Post on 16-Jan-2016
Apache Hadoop MapReduce
What is it ? Why use it ? How does it work Some examples Big users
MapReduce – What is it ?
Processing engine of Hadoop Developers create Map and Reduce jobs Used for big data batch processing Parallel processing of huge data volumes Fault tolerant Scalable
MapReduce – Why use it ?
Your data in Terabyte / Petabyte range You have huge I/O Hadoop framework takes care of
Job and task managementFailuresStorageReplication You just write Map and Reduce jobs
MapReduce – How does it work ?
Take word counting as an example, something that Google does all of the time.
MapReduce – How does it work ?
Input data split into shards Split data mapped to key,value pairs i.e. Bear,1 Mapped data shuffled/sorted by key i.e. Bear Sorted data reduced i.e. Bear, 2 Final data stored on HDFS There might be extra map layer before shuffle JobTracker controls all tasks in job TaskTracker controls map and reduce
MapReduce - Some examples
A visual example with colours to show you the cycleSplit -> Map -> Shuffle -> Reduce
MapReduce - Some examples
A visual example of MapReduce with job and task trackers added to individual map and reduce jobs.
Hadoop MapReduce – Big users
UsersFacebook Yahoo Amazon Ebay