Navigating NoSQL in cloudy skies
•
0 likes•863 views
NoSQL is not a buzzword anymore. The array of non- relational technologies have found wide-scale adoption even in non-Internet scale focus areas. With the advent of the Cloud...the churn has increased even more yet there is no crystal clear guidance on adoption techniques and architectural choices surrounding the plethora of options available. This session initiates you into the whys & wherefores, architectural patterns, caveats and techniques that will augment your decision making process & boost your perception of architecting scalable, fault-tolerant & distributed solutions.
Report
Share
Navigating NoSQL in cloudy skies
- 2. Presented at: Chicago IT Architects Group Jan 15, 2013
- 3. shankar ramachandran works with: Microsoft Web Stack of Love Microsoft SQL Server also works with:
- 4. Skipping essential steps, just creates an illusion of speed & growth.
- 5. simple. 5
- 6. Agenda • What NoSQL is & What it is not • Why NoSQL – 2 specific reasons • Conceptual Fundamentals & Grounding • 3 techniques to classify & choose • Way ahead
- 7. What • Variety of non- relational database systems • Usually schema-less • Mostly open-source • Not anti-RDBMS • Not a replacement
- 8. No – relational tables – were harmed in the making of this presentation.
- 9. Why NoSQL?
- 10. Reason #1
- 11. Big Data
- 12. “Big Data”
- 15. 4 Vs of Big Data Volume Velocity • Terabytes and Petabytes • Time sensitive real-time data processing & decision making Variety Value • Of structured and • Inherent value always unstructured data
- 16. RDBMS can handle all that. Right?? • Scaling up has a limit. • Sharding - spread data across servers. • Denormalization - potentially duplicates data in the database, requiring updates to multiple tables when a . duplicated data item is changed • Distributed Caching - caching recently accessed data in memory and storing that data across any number of servers . or virtual machines. Think Memcached.
- 17. RDBMS tactics - Downside & Pitfalls • Re-sharding is disruptive. • Maintain schema on every server • Distributed Caching accelerates just the reads • You lose relational benefits anyway.
- 18. aggregate-oriented vs. aggregate-ignorant
- 19. Aggregate-orientation • Unit of data can have a more complex structure than a set of simple tuples. • Excellent fit to run on a cluster. • Atomic manipulation of single aggregate. • Application code takes precedence.
- 20. Reason #2
- 22. • Difference between relational model & in-memory data structures • Simple tuples • ORMs provide a bridge ; complicate query performance.
- 23. { product : "Tintin Statue", created : Date(’11-16-2010’), title : "Brass replica of Tintin", tags : [ "tintin", "herge", "snowy"], comments : [ { author : ‘Shankar', comment : 'I love it' }, { author : ‘Skeet', comment : 'me too!!' } ] }
- 24. Concepts
- 25. 3 properties of distributed databases • Consistency means that each client always has the same view of the data. • Availability - node always available for read and write. • Partition tolerance means that the system works well across physical network partitions.
- 26. consistency availability partition-tolerance only-2-out-of-3 CAP Theorem
- 27. consistency availability partition-tolerance This is incorrect
- 28. consistency availability partition-tolerance
- 29. horizontal-partitioning multiple-instances shared-nothing sharding
- 30. commodity-hardware distributed infinite-expansion horizontal-scalability
- 31. google-patented-framework map: chop data reduce: fold data MapReduce
- 32. low-latency order-of-reads delayed-gratification eventual-consistency
- 33. For the academically inclined: Proprietary DB high-performance Google App. Engine Google BigTable Amazon Dynamo Proprietary system high-availability AWS key-value
- 34. quick shout-out
- 35. Object oriented Faster and Declarative. Lack of interoperability and recovery standards. End-to-end development, database & deployment platform Embeddable and fast. Lack of querying capabilities.
- 36. XML Native XML database systems. Typically XQuery used as querying mechanism. Advantage or Disadvantage based on XML affinity. Sedna Tamino
- 37. Choice By Data Model
- 39. Graph Graph-data structure associative-datasets node/edges Small records with complex interconnections. GraphDB
- 41. Key/Value in-memory processing/caching hyper-efficient associative storage Voldemort
- 42. Wide-Column horizontally-partition fully distributed Dynamo + BigTable
- 43. Document-oriented schema-less collection-based-JSON-like dynamic-indexing
- 45. Choice By CAP CA AP CP RDBMS Riak MongoDB Dynamo Hbase Cassandra Redis CouchDB Hypertable Voldemort
- 46. C C++ C# Erlang Java Redis MongoDB RavenDB CouchDB Cassandra Hypertable GraphDB Couchbase Hadoop Kyoto Riak HBase Tycoon Scalaris neo4J Voldemort
- 47. Way Ahead
- 48. What are Microsoft & Oracle up to?
- 50. Oracle NoSQL
- 52. polyglot persistence … a highly possible future
- 53. We learnt that ... RDBMSs are here to stay. NoSQL is not creating a paradigm shift. NoSQL provides a set of non-relation data stores & technologies that have affinity for being processed in a clustered environment. Some of them NoSQL databases also offer a solution to Impedance Mismatch thus increasing application developer productivity. What Aggregate-Orientation in data modeling means. What the different types of database types are. And most importantly ... we now know that RDBMS systems need DBAs - Database Architects & Admins. NoSQL systems need DBAs too - Developers Beyond Awesome!
- 55. Computers are useless. They can only give you answers. Pablo Picasso Cubist painter (1881 - 1973) ?