NoSql (It’s “Not only SQL” not “No to Sql”)

This is my first post on NoSql database technologies. There have been drastic changes in database technologies over the few years. Increase in user’s requests, high availability of applications, real time performance forced to think on different database technologies. We have traditional RDBMS, memory and NoSql databases available in market to suffice particular business needs. Here I’ll illustrate some of key aspects of NoSql databases like what is NoSql, why we need it, advantages and disadvantages of NoSql.

What is NoSql Movement?

It’s a different way of thinking in database technologies. It is unlike relational database management system where we have tables, procedures, functions, normalization concepts. NoSql databases are not built primarily on tables and don’t use sql for manipulation or querying database.

NoSql databases have specific purpose to achieve, that means NoSql database might not support all the features like in relational databases.

NoSql databases are based on CAP Theorem.

  • Consistency: Most of the applications or services attempt to provide strong consistent data. Interactions with applications/services are expected to behave in transactional manner ie. Operation should be atomic (succeed or failure entirely), uncommitted transactions should be isolated from each other and transaction once committed should be permanent.
  • Availability: Load on services /applications are increasing and eventually services should be highly available to users. Every request should be succeed.
  • Partition tolerant: Your services should provide some amount of fault tolerance in case of crash, failure or heavy load. It is important that in case of these circumstances your services should still perform as expected. Partition tolerant is one of desirable property of service. Services can serve request from multiple nodes

Why NoSql?

Since NoSql databases are using for specific purpose. They are normally using for huge data where performance matters. Relational database systems are hard to scale out in case of write operation. We can load balance database servers by replicating on multiple servers, in this case read operation can be load balance but write operation needs consistency across multiple servers. Writes can be scaled only by partitioning the data. This affects reads as distributed joins are usually slow and hard to implement. We can support increase in no. of users or requests by scaling up relational databases which means we need more hardware support, licensing, increase in costs etc.

Relational databases are not good option on heavy load which are doing read and write operations simultaneously like Facebook, Google, Amazon, Twitter etc.

A NoSQL implementation, on the other hand, can scale out, i.e. distribute the database load across more servers.



Common characteristic in NoSql databases

· Aggregating (supported by column databases): Aggregation usage to calculate aggregated values like Count, Max, Avg, Min etc. Some of NoSql provides support for aggregation framework which have inbuilt aggregation of values. Approach in column databases is to store values in columns instead rows (de-normalized data). This kind of data mainly used in data analytics and business intelligence. Google’s BigTable and Apache’s Cassandra supports some feature of column databases.

· Relationships (support by graph databases): A graph database uses graph structures with nodes, edges and properties. Every element contains a direct pointer to adjacent element; in this case it doesn’t need to lookup indexes or scanning whole data. Graph databases are mostly use in relational or social data where elements are connected. Eg. Neo4j, BigData, OrientDB.



Source: wikipaedia


· Document based. Document databases are considered by many as the next logical step from simple key-/value-stores to slightly more complex and meaningful data structures as they at least allow encapsulating key-/value-pairs in documents. Eg. CouchDb, MongoDb.

Mapping of document based db vs relational db


Document Based Databases Relational databases
Collections Table
Document Row


· Key- Value Store: Values are stored as simply key-value pairs. Values only stored like blob object and doesn’t care about data content. Eg. Dynamo DB, LevelDB, RaptorDB.

· Databases Scale out: When the load increases on databases, database administrators were scaling up tradition databases by increasing hardware, buying bigger databases- instead of scale out i.e. distributing databases on multiple nodes /servers to balance load. Because of increase in transactions rates and availability requirements and availability of databases on cloud or virtual machine, scaling out is not economic pain in increasing cost anymore.

On the other hand, NoSql databases can scale out by distributing on multiple servers. NoSQL databases typically use clusters of cheap commodity servers to manage the exploding and transaction volumes.  The result is that the cost per gigabyte or transaction/second for NoSQL can be many times less than the cost for RDBMS, allowing you to store and process more data at a much lower price;

Now question here is why scaling out in RDBMS is hard to implement. Traditional databases support ACID properties that guarantee that database transactions are processed reliably. A transaction can have write operations for multiple records, so to keep consistency across multiple nodes is slow and complex process, because multiple servers would need to communicate back and forth to keep data integrity and synchronize transactions while preventing deadlock. On the other hand NoSql databases supports single record transaction and data is partitioned on multiple nodes to process transactions fast.

· Auto Sharding (Elasticity): NoSql databases support automatic data sharding (horizontal partitioning of data), where database breaks down into smaller chunks (called shard) and can be shared across distributed servers or cluster. This feature provides faster responses to transactions and data requests.


· Data Replication: Most of NoSql supports data-replication like relational databases to support same data-availability across distributed servers.


· No schema required (Flexible data model): Data can be inserted in a NoSQL DB without first defining a rigid database schema. The format of the data being inserted can be changed at any time, without application disruption. This provides greater application flexibility, which ultimately delivers significant business flexibility.


· Caching: Most of NoSql databases supports integrated caching to support low latency and high throughput. This behavior is contrast with traditional database management systems where it needs separate configuration or development to support.

Challenges of No-SQL

Till now we have seen significant advantages of NoSql over RDBMS, however there are many challenges to implement NoSql.

Maturity: Most of the NoSql databases are in open source or in pre-production stage. In this case it might be risk to adopt these databases on enterprise level. For small business or use case it might be better to consider. On the other hand RDBMS databases are matured, providing many features and having good documentations or resources.

Support: Most of RDBMS are not open source that means they come with commitment and assurance in case of failure. They are reliable products and properly tested. Most of NoSql databases are open source and not widely adopted by organizations. It is very hard to get effective support from open sources databases. Some of NoSql databases created by small startups for specific needs, not for global reach.

Tools: RDBMS databases have lot of tools to monitor databases, queries analyzing, optimizations, performance profiling, analytics and Business Intelligence. Objective of NoSql databases are to minimize use of admin tools which has not achieved fully yet, still there are certain things which need skills and tools to monitor database activities.

When to consider NoSql

Following are some of indicators you can consider while choosing NoSql database for your application:

· If your application needs high performance databases.

· Need less or zero administration of databases.

· You want flexible data model. Minor of major changes should not impact whole system.

· Application that needs less complex transactions.

· High availability.

· Not or less consideration on Business Intelligence and analytics.






Enhanced by Zemanta

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.