Big Data

The purpose of this project is to explore hadoop ecosystem (ex: MapReduce, Hive) and solve specific problems.

Scope

• Graphs
  • 'Dynamic Connectivity' and 'Connected Components' problems
  • Implementing 'Weighted Quick Union with Path Compression' algorithm (for 'Dynamic Connectivity' problem) on Hadoop (as HIVE UDAF)
    • module udaf, hql/custom_udaf.sql
  • PEGASUS library - http://www.cs.cmu.edu/~pegasus/
• Hive
  • User Defined Functions
    • Handling rows with a header (udf) - module udf, hql/custom_udf.sql
  • DDL - Create, show, drop, alter
  • Ingestion - Load, CTAS, Insert into Select
  • Serde - Regular Expression ex: Apache log
  • Columnar - ORC and Vectorized execution.
  • Select - count(*), distinct, nested select, filters
  • Functions - UDF, UDTF, Custom UDF, Custom UDAF
  • Aggregation functions - count, sum
  • Windowing and analytic functions - row_number(), rank(), over with partition by and order by,
  • Export - overwrite local
  • Unclassified Features - Lateral Views, Parameterizing scripts
  • Windowing and Analytics Functions
    • rank within a group
    • top 3 within a group
• Leverage Tez execution engine library (use "-e tez" option while running server and client commands)
• Data warehouse patterns (and possibly anti-patterns)
• MapReduce programs

How-to use/run this project

• Prerequisites:

  • Install JDK 6 or higher verssion
  • Download Hadoop tarball
    • hadoop-2.7.1.tar.gz from http://archive.apache.org/dist/hadoop/core/hadoop-2.7.1/
    • and extract contents to a folder say /home/sandeep/tools/hadoop/2.7.1
  • Download Hive tarball
    • apache-hive-1.2.1-bin.tar.gz from https://archive.apache.org/dist/hive/hive-1.2.1/
    • and extract contents to a folder say /home/sandeep/tools/hive/1.2.1
  • Download PEGASUS
    • from http://www.cs.cmu.edu/~pegasus/PEGASUSH-2.0.tar.gz
    • and extract contents to a folder say /home/sandeep/tools/PEGASUS
  • Install protobuf - http://www.confusedcoders.com/random/how-to-install-protocol-buffer-2-5-0-on-ubuntu-13-04
  • Clone Tez - https://github.com/apache/tez
    • and install it - http://tez.apache.org/install.htm

• Command to start hadoop

    sudo ./run.sh -m server -c start -e mr
  

• Command to run wordcount MapReduce program and get hdfs REPL

    sudo ./run.sh -m client -c mapreduce -e mr
  

• Command to run all hive scripts including connected components programs

    sudo ./run.sh -m client -c start -e mr
  

• Command to run GIMV-CC, PEGASUS CC and get PEGASUS REPL

    sudo ./run.sh -m client -c pegasus -e mr
  

• Command to stop hadoop

    sudo ./run.sh -m server -c stop -e mr
  

Technology Stack

• OS Version : Ubuntu wily(15.10)
• Hadoop Version : 2.7.1
• Hive Version : 1.2.1
• HBase Version : 0.98.18 (changed from 1.2.0 as hive support)
• Tez Version : 0.8.2-SNAPSHOT
• PEGASUS : 2.0

Algorithms & Papers

• Weighted Quick Union With Path Compression : An attempt has been made to parallelize it in the form of a hive UDAF.

    * Union Find - http://algs4.cs.princeton.edu/15uf/
  

• GIM-V based algorithms - http://www.cs.cmu.edu/~pegasus/

    * Paper - http://www.cs.cmu.edu/~ukang/papers/PegasusICDM2009.pdf
    * Getting Started - http://www.cs.cmu.edu/~pegasus/getting%20started.htm
  

• GIMV HBASE

    * Paper - Fast Graph Mining with HBase - http://datalab.snu.ac.kr/~ukang/papers/LeeSKunicorn.pdf
  

• 'MapReduce' - programming model

    *  Paper and presentation - http://research.google.com/archive/mapreduce.html
  

DataSets

• India power generation stats for FEB-2013 from http://data.gov.in/
• Apache logs from http://tingri.me

TODO

• Kylin
• Giraph
• Drill
• More SQL patterns