Apache Falcon – Basic Concepts

Apache Falcon use only three types of entities to describe all data management policies and pipelines. These entities are:

·         Cluster: Represents the “interfaces” to a Hadoop cluster

·         Feed: Defines a “dataset” File, Hive Table or Stream

·         Process: Consumes feeds, invokes processing logic & produces feeds

Using these three types of entities only we can manage replication, archival, retention of data and also handle job/process failures and late data arrival.

These Falcon entities–

• Are easy and simple to define using XML.
• Are modular - clusters, feeds & processes defined separately and then linked together and easy to re-use across multiple pipelines.
• Can be configured for replication, late data arrival, archival and retention.

Using Falcon a complicated data pipeline like below

can be simplified to a few Falcon entities (which are further converted to multiple Oozie workflows by Falcon engine itself)

In my next post I will explain how we can define a Falcon process and perquisites for that.

Reference - http://www.slideshare.net/Hadoop_Summit/driving-enterprise-data-governance-for-big-data-systems-through-apache-falcon

Introduction to Apache Falcon

While working for a project recently I got chance to work on Apache Falcon, which is a data governance engine that defines, schedules, and monitors data management policies. Falcon allows Hadoop administrators to centrally define their data pipelines, and then Falcon uses those definitions to auto-generate workflows in Apache Oozie.

I faced some issues in setting up my first Falcon job so in my next few posts I will write more about Falcon.

What does Apache Falcon do?

Apache Falcon simplifies complicated data management workflows into generalized entity definitions. Falcon makes it far easier to:

• Define data pipelines
• Monitor data pipelines in coordination with Ambari, and
• Trace pipelines for dependencies, tagging, audits and lineage.

Apache Oozie is Hadoop’s workflow scheduler, but mature Hadoop clusters can have hundreds to thousands of Oozie coordinator jobs. At that level of complexity, it becomes difficult to manage so many data set and process definitions.

This results in some common mistakes. Processes might use the wrong copies of data sets. Data sets and processes may be duplicated, and it becomes increasingly more difficult to track down where a particular data set originated.

Falcon addresses these data governance challenges with high-level and reusable “entities” that can be defined once and re-used many times. Data management policies are defined in Falcon entities and manifested as Oozie workflows.

Falcon Features -

1.      Centrally Manage Data Lifecycle

Ø  Centralized definition & management of pipelines for data ingest, process & export across different clusters.

2.      Business Continuity & Disaster Recovery

Ø  Out of the box policies for data replication, retention and archival.

Ø  Configuration and management of late data handling and exception handling.

Ø  Handles process failures and retries.

Ø  End to end monitoring of data pipelines.

3.      Address audit & compliance requirements

Ø  Visualize data pipeline lineage.

Ø  Track data pipeline audit logs.

Ø  Tag data with business metadata.

 

Reference - http://www.slideshare.net/Hadoop_Summit/driving-enterprise-data-governance-for-big-data-systems-through-apache-falcon

A Quick Introduction to Hive Java APIs

Hive provides some Java APIs like HCatClient and Driver which can be used to execute many Hive related functions.

HCatClient API can be used for many Hive related functions like -

1. Create/Drop Hive schemas/tables.
2. Update Hive table schema.
3. Add/Drop Partitions.
4. Update Table name.

Driver API can be used to -

1. Execute Hive query.
2. Get query plan.

The below program illustrates the use of HCatClient to create a database and create a table -


Now we can modify the above table schema using HCatClient -


Now we can use Driver API to drop table and database
 

Hive - Initialization Script

In many of Hive scripts there is use of some Hive UDFs and Hive configuration settings. Declaring these UDFs and Hive configuration variable across many scripts and/or at start of each Hive session is very cumbersome task and maintaining multiple scripts becomes more and more difficult with time.

To overcome these problems we can follow either below methods -

1. Create an environment file which executed on start of each Hive cli session.
2. Create an initialization script file which is initialized before start of Hive cli session or before execution of other Hive script.

Let us see each of the above methods in more details -
1. Create an environment file -
Hive executes an environment file by name .hiverc, present in home directory of user. All commands present in this file are executed before start of each Hive cli seesion, hive -e or hive -f options.
By default .hiverc file will not be present in user's home directory and user needs to create this.

2. Create an initialization script file
This option is more useful when we want to execute different set of initialization commands for different set of scripts.
To use this option create a text file with all initialization commands and use it in hive using hive -i <filename> option. This option can be used with Hive/beeline cli or -e or -f non-interactive modes.

An example initialization script looks like this (lets assume filename is hive_init.hql) -

SET hive.exec.dynamic.partition=true;
SET hive.exec.dynamic.partition.mode=nonstrict;

SET hive.vectorized.execution.enabled=false;
SET hive.vectorized.execution.reduce.enabled=false;

SET hive.enforce.bucketing=true;

SET hive.exec.parallel=true;

SET hive.auto.convert.join=false;
SET hive.enforce.bucketmapjoin=true;
SET hive.optimize.bucketmapjoin.sortedmerge=true;
SET hive.optimize.bucketmapjoin=true;

ADD JAR /hadoop/rishav/udfs-0.0.1-SNAPSHOT.jar;
create temporary function getEvent as 'com.rishav.bigdata.hive.udfs.GetEventUDF';
create temporary function get_last_day as 'com.rishav.bigdata.hive.udfs.LastDayUDF';

To use this in different modes use

Hive CLI - hive -i hive_init.hql
Hive non-interactive command mode - hive -i hive_init.hql -e "sql command;"
Hive non-interactive script mode - hive -i hive_init.hql -f <script_file>

Implementing Apriori Algorithm In Hadoop-HBase - Part 2 : Conversion to MapReduce Codes

Let us assume that the transaction data which we are getting is in csv format like this - tId,pId
where tId is transaction Id
and pId is product Id
a single transaction can have one or more product Ids spread across one or multiple csv records e.g.
101,701
101,702
101,703
102,701
103,705
103,707

I have implemented Apriori algorithm for 2-itemset using 3 MapReduce jobs. The jobs and their functions are described below -

1. PopulateTranBasketAndProdPairJob - The mapper class of this job reads transaction records from specified csv file and emits (tId, pId). This job's reducer class gets (tId, <pId1, pId2,..., pIdn>) as input, then it makes product pairs available for this tId and writes individual pId(s) and product-pair(s) to HBase table 'tranBasket' with tId as rowkey.

2. SupportCountJob - This job reads the 'tranBasket' table and calculates the support counts for all pId and product pair(s). The support counts of individual products are stored in 'pCount' table with pId as rowkey and the support counts for product pairs are stored in 'ppCount' table with product pair as rowkey. At the end of this job transaction count is also printed to screen which acts as input to next job.

3.CalcSupportConfidenceJob - This is the last job in this series and gives us support, confidence and lift values for different product pairs. This job takes transaction count from the previous job as input to calculate support values. In this job only mapper is there, which reads complete 'pCount' table in memory and then reads 'ppCount' table row by row and performs calculation of different Apriori measures like support, confidence and lift and writes the result to HBase table 'apprOut'.
For verifying the results we can check mapper sysout files which have the values in readable format.

The source code is available here.

Note - This is just a simple demo application and there is scope for improvements. Some modifications which I can think of now are -

1. Generally transaction ids are sequential numbers and if they are stored in HBase as such we will experience region hot spotting. Hence rowkey design has to be reworked.
2. HBase scanner caching value needs to be checked and optimised.
3. Currently pId and tId are stored as Text which can be changed to Long.

References -

• http://rakesh.agrawal-family.com/papers/vldb94apriori.pdf
• http://www.slideshare.net/deepti92pawar/the-comparative-study-of-apriori-and-fpgrowth-algorithm
• http://www3.cs.stonybrook.edu/~cse634/lecture_notes/07apriori.pdf