Tuesday, August 19, 2014

NoSQL Databases

       NoSQL is next generation DBMS which is used to store and retrieve the data which does not have specific format or table like structure. NoSQL database is also called as “Not only SQL” to emphasize that it does not use SQL language but It uses SQL-like language. The original intension has been modern web scale databases. The movement began in early 2009 and growing rapidly. NoSQL databases are mostly addressing some of the points:-
·         Non-relational
·         Distributed
·         Open source
·         Horizontally scale

          As we Know there are different types of Database Management System and RDBMS is most commonly used database management system. If we go back in history, first flat file systems were created then in 1970's Codd came up with Relational theory and based on that relational databases were developed. The problem was that in flat file system there was no standard way of storing data and no standard way of communicating with data. Everybody has implementing their own protocol and that was creating lots of inefficiency. So relational database standardized the way we communicate with database. Then life moved on and everything was going well and suddenly, we entered into Big data scenario. Relational databases are unable to handle huge amount of data so the answer was “NoSQL”. So NoSQL was created because of limitations of relational databases.
     The data is structured in relational databases. We have to define table structure in advance, we have to tell the system in advance like table will have 10 rows and 10 columns and each column will take particular data type and this would be the maximum value one can enter and so on. This way we can't handle unstructured data where there is no fixed format or format is always changing. So NoSQL handles both structured and unstructured data.
-NoSQL focused to provide:
·         Scalability
·         Performance
·         High Availability
·         Simple API
·         Schema free
·         Easy replication support
·         Eventually consistent(not ACID)
      NoSQL can handle large amount of data.  As data is keep on growing it provides scalability and high availability in terms of hardware failure. Ability of NoSQL is to handle large amount of data with amazing performance with comparison they offers less functionality than RDBMS.

     NoSQL is category which can be commonly further divided into three categories given below with example:
·         Key Value Store : Memcached,Coherence, Redis
·         Tabular : BigTable,Hbase,Accumulo
·         Document oriented : Mango,CouchDB,Cloudant
If you search you will find more categories. Here is the list of some:
·         Column store/Column family
·         Document store
·         Key value store/Tuple sore
·         Graph database
·         Multimodel database
·         Object database
·         Grid and Cloud solution
·         XML database
·         Multidimensional database
·         Event sourcing
·         Network model
·         Other NoSQL related databases
·         Unsolved and uncategorized

      So what is missing from NoSQL databases if you compare them with relational databases? Well Joins are not there because of this joins relational databases are not so scalable. So in NoSQL databases this functionality is not implemented hence scalability and performance comes in. The support for complex transaction is not there for example you can’t not do insert three records then update two records and check something if it’s not then rollback everything. Again constraint support is not there. Transaction support and constraint support is not there on database level but transaction support and constraint support can be applied on application level.

 What are the situations or scenarios where one can use NoSQL database?
·         The ability to store and retrieve great quantity of data is important.
·         Storing relationship between elements is not important.
·         Dealing with growing list of elements, example: Twitter post, Internet server logs, Blogs.
·         The data is not structured or the structure is changing with the time.
·          Prototype or fast application need to be developed
·         Constraints and validation logic is not required to be implemented in database.
What are the situations or scenarios where not to use NoSQL database?
·         Complex transactions need to be handled
·         Joins must be handled by databases
·         Validations must be handled by databases