Your SlideShare is downloading. ×
Big data
Big data
Big data
Big data
Big data
Big data
Big data
Big data
Big data
Big data
Big data
Big data
Big data
Big data
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×
Saving this for later? Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime – even offline.
Text the download link to your phone
Standard text messaging rates apply

Big data

1,735

Published on

a quick talk i gave at the meetup in boulder, colorado

a quick talk i gave at the meetup in boulder, colorado

Published in: Technology, Design
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
1,735
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
23
Comments
0
Likes
0
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide

Transcript

  • 1. Cassandra and Hadoop
    • Kevin Cawley, Engineer Linksmart
    • 2. Cassandra – been actively using for 2+ years
    • 3. Hadoop – 1 yr experience, sort of
  • 4. Problem For Today
    • Running a survey for discovering nosql preference
    • 5. Options: mongodb, redis, cassandra, couch, hbase, riak, voldermort, dynamodb
    • 6. We're gonna get billions of responses – RDBMS is going to fall over
    • 7. We need nosql... what the hell is that??
  • 8. Problem For Today, cont. Kevin, response=cassandra, kevin@foo.com Emma, response=redis, emma@foo.com Asher, response=cassandra, [email_address] … … … BILLIONS AND BILLIONS OF THESE!!!!
  • 9. Cassandra
    • Linear scalability, high availability & performant database
    • 10. Key Value store
    • 11. Ring architecture w/ replication 2^217 tokens
    Node 1 Node 2 Node 4 Node 3
  • 12. Cassandra
    • Keypace
    • 13. Column Families – std, dynamic (mo better)
    name preference 100 kevin cawley cassandra 101 asher cawley cassandra 102 emma cawley redis 202 201 redis ['joe','bob'] ['matthias'] cassandra ['kevin', 'asher'] ['tom'] mongodb ['holly'] ['dan'] assume User keys as utf8;
  • 14. Super columns
    • Not so super
      • Nice on paper, can be catastrophic in practice
      • 15. Fanning – not the cool refreshing kind
      • 16. Getting phased out
    202 201 redis {'joe' => 'joe@foo.com , 'bob' => 'bob@boo.com'} {'matthias' => 'matthias@foo.cm', 'tom' => 'tom@boo.com'}
  • 17. Secondary Indexes
    • Indexes on column values
    • 18. Replacement for not so super, super columns
      • Composite columns US:colorado:cassandra => kevin
    • Demo 1
  • 19. Counters
    • Yes! Yum. Counters good
    • 20. We built our own – now free
    • 21. Cassandra is eventually consistent makes this hard
    • 22. Be clever and you will win
    • 23. Demo 2
  • 24. Counters Counter cassandra 30333 redis 22098 mongodb 24567 couch 12340 ...
  • 25. Hadoop
    • Distributed processing of large data sets across clusters of computers – too good to be true?
  • 26. Map Reduce
    • Acronym soup
      • hadoop common, hdfs, map reduce
      • 27. hbase, pig, hive, zookeeper
    • Map Reduce is @ the heart
    • 28. Map - processes a key/value pair to generate a set of intermediate key/value pairs
    • 29. Reduce - function that merges all intermediate values associated with the same intermediate key
  • 30. Map Reduce – Our Example Kevin, response=cassandra, kevin@foo.com Emma, response=redis, emma@foo.com Asher, response=cassandra, [email_address] ...
    • Map:
      • cassandra kevin
      • 31. cassandra asher
      • 32. redis emma
    • Reduce:
    AND the winner is cassandra w/ 2 votes!!!
  • 34. Cassandra Hive
    • Hadoop/Brisk on Cassandra – no luck
    • 35. Hive
      • Data warehouse built on top of cassandraFS leveraging map reduce
      • 36. Query the data using a SQL-like language called HiveQL
    • Demo 3
  • 37. Summary
    • Cassandra
      • Awesome for storing massive amounts of data
      • 38. Dangerous if you don't know what you are doing
      • 39. Schemaless – ironically modelling is extermely imp.
      • 40. Ad-hoc questions are hard to answer fast
    • Hadoop/Brisk
      • Great for answering ad hoc questions reasonably fast
    • What you really want is Cassandra ↔ Hadoop ↔ RDBMS

×