Distributed processing
•
0 likes•1,677 views
This document discusses using distributed processing to increase computational power. It introduces the concepts of using a distributed file system like HDFS to store large files across multiple machines. MapReduce is presented as a way to process large amounts of data in parallel by splitting it into chunks, mapping functions to each chunk, and then recombining the results. Hadoop is introduced as an open-source MapReduce framework that uses HDFS for storage and allows processing of data across a cluster of machines. Examples are given of writing Map and Reduce functions and running jobs on Hadoop to demonstrate distributed processing of large datasets.
Report
Share
Related slideshows
Distributed processing
- 1. Increase computational power with distributed processing Neil Stein 03 Nov 2012
- 3. A Discussion Example…….. Getting the data, and ordering it as needed….. Familiar with grep and sort? — “grep” extracts all the matching lines — “sort” sorts all the lines grep “some_record_parameters” hl7_transfer.data-file | sort [2012/02/25/ 9:15] records sent to healthcare-1 [2012/02/28/ 6:15] records sent to healthcare-2 [2012/03/12/ 10:30] records sent to healthcare-3
- 4. A Discussion Example…….. — As the amount of data increases, process requires more and more resources — What if hl7_transfor.data-file is 500GB or bigger? — What if there are hundreds or thousands of data files? — What if there are multiple types of data files? grep “provider 1” hl7_transfor.data-file | sort — Ignoring the process for a moment, how do we write all the data to disk in the first place? Need to rethink the process
- 6. Distributed File-System – “the cloud” — Files can be stored across many machines — Files can be replicated across many machines — Files can be in a hyrbid-cloud model — Share the file-system transparently — You simply see the usual file structure — Opportunity to leverage private and public cloud environments
- 8. Map-Reduce – the cloud — A way of processing large amounts of data across many machines — Must be able to split-up the data in chunks for processing, (Map) — Recombined after processing (Reduce) — Requires a constant flow of data from one simple state to another — Allows for a simple way of breaking down a large task into smaller manageable tasks — Increase the available computational power
- 9. A look at Hadoop
- 10. What is Hadoop — A Map-Reduce framework — Designed to run applications on clusters of local and remote systems — HDFS — The file system of Hadoop (Hadoop Distributed File System) — Designed to access clusters of local and remote systems
- 11. Putting the pieces together….
- 12. First, we need some code…… Map Reduce
- 13. Map Hadoop streams information on STDIN Separate value with a newline (for Hadoop)
- 14. Reduce Hadoop streams back to us on STDIN Output the aggregated records
- 15. Sanity Checking Command Results This should work with small data-sets
- 16. Push file to “the distributed file system” Put file on the DFS Check that the file is in the cloud
- 17. Running in “the distributed environment” Call the Hadoop streaming command Pass the appropriate parameters
- 18. Running in “the distributed environment”
- 19. Running in “the distributed environment”
- 20. Running in “the distributed environment”
- 21. Running in “the distributed environment”
- 22. Checking Status — Cluster Summary — Running Jobs — Completed Jobs — Failed Jobs — Job Statistics — Detailed Job Logs
- 23. Checking Distributed Cluster Health — List Data-Nodes — Dead Nodes — Node Heart-beat information — Failed Jobs — Job Statistics — Detailed Job Logs
- 24. Conclusion — A different paradigm for solving large-scale problems — Designed to solve specific problems that can be defined in a focused map-reduce manner