RDBMS is relational database management system. In the early 1970s, Codd invented the relational database management system. It was an advancement to spur DBMS movement they had at that time by implementing the cardinality and normalization to the database. Codd conceptualized and created 12 rules for a traditional RDBMS. Though the rules are laid out, it made the database more flexible and integrated with these principles.
(a) RDBMS deems all database management systems to accede these rules. A majority of the commercial database products fulfill the majority of the rules;
(b) the database system is defined with exemplary structured data. The entire database resides as values organized in immaculate and structured columns and rows;
(c) the data in a table must be queryable. The database table should have a primary index. The query should be against the values in the fields with a combination of the column fields that can be specified in the query. A table can have multiple secondary indices. The table can be queried against those indices as well with composite combination of other column fields;
(d) the database columns should have particular schema definition on the table to bolster any null values to be written into the database; (e) the SQL query language built surrounding such RDBMS systems should be able to aid retrieving the structure of the table as well;
(f) the SQL query language should buoy all the database operations (read, write, and update) with security concepts. The database also should have integrity for attaining relational database approach;
(g) the system should be able to generate views combining complex database tables. The views can be updated similar to the updates to the database tables;
(h) the transactional commits involving update, insert, and delete operations on the database should be able to perform mass transactional updates;
(i) the updates should be isolated;
(j) the database updates should be strict to the database level, but not to the application stack or kernel level of the database;
(k) setting up foreign keys between multiple database tables will create logical relationships for building validation rules in the database; (l) the updates should be distributed across the system with seamless experience (W3Resource, 2013).
RDBMS majorly supports historic schema-based data structured types such as numbers, characters, dates, alphanumeric combinations, binary large objects, and decimal points (Oracle, 2015). RDBMS has static schema definitions to support well-structured database formats and eliminates scope for ambiguity. The design is a crucial factor to RDBMS to build sustainable enterprise resource planning database systems for storing the corporate data. With the diffusion of Internet, there is a massive amount of data created in a corporation from multiple heterogeneous systems. RDBMS cannot directly integrate these systems. It requires conversions to well-defined text files to transform the data into structured transactional database updates. RDBMS was not built to handle large-scale data sets. The infrastructure and storage hardware costs are high for RDBMS databases. Traditional RDBMS databases cannot handle the unstructured data, velocity, and volume of the data generated outside the corporate firewall. They also lack parallel computing as there are limited number of cores on the processor of the server (Labs, 2014).
The unstructured big data is schema-less and comes from disparate sources of data types that cannot be stored on the RDBMS database system due to costs per byte equation. Traditional RDBMS have data categories for storing the data content. However, big data needs a staging pad to save all the data oceans in liquid form unaltered, unclassified, uncategorized, unprocessed, and ungoverned in its original raw form to retain the data formats for processing. These staging pads are dubbed as data lakes (Aspili, 2015). The heterogeneous data types in semi-structured and unstructured formats arise from multimedia content such as social or business text messages, video, audio, emails, website pages, and a bevy of corporate documents. RDBMS cannot fit these data types in pre-architected database formats. Big data is dynamic and does not comply Codd’s 12 rules. Big data is raw, dissimilar, unconnected with other database tables (Philip, 2014).
Hadoop framework was invented as a distributed freeware that runs on clusters of vendible distributed servers that considers cost per byte as the critical factor. Hadoop framework is chiefly a Java featured file system that can store and process gargantuan volumes of big data. The files can store 100s of petabytes of data efficiently spanning over the cluster of 1000s of servers. Hadoop framework can connect to data workloads of more than a billion number of files. The data on Hadoop can reside on cloud-based data centers or traditional on-premise data centers. Hadoop file system is a container where the data can be filled in raw format to process it on commodity servers with economies of scale for mining and refining the data into valuable insights (Awadallah & Graham, 2012).
Corporations can frequently add tons of petabytes of corporate data to the clusters of commodity hardware on a daily basis. Hadoop is a new breed of transformation technology framework when contrasted with traditional data warehouse architecture that processes the data with conversion, extraction, and loading processes. Hadoop is a funnel for transformation layer after connecting the dots with all the semi-structured and unstructured data with a logical mapping integration. Hadoop is capable of distributing the monumental data across the clusters of voluminous servers. The advantages of Hadoop is the flexibility of writing the code in a bevy of programming languages for extensive parallelization process on the same cluster connecting a series of blocks and files and process them concurrently (Awadallah & Graham, 2012).
Organization can store big data on the commodity servers of Hadoop similar to storing the traditionally archived data on hard tapes. It is easier for the companies to spin up and spin down a number of commodity servers on physical data centers or cloud-based segmented and partitioned commodity servers. Massive parallelism is the main advantage of Hadoop framework to distribute the data workloads on multiple commodity servers for processing millions of files on each system and glean the data such as text and add the extracted data to the new file. Publishing companies such as New York Times implemented Apache Hadoop framework on Amazon Elastic Cloud computing to glean the data out of some eleven million portable document files under a day with a query on specific text with massive parallel processing mechanism (Snowpoch, 2013).
Corporations need a combination of economies of flash and economies of big data with parallel computing abilities. This combination can be achieved by implementing SAP HANA with Apache Hadoop framework. Companies can store a treasure trove of their untamed big data on Apache Hadoop and leverage SAP HANA for wrangling the data with massive parallelization options. Integration of SAP HANA with Apache Hadoop results in resolving several business conundrums. Corporations can selectively choose the untamed, unstructured data and distribute it over the commodity servers of Apache Hadoop and leverage massive parallel processing from Hadoop to run analytics by gleaning the refined data into SAP HANA calculation engine. Few advantages of parallel processing include SAP HANA providing connection to Apache Hadoop through smart data access connector without having to store physically, convert or extract the data into SAP HANA. Once the connections are established, SAP HANA can connect to any Hadoop remote files and leverage parallelization technique and run SQL queries from SAP HANA Sybase database still keeping the data on Hadoop with parallel computing and return the result sets back into SAP HANA for calculation and analytics purposes (SAP, 2015).
Aspili, A. (2015). How to Ensure Data Lakes Success. Retrieved October 26, 2015, from http://www.smartdatacollective.com/alleliaspili/353161/how-ensure-data-lakes-success
Awadallah, A., & Graham, D. (2012). Hadoop and the Data Warehouse:When to Use Which. Retrieved October 26, 2015, from http://assets.teradata.com/resourceCenter/downloads/WhitePapers/EB-6448.pdf?processed=1
Labs, F. (2014). Why RDBMS Fails To Support Big Data? Retrieved October 26, 2015, from http://blog.flux7.com/blogs/flux7-labs/why-rdbms-fails-to-support-big-data
Oracle (2015). RDBMS Event Generator. Retrieved October 26, 2015, from https://docs.oracle.com/cd/E13214_01/wli/docs81/rdbmseg/datatypes.html
Philip, N. (2014). Hadoop vs. Traditional Database: Which Better Serves Your Big Data Business Needs? Retrieved October 26, 2015, from https://www.qubole.com/blog/big-data/hadoop-vs-traditional/
SAP (2015). Improve business performance in real time with the industry’s leading Big Data platform. Retrieved October 27, 2015, from http://www.sap.com/solution/big-data/software/platform.html
Snowpoch (2013). Using Hadoop for Parallel Processing rather than Big Data. Retrieved October 27, 2015, from http://stackoverflow.com/questions/15743943/using-hadoop-for-parallel-processing-rather-than-big-data
W3Resource (2013). Codd’s 12 rules. Retrieved October 26, 2015, from http://www.w3resource.com/sql/sql-basic/codd-12-rule-relation.php