Big Data simply refers to the collection of data sets which are so huge that it is impossible to process them through the traditional data management tools and data processing applications.
Along with the traditional data big data integrates many new types of data and data management. While Big Data has existed for quite a while, but is only now due to lower costs of computation and massive explosion of data which has encouraged the adoption of the big data. Apache hadoop and NoSQL database are two storage and foundational techniques which has been founded by the Big Data.
Spark is essentially a framework used to perform general analytics on Hadoop. Spark is responsible for memory computations for increase speed and access the Hadoop Data Store.
Spark is a processing engine built by the Apache team which offers speed, ease of use and analytics as its core functionality. For data which requires low latency, Spark is perfectly suitable. Spark is 100 times faster than the Map reduce and supports python, java and Scala APIs for the ease of development.
Sparks biggest point is that includes streaming, SQL and complex analytics together in the same application and is capable of providing the wide range of data processing scenarios. It is capable of running on Hadoop, Mesos, and standalone or even in the cloud and has the ability to access the diverse data sources such as HDFS, Cassandra, HBase or S3.
Now, the three S of spark framework which are the main reasons to choose this framework for Big Data Analytics.
Spark has a rich set of APIs which increases its capabilities and makes it more accessible for interacting quickly and easily with the data. All the APIs are documented in detail in a structured and straight format which helps data scientists and application developers to put spark to work quickly and easily.
Speed in another intrinsic characteristic of Spark which gives excellent result both in memory and on disk. Spark is capable of processing 100 terabytes of data stored on solid state drives in merely 23 minutes. The previous winner was Hadoop which took 72 minutes for processing the data. While supporting the interactive queries on data the performance of Spark can even be greater and this can be 100 times faster than the Hadoop and Map reduce.
Spark supports variety of programming languages such as Java, Python, R and Scala. Though, Spark is closely associated with Hadoop’s storage system HDFS, it also provides tight integration with the leading storage solutions due to its native support in the Hadoop ecosystem and beyond. The community of Apache Spark is growing by the day and is international along with being very active. Some commercial vendors like IBM, Databricks, main Hadoop vendors and many others are providing comprehensive support for the Spark based solutions.