). As, a general thumb rule, wait for the installation to complete and confirmation screen will tell you the host your sandbox resolves to. For example:
In case of VirtualBox: host would be 127.0.0.1
If you are using a private cluster or a cloud to run sandbox. Please find the host your sandbox resolves to.
Append the port number :8888 to your host address, open your browser, and access Sandbox Welcome page at http://host:8888/.
Navigate to Ambari welcome page using the url given on Sandbox welcome page.
Both the username and password to login are admin.
NOTE**
- login using username as “root” and password as “hadoop”.
- Type ifconfig and look for inet address under eth.
- Use the inet address, append :8080 and open it into a browser. It shall direct you to Ambari login page.
- This inet address is randomly generated for every session and therefore differs from session to session.
The following table has some useful URLs as well:
Service URL
Sandbox welcome page http://host:8888
Ambari Dashboard http://host:8080
Ambari Welcome http://host:8080/views/ADMIN_VIEW/2.1.0/INSTANCE/#/
Hive User View http://host:8080/#/main/views/HIVE/1.0.0/Hive
Pig User View http:/host:8080/#/main/views/PIG/0.1.0/MyPig
FIle User View http://host:8080/#/main/views/FILES/0.2.0/MyFiles
SSH web Client http://host:4200
Hadoop Configuration
http://host:50070/dfshealth.html
http://host:50070/explorer.html
Service User Password
Ambari admin admin
Linux OS root hadoop
Enter the Ambari Welcome URL and then you should see a similar screen:
There are 5 key capabilities to explore in the Ambari Welcome screen:
- “Operate Your Cluster” will take you to the Ambari Dashboard which is the primary UI for Hadoop Operators
- “Manage Users + Groups” allows you to add & remove Ambari users and groups
- “Clusters” allows you to grant permission to Ambari users and groups
- “Ambari User Views” list the set of Ambari Users views that are part of the cluster
- “Deploy Views” provides administration for adding and removing Ambari User Views
Take a few minutes to quickly explore these 5 capabilities and to become familiar their features.
Enter the Ambari Dashboard URL and you should see a similar screen:
Briefly skim through the Ambari Dashboard links (circled above) by clicking on
- Metrics, Heatmap and Configuration
and then the
- Dashboard, Services, Hosts, Alerts, Admin and User Views icon (represented by 3×3 matrix ) to become familiar with the Ambari resources available to you.
NOTE**
If you have questions, feedback or need help getting your environment ready visit developer.hortonworks.com. Please also explore the HDP documentation. To ask a question check out the Hortonworks Forums.
Lab 1: HDFS - Loading Data
Introduction:
In this section you will download the sensor data and load that into HDFS using Ambari User Views. You will get introduced to the Ambari Files User View to manage files. You can perform tasks like create directories, navigate file systems and upload files to HDFS. In addition you’ll perform a few other file-related tasks as well. Once you get the basics, you will create two directories and then load two files into HDFS using the Ambari Files User View.
Prerequisites:
The tutorial is a part of series of hands on tutorial to get you started on HDP using Hortonworks sandbox. Please ensure you complete the prerequisites before proceeding with this tutorial.
- Lab 0 (Hortonworks sandbox set up)
- Allow yourself around half an hour to complete this tutorial.
Outline:
- HDFS backdrop
- Step 1.1: Download data – Geolocation.zip
- Step 1.2: Load Data into HDFS
- Suggested readings
HDFS backdrop:
A single physical machine gets saturated with its storage capacity as the data grows. Thereby comes impending need to partition your data across separate machines. This type of File system that manages storage of data across a network of machines is called Distributed File Systems. HDFS is a core component of Apache Hadoop and is designed to store large files with streaming data access patterns, running on clusters of commodity hardware. With Hortonworks Data Platform HDP 2.2, HDFS is now expanded to support heterogeneous storage media within the HDFS cluster.
Step 1.1: Download and Extract the Sensor Data Files
- You can download the sample sensor data contained in a compressed (.zip) folder here: Geolocation.zip
- Save the Geolocation.zip file to your computer, then extract the files. You should see a Geolocation folder that contains the following files:
- geolocation.csv – This is the collected geolocation data from the trucks. it contains records showing truck location, date, time, type of event, speed, etc.
- trucks.csv – This is data was exported from a relational database and it shows info on truck model, driverid, truckid, and aggregated mileage info.
Step 1.2: Load the Sensor Data into HDFS
- Go to the Ambari Dashboard and open the HDFS User View by click on the User Views icon and selecting the HDFS Files menu item.
- Starting from the top root of the HDFS file system, you will see all the files the logged in user (admin in this case) has access to see:
- Click tmp. Then click
button to create the
/tmp/admindirectory and then create the/tmp/admin/datadirectory.
- Now traverse to the
/tmp/admin/datadirectory and upload the corresponding geolocation.csv and trucks.csv files into it.
You can also perform the following operations on a file by right clicking on the file: Download, Move, Permissions, Rename and Delete.
**Suggested Readings **
NOTE
**
Ambari Files View
HDFS Administration
Lab 1: HDFS - Loading Data
Introduction:
In this section you will download the sensor data and load that into HDFS using Ambari User Views. You will get introduced to the Ambari Files User View to manage files. You can perform tasks like create directories, navigate file systems and upload files to HDFS. In addition you’ll perform a few other file-related tasks as well. Once you get the basics, you will create two directories and then load two files into HDFS using the Ambari Files User View.
Prerequisites:
The tutorial is a part of series of hands on tutorial to get you started on HDP using Hortonworks sandbox. Please ensure you complete the prerequisites before proceeding with this tutorial.
- Lab 0 (Hortonworks sandbox set up)
- Allow yourself around half an hour to complete this tutorial.
Outline:
- HDFS backdrop
- Step 1.1: Download data – Geolocation.zip
- Step 1.2: Load Data into HDFS
- Suggested readings
HDFS backdrop:
A single physical machine gets saturated with its storage capacity as the data grows. Thereby comes impending need to partition your data across separate machines. This type of File system that manages storage of data across a network of machines is called Distributed File Systems. HDFS is a core component of Apache Hadoop and is designed to store large files with streaming data access patterns, running on clusters of commodity hardware. With Hortonworks Data Platform HDP 2.2, HDFS is now expanded to support heterogeneous storage media within the HDFS cluster.
Step 1.1: Download and Extract the Sensor Data Files
- You can download the sample sensor data contained in a compressed (.zip) folder here: Geolocation.zip
- Save the Geolocation.zip file to your computer, then extract the files. You should see a Geolocation folder that contains the following files:
- geolocation.csv – This is the collected geolocation data from the trucks. it contains records showing truck location, date, time, type of event, speed, etc.
- trucks.csv – This is data was exported from a relational database and it shows info on truck model, driverid, truckid, and aggregated mileage info.
Step 1.2: Load the Sensor Data into HDFS
- Go to the Ambari Dashboard and open the HDFS User View by click on the User Views icon and selecting the HDFS Files menu item.
- Starting from the top root of the HDFS file system, you will see all the files the logged in user (admin in this case) has access to see:
- Click tmp. Then click
/tmp/admindirectory and then create the/tmp/admin/datadirectory.
- Now traverse to the
/tmp/admin/datadirectory and upload the corresponding geolocation.csv and trucks.csv files into it.
You can also perform the following operations on a file by right clicking on the file: Download, Move, Permissions, Rename and Delete.
**Suggested Readings **
NOTE
**
Ambari Files View
HDFS Administration
Lab 3: Pig - Risk Factor
In this tutorial you will be introduced to Apache Pig. In the earlier section of lab you learned how to load data into HDFS and then manipulate it using Hive. We are using the Truck sensor data to better understand risk associated with every driver. This section will teach you to compute risk using Apache Pig.
Prerequisites:
The tutorial is a part of series of hands on tutorial to get you started on HDP using Hortonworks sandbox. Please ensure you complete the prerequisites before proceeding with this tutorial.
- Lab 0 (Hortonworks sandbox set up)
- Lab 1: Loading sensor data into HDFS
- Lab 2: Data Manipulation with Apache Hive
- Allow yourself around one hour to complete this tutorial.
Outline:
- Pig basics
- Step 3.1: Define Table schema
- Step 3.2: Create Pig Script
- Step 3.3: Quick Recap
- Step 3.4: Execute Pig Script on Tez
- Suggested readings
Pig Basics:
Pig is a high level scripting language that is used with Apache Hadoop. Pig enables data workers to write complex data transformations without knowing Java. Pig’s simple SQL-like scripting language is called Pig Latin, and appeals to developers already familiar with scripting languages and SQL.
Pig is complete, so you can do all required data manipulations in Apache Hadoop with Pig. Through the User Defined Functions(UDF) facility in Pig, Pig can invoke code in many languages like JRuby, Jython and Java. You can also embed Pig scripts in other languages. The result is that you can use Pig as a component to build larger and more complex applications that tackle real business problems.
Pig works with data from many sources, including structured and unstructured data, and store the results into the Hadoop Distributed File System.
Pig scripts are translated into a series of MapReduce jobs that are run on the Apache Hadoop cluster.
Step 3.1: Define table schema
Now we have refined the truck data to get the average mpg for each truck. The next task is to compute the risk factor for each driver which is the total miles driven/abnormal events. We can get the event information from the geolocation table.
If we look at the truck_mileage table, we we have the driverid and the number of miles for each trip. To get the total miles for each driver, we can group those records by driverid and then sum the miles.
- We will start by creating a table named driver_mileage that is created from a query of the columns we want from truck_mileage. The following query groups the records by driverid and sums the miles in the select statement. Execute this query in a new Worksheet:
— Create table DriverMileage from existing truck_mileage data
CREATE TABLE DriverMileage STORED AS ORC AS SELECT driverid, sum(miles) totmiles FROM truck_mileage GROUP BY driverid;
- View the data generated by the script by clicking the Load sample data icon in the Database Explorer next to drivermileage. The results should look like:
- Next, you will use Pig to compute the risk factor of each driver. Before we can run the Pig code, one of the requirements for the HCatStorer() class is that the table must already exist in Hive. The Pig code expects the following structure for a table named riskfactor. Execute the following DDL command:
— Create table avg_mileage from existing trucks_mileage data
CREATE TABLE riskfactor (driverid string,events bigint,totmiles bigint,riskfactor float)STORED AS ORC;
- Verify the riskfactor table was created successfully. It will be empty now, but you will populate it from a Pig script. You are now ready to compute the risk factor using Pig. Let’s take a look at Pig and how to execute Pig scripts from within Ambari.
Step 3.2: Create Pig Script
In this tutorial we create and run a Pig script. We will use the Ambari Pig User View. Let’s get started…
NOTE
**** a. Log in to Ambari Pig User Views****
To get to the Ambari Pig User View, click on the User Views icon at top right and select Pig:
This will bring up the Ambari Pig User View interface. Your Pig View does not have any scripts to display, so it will look like the following:
On the left is a list of your scripts, and on the right is a composition box for writing scripts. A special feature of the interface is the Pig helper at the bottom. The Pig helper will provide us with templates for the statements, functions, I/O statements, HCatLoader() and Python user defined functions. At the very bottom are status areas that will show the results of our script and log files.
The following screenshot shows and describes the various components and features of the Pig User View:
Let’s enter a Pig script. Click the New Script button in the upper-right corner of the view:
Name the script riskfactor.pig, then click the Create button:
We are going to use HCatalog to load data into Pig. HCatalog allows us to share schema across tools and users within our Hadoop environment. It also allows us to factor out schema and location information from our queries and scripts and centralize them in a common repository. Since it is in HCatalog we can use the HCatLoader() function. Pig makes it easy by allowing us to give the table a name or alias and not have to worry about allocating space and defining the structure. We just have to worry about how we are processing the table.
-
We can use the Pig helper at the bottom of the screen to give us a template for the line. Click on Pig helper -> HCatalog->load template
-
The entry %TABLE% is highlighted in red for us. Type the name of the table which is geolocation.
-
Remember to add the a = before the template. This saves the results into a. Note the ‘=’ has to have a space before and after it.
-
Our completed line of code will look like:
a = LOAD 'geolocation' using org.apache.hive.hcatalog.pig.HCatLoader();
Copy-and-paste the above Pig code into the riskfactor.pig window.
**NOTE **
- You can use the command
DUMP %VAR%if you want to view the data inside. - Just replace
%VAR%with the variable you wish to view.
** d. Filter your data set**
The next step is to select a subset of the records so that we just have the records of drivers for which the event is not normal. To do this in Pig we use the Filter operator. We tell Pig to Filter our table and keep all records where event !=“normal” and store this in b. With this one simple statement Pig will look at each record in the table and filter out all the ones that do not meet our criteria.
-
We can use Pig Help again by clicking on Pig helper->Relational Operators->FILTER template
-
We can replace %VAR% with “a” (hint: tab jumps you to the next field)
-
Our %COND% is “event !=’normal’; ” (note: single quotes are needed around normal and don’t forget the trailing semi-colon)
-
Complete line of code will look like:
b = filter a by event != 'normal';
Copy-and-paste the above Pig code into the riskfactor.pig window.
** e. Iterate your data set**
Now that we have the right set of records we can iterate through them. We use the “foreach” operator on the grouped data to iterate through all the records. We would also like to know how many times a driver has a non normal event associated with him. to achieve this we add ‘1’ to every row in the data set.
-
Pig helper ->Relational Operators->FOREACH template will get us the code
-
Our %DATA% is b and the second %NEW_DATA% is “driverid,event,(int) ‘1’ as occurance;”
-
Complete line of code will look like:
c = foreach b generate driverid, event, (int) '1' as occurance;
Copy-and-paste the above Pig code into the riskfactor.pig window:
** f. Calculate the total non normal events for each driver**
The group statement is important because it groups the records by one or more relations. In this case we would like to group by driver id and iterate over each row again to sum the non normal events.
-
Pig helper ->Relational Operators->GROUP %VAR% BY %VAR% template will get us the code
-
First %VAR% takes “c” and second %VAR% takes “driverid;”
-
Complete line of code will look like:
d = group c by driverid;
Copy-and-paste the above Pig code into the riskfactor.pig window.
-
Next use Foreach statement again to add the occurance.
e = foreach d generate group as driverid, SUM(c.occurance) as t_occ;
** g. Load drivermileage table and perform a join operation**
In this section we will load drivermileage table into Pig using Hcatlog and perform a join operation on driverid. The resulting data set will give us total miles and total non normal events for a particular driver.
-
Load drivermileage using HcatLoader()
g = LOAD 'drivermileage' using org.apache.hive.hcatalog.pig.HCatLoader();
-
Pig helper ->Relational Operators->JOIN %VAR% BY template will get us the code
-
Replace %VAR% by ‘e’ and after BY put ‘driverid, g by driverid;’
-
Complete line of code will look like:
h = join e by driverid, g by driverid;
Copy-and-paste the above two Pig codes into the riskfactor.pig window.
** h. Compute Driver Risk factor**
In this section we will associate a driver risk factor with every driver. Driver risk factor will be calculated by dividing total miles travelled by non normal event occurrences.
-
We will use Foreach statement again to compute driver risk factor for each driver.
-
Use the following code and paste it into your Pig script.
final_data = foreach h generate $0 as driverid, $1 as events, $3 as totmiles, (float) $3/$1 as riskfactor;
-
As a final step store the data into a table using Hcatalog.
store final_data into ‘riskfactor’ using
Here is the final code and what it will look like once you paste it into the editor.
— Geolocation has data stored in ORC format
a = LOAD 'geolocation' using org.apache.hive.hcatalog.pig.HCatLoader();
b = filter a by event != 'normal';
c = foreach b generate driverid, event, (int) '1' as occurance;
d = group c by driverid;
e = foreach d generate group as driverid, SUM(c.occurance) as t_occ;
g = LOAD 'drivermileage' using org.apache.hive.hcatalog.pig.HCatLoader();
h = join e by driverid, g by driverid;
final_data = foreach h generate $0 as driverid, $1 as events, $3 as totmiles, (float) $3/$1 as riskfactor;
store final_data into 'riskfactor' using org.apache.hive.hcatalog.pig.HCatStorer();
Save the file riskfactor.pig by clicking the Save button in the left-hand column.
Step 3.3: Quick Recap
Before we execute the code, let’s review the code again:
- The line a= loads the geolocation table from HCatalog.
- The line b= filters out all the rows where the event is not ‘Normal’.
- Then we add a column called occurrence and assign it a value of 1.
- We then group the records by driverid and sum up the occurrences for each driver.
- At this point we need the miles driven by each driver, so we load the table we created using Hive.
- To get our final result, we join by the driverid the count of events in e with the mileage data in g.
- Now it is real simple to calculate the risk factor by dividing the miles driven by the number of events
You need to configure the Pig Editor to use HCatalog so that the Pig script can load the proper libraries. In the Pig arguments text box, enter -useHCatalog and click the Add button:
NOTE-useHCatalog is case sensistive
The Arguments section of the Pig View should now look like the following:
Step 3.4: Execute Pig Script on Tez
- You are now ready to execute the script. Click Execute on Tez checkbox and finally hit the blue Execute button to submit the job. Pig job will be submitted to the cluster. This will generate a new tab with a status of the running of the Pig job and at the top you will find a progress bar that shows the job status.
- Wait for the job to complete. The output of the job is displayed in the Results section. Your script does not output any result – it stores the result into a Hive table – so your Results section will be empty.
Click on the Logs twisty to see what happened when your script ran. This is where you will see any error messages. The log may scroll below the edge of your window so you may have to scroll down.
- Go back to the Ambari Hive User View and browse the data in the riskfactor table to verify that your Pig job successfully populated this table. Here is what is should look like:
At this point we now have our truck miles per gallon table and our risk factor table. The next step is to pull this data into Excel to create the charts for the visualization step.
Suggested Readings
NOTE ****
Lab 4: Spark - Risk Factor
Introduction:
In this tutorial you will be introduced to Apache Spark. In the earlier section of lab you learned how to load data into HDFS and then manipulate it using Hive. We are using the Truck sensor data to better understand risk associated with every driver. This section will teach you to compute risk using Apache spark.
NOTE
Prerequisites:
The tutorial is a part of series of hands on tutorial to get you started on HDP using Hortonworks sandbox. Please ensure you complete the prerequisites before proceeding with this tutorial.
- Lab 0 (Hortonworks sandbox set up)
- Lab 1: Loading sensor data into HDFS
- Lab 2: Data Manipulation with Apache Hive
- Allow yourself around one hour to complete this tutorial.
- Apache Spark backdrop
- Apache Spark basics
- Step 4.1: Configure Apache Spark services using Ambari
- Step 4.2: Create a Hive Context
- Step 4.3: Create RDD from Hive Context
- Step 4.4: RDD transformations and actions
- Step 4.5: Load and save data into Hive as ORC
- Suggested readings
Apache Spark backdrop:
MapReduce has been useful, but the amount of time it takes for the jobs to run can at times be exhaustive. Also, MapReduce jobs only work for a specific set of use cases. There is a need for computing framework that works for a wider set of use cases.
Therefore Apache Spark was designed as a computing platform to be fast, general-purpose, and easy to use. It extends the MapReduce model and takes it to a whole other level. The speed comes from the in-memory computations. Applications running in memory allows for a much faster processing and response.
Apache Spark:
Apache Spark is a fast, in-memory data processing engine with elegant and expressive development APIs in Scala,Java, and Python that allow data workers to efficiently execute machine learning algorithms that require fast iterative access to datasets. Spark on Apache Hadoop YARN enables deep integration with Hadoop and other YARN enabled workloads in the enterprise.
You can run batch application such as MapReduce types jobs or iterative algorithms that builds upon each other. You can also run interactive queries and process streaming data with your application. Spark also provides number of libraries which you can easily use to expand beyond the basic Spark capabilities such as Machine Learning algorithms, SQL, streaming, and graph processing. Spark runs on Hadoop clusters such as Hadoop YARN or Apache Mesos, or even as a standalone with its own scheduler.
Lets get started…!!
Step 4.1: Configuring Spark services using Ambari
- Log on to Ambari Dashboard and click on Actions tab at the bottom left corner. Hit Start All to ensure Spark is running. Ambari will take some time to start all services and you can monitor the progress of it.
- Close the Ambari browser and we will get running with some codes on Spark. ssh into the sandbox using root as login and hadoop as password.
login: root
password: hadoop
Optionally, if you don’t have an SSH client installed and configured you can use the built-in web client which can be accessed from here:** http://(host):4200 **(use the same username and password provided above)
- Type the command spark-shell
This will load the default Spark Scala API.
Notice it is already starting with Hive integration as we have preconfigured it on the Hortonworks Sandbox.
Step 4.2: Create a HiveContext
For improved Hive integration, HDP 2.3 offers ORC file support for Spark. This allows Spark to read data stored in ORC files. Spark can leverage ORC file’s more efficient columnar storage and predicate pushdown capability for even faster in-memory processing. HiveContext is an instance of the Spark SQL execution engine that integrates with data stored in Hive. The more basic SQLContext provides a subset of the Spark SQL support that does not depend on Hive. It reads the configuration for Hive from hive-site.xml on the classpath.
-
Import these sql libraries.
import org.apache.spark.sql.hive.orc._ import org.apache.spark.sql._
-
Instantiate HiveContext
val hiveContext = new org.apache.spark.sql.hive.HiveContext(sc)
NOTE**
Step 4.3: Creating a RDD from HiveContext
What is RDD?
Spark’s primary core abstraction is called Resilient Distributed Dataset or RDD. Essentially it is just a distributed collection of elements that is parallelized across the cluster. What this means is which is that RDD is an immutable collection of objects that is partitioned and distributed across multiple physical nodes of a YARN cluster and that can be operated in parallel.
There are three methods for creating a RDD:
- You can parallelize an existing collection. This means that the data already resides within Spark and can now be operated on in parallel.
- The second method to create a RDD, is to reference a dataset. This dataset can come from any storage source supported by Hadoop such as HDFS, Cassandra, HBase etc.
- The third method to create a RDD is from transforming an existing RDD to create a new RDD. We will be using the later two methods in our tutorial.
**** 1. Use a simple show command to see the list of tables in Hive warehouse.****
hiveContext.sql("show tables").collect.foreach(println)
You will notice that geolocation table and driver mileage table that we created in earlier tutorial are already listed in Hive metastore and can be directly queried upon.
**** 2. Query tables to build Spark RDD****
We will do a simple select query to fetch data from geolocation and drivermileage tables to a spark variable. Getting data into Spark this way also allows to copy table schema to RDD.
val geolocation_temp1 = hiveContext.sql("select * from geolocation")
val drivermileage_temp1 = hiveContext.sql("select * from drivermileage")
Make sure that the RDD`s carry the exact data. You can verify through following command
geolocation_temp1.take(10)
drivermileage_temp1.take(10)Both these commands will return 10 rows from respective RDD`s.
**** 3. Registering a Temporary table****
Now let’s give this RDD a name, so that we can use it in Spark SQL statements
geolocation_temp1.registerTempTable("geolocation_temp1")
drivermileage_temp1.registerTempTable("drivermileage_temp1")
Step 4.4: RDD transformations and Actions
Typically, RDDs are instantiated by loading data from a shared filesystem, HDFS, HBase, or any data source offering a Hadoop InputFormat on a YARN cluster.
Once an RDD is instantiated, you can apply a series of operations. All operations fall into one of two types:transformations or actions.
- Transformation operations, as the name suggests, create new datasets from an existing RDD and build out the processing DAG that can then be applied on the partitioned dataset across the YARN cluster. Transformations do not return a value. In fact, nothing is evaluated during the definition of these transformation statements. Spark just creates these Direct Acyclic Graphs or DAG, which will only be evaluated at runtime. We call this lazy evaluation.
- An Action operation, on the other hand, executes DAG and returns a value.
**** 1. Querying against the table****
Now that our schema’s RDD with data has a name, we can use Spark SQL commands to query it. Remember the table below is not a Hive table, it is just a RDD we are querying with SQL.
-
Here we will try to perform iteration and filter operation. First, we need to filter drivers that have non- normal events associated to them and then count the number for non- normal events for each driver.
val geolocation_temp2 = hiveContext.sql("SELECT driverid, count(driverid) occurance from geolocation_temp1 where event!='normal' group by driverid")
NOTE**
-
The resulting table will have count of total non normal events associated to each driver. Register this filtered table as a temporary table so that subsequent SQL queries can be applied on it.
geolocation_temp2.registerTempTable("geolocation_temp2")
-
You can view the result by doing an action operation on RDD.
geolocation_temp2.collect.foreach(println)
- Perform join operation
In this section we will perform a join operation geolocation_temp2 table has details of drivers and count of their respective non-normal events. drivermileage_temp1 table has details of total miles travelled by each driver.
-
We will join two tables on common column, which in our case is driverid.
val joined= hiveContext.sql("select a.driverid,a.occurance,b.totmiles from geolocation_temp2 a,drivermileage_temp1 b where a.driverid=b.driverid")
-
The resulting data set will give us total miles and total non normal events for a particular driver. Register this filtered table as a temporary table so that subsequent SQL queries can be applied on it.
joined.registerTempTable("joined")
-
You can view the result by doing an action operation on RDD.
joined.collect.foreach(println)
- **Compute Driver Risk Factor **
In this section we will associate a driver risk factor with every driver. Driver risk factor will be calculated by dividing total miles travelled by non normal event occurrences.
val risk_factor_spark=hiveContext.sql("select driverid, totmiles,occurance, totmiles/occurance riskfactor from joined")
-
The resulting data set will give us total miles and total non normal events and what is a risk for a particular driver. Register this filtered table as a temporary table so that subsequent SQL queries can be applied on it.
risk_factor_spark.registerTempTable("risk_factor_spark")
-
View the results
risk_factor_spark.collect.foreach(println)
Step 4.5: Load and Save Data into Hive as ORC
In this section we will try to store data in orc format in Hive from Spark.ORC is a self-describing type-aware columnar file format designed for Hadoop workloads. It is optimized for large streaming reads and with integrated support for finding required rows fast. Storing data in a columnar format lets the reader read, decompress, and process only the values required for the current query. Because ORC files are type aware, the writer chooses the most appropriate encoding for the type and builds an internal index as the file is persisted.
Predicate pushdown uses those indexes to determine which stripes in a file need to be read for a particular query and the row indexes can narrow the search to a particular set of 10,000 rows. ORC supports the complete set of types in Hive, including the complex types: structs, lists, maps, and unions.
- Create an ORC table
Create a table and store it as ORC. Specifying as orc at the end of the SQL statement below ensures that the Hive table is stored in the ORC format.
hiveContext.sql("create table finalresults( driverid String, occurance bigint,totmiles bigint,riskfactor double) stored as orc").toDF()
- ** Load data into ORC table**
Before we load the data into hive table that we created above, we will have to convert our data file into orc format too.
risk_factor_spark.saveAsOrcFile("risk_factor_spark")
Load the data into Hive table using load data command.
hiveContext.sql("load data inpath 'risk_factor_spark' into table finalresults")
-
** **Execute a select query to verify your table has been successfully stored.You can go to Ambari Hive user view to check whether the Hive table you created has the data populated in it.
hiveContext.sql("select * from finalresults")
Lab 5a: Reporting With Microsoft Excel
Introduction:
This step is optional as it requires you to have Excel and Power View, however feel free to connect from any reporting tool to do a similar exercise. In this section, we will use Microsoft Excel Professional Plus 2013 to access the refined data. We will be using the ODBC connection.
Prerequisites:
The tutorial is a part of series of hands on tutorial to get you started on HDP using Hortonworks sandbox. Please ensure you complete the prerequisites before proceeding with this tutorial.
- Lab 0 (Hortonworks sandbox set up)
- Lab 1: Loading sensor data into HDFS
- Lab 2: Data Manipulation with Apache Hive
- Lab 3: Use Pig to compute Driver Risk Factor/ Lab4: Use Spark to compute Driver Risk Factor
- Please configure ODBC drivers on your system with the help of following tutorial:
- Allow yourself around half an hour to complete this tutorial.
Outline:
- Step 5.b.1: Access Data in Microsoft Excel
- Step 5.b.2: Visualize data with Excel
- Suggested readings
Step 5.b.1: Access the Refined Data with Microsoft Excel
The Hive ODBC driver can be found at the Hortonworks Add-on page. For Windows ODBC driver setupfollow these instructions and these are the Mac ODBC driver setup instructions. Open the ODBC connection manager and open the connection you setup up. It should look like this. Click on the test button and it should report success. If the test fails you will need to troubleshoot the connection before you can go on.
For Mac Users:-
- Open a new blank workbook. Select Data tab at the top then select “Get External Data” and then select “New Database query”.
- Chose Sample Hortonworks Hive DSN and hit Ok.
- Next prompt shall ask you to enter login and password. Type “sandbox” in login and the leave the password empty. Hit Ok.
- Choose the data table avg_mileage table from the table chart on the right and hit Add Table. avg_mileage should now appear under Query View tab on the left. Choose avg_mileage.* in the Field column. You are set to import data from your Hive table. As a final step, Hit Test!
- Click on Return Data to import your table into an excel. Submit OK to populate your excel.
- Once the data is placed you will see the avg_mileage table imported into your spreadsheet.
For Windows users:-
- Open a new blank workbook. Select Data tab at the top then select “Get External Data” and then select “From Other Data Sources”. Then at the bottom select “From Microsoft Query”. Choose your data source and ours is called Hadoop and you will then see the Query Wizard. We will import the avg_mileage table.
- Accept the defaults for everything and click through till you hit the Finish button. After you click on Finish, Excel will send the data request over to Hadoop. It will take awhile for this to happen. When the data is returned it will ask you to place the data in the workbook. We want to be in cell $A$1 like this.
- The data is placed you will see the avg_mileage table imported into your spreadsheet.
Step 5.b.2: Visualize Data with Microsoft Excel
- So now we are going to insert a Power View report. We do this by selecting the “Insert” tab at the top and select “Power View Reports” button in the middle. This will create a new tab in your workbook with the data inserted in the Power View page.”
- Select the design tab at the top and then select a column chart and use the stacked column version in the drop down menu. This will give you a bar chart. Grab the lower right of the chart and stretch it out to the full pane. Close the filter tab and the chart will expand and look like this.
-
So to finish off the tutorial I am going to create a map of the events reported in the geolocation table. I will show you how you can build up the queries and create a map of the data on an ad hoc basis.
-
For a map we need location information and a data point. Looking at the geolocation table I will simply plot the location of each of the events. I will need the driverid, city and state columns from this table. We know that the select statement will let me extract these columns. So to start off I can just create the select query in the Query Editor.
-
— Query subset of geolocation columns
select driverid, city, state from geolocation;
- After I execute the query I see what results are returned. In a more complex query you can easily make changes to the query at this point till you get the right results. So the results I get back look like this.
- Since my results look fine I now need to capture the result in a table. So I will use the select statement as part of my CTAS (create table select as) pattern. I will call the table events and the query now looks like this.
— Create table avg_mileage from existing trucks_mileage data
CREATE TABLE events
STORED AS ORC
AS
SELECT driverid, city, state FROM geolocation;
- I can execute the query and the table events gets created. As we saw earlier I can go to Excel and import the table into a blank worksheet. The imported data will look like this.
- Now I can insert the PowerView tab in the Excel workbook. To get a map I just select the Design tab at the top and select the Map button in the menu bar.
- Make sure you have a network connection because Power View using Bing to do the geocoding which translates the city and state columns into map coordinates. If we just want to see where events took place we can uncheck the driverid. The finished map looks like this.
We’ve shown how the Hortonworks Data Platform (HDP) can store and analyze geolocation data. In addition I have shown you a few techniques on building your own queries. You can easily plot risk factor and miles per gallon as bar charts. I showed you the basics of creating maps. A good next step is to only plot certain types of events. Using the pattern I gave you it is pretty straight forward to extract the data and visualize it in Excel.
Lab 5b: Reporting With Apache Zeppelin Data Visualization using Apache Zeppelin Introduction: Apache Zeppelin provides a powerful web-based notebook platform for data analysis and discovery. Behind the scenes it supports Spark distributed contexts as well as other language bindings on top of Spark.
In this tutorial we will be using Apache Zeppelin to run SQL queries on our geolocation, trucks, and risk factor data that we’ve collected earlier and visualize the result through graphs and charts.
NOTE: We can also run queries via various interpreters for the following (but not limited to) spark, hawq and postgresql.
Prerequisites:
– Working Zeppelin installation. – Make sure Hive services are up and running. (Check in Ambari dashboard) – Make sure Spark services are up and running.(Check in Ambari dashboard) – Lab 3 or 4.
Step 1: Creating a Zeppelin Notebook Click on Ambari User view icon and navigate to Zeppelin notebook.
– NOTE: You can also navigate to http://sandbox.hortonworks.com:9995 directly to open the Zeppelin interface. – You can also change the port that the zeppelin server runs on by Navigating to the Ambari service > configs > Advanced zeppelin config and then changing the zeppelin.server.port. – You may also need to modify the forwarded ports on the virtual machine as well if you do this.
Click on create notebook and a new workbook shall get started. Name the note Driver Risk Factor
After creating the new note, use the dropdown menu from the Notebook button to open the new Notebook
Step 2: Running a Hive query In the previous Spark and Pig tutorials you already created a table finalresults which gives the risk factor associated with every driver. We will use the data we generated in this table to visualize which drivers have the highest risk factor. Copy and paste the code below into your Zeppelin note.
%hive SELECT * FROM finalresults
Click the arrow on the right hand side to run the query in the Zeppelin notebook. ( or press Shift+Enter )
Initially, the query will produce the data in tabular format as shown.
Step 3: Making charts using Zeppelin
– Iterate through each of the tabs that appear underneath the query. Each one will display a different type of chart depending on the data that is returned in the query. – After clicking on a chart, we can view extra advanced settings to tailor the view of the data we want
Click settings to open the advanced chart features.
To make the same chart as the one above, drag the table relations into the boxes as shown in the image below.
You should now see an image like the one below.
If you hover on the peaks, each will give the driverid and riskfactor.
Try experimenting with the different types of charts as well as dragging and dropping the different table fields to see what kind of results you can obtain.
Let’s try a different query to find which cities and states contain the drivers with the highest risk factors.
%hive SELECT a.driverid, a.riskfactor, b.city, b.state FROM finalresults a, geolocation b where a.driverid=b.driverid
Run the query above using the keyboard shortcut Shift+Enter. You should eventually end up with the results in a table below.
After changing a few of the settings we can figure out which of the cities has the high risk factors. Try changing the chart settings by clicking the scatterplot icon. Then make sure that they keys a.driverid is within the xAxis field, a.riskfactor is in the yAxis field, and b.city is in the group field. The chart should look similar to the following.
The graph shows that driver id number A39 has a high risk factor of 652417 and drives in Santa Maria.
Outro Now that we know how to use Apache Zeppelin to obtain and visualize our data, we can use the skills we’ve learned from our Hive, Pig, and Spark labs and apply them to new kinds of data to try to make better sense and meaning from the numbers!
Suggested Reading
– Data Science With Spark on Zeppelin