Hi!
I have a question about Selecting Type of Database for a system which may encounter to a BigData in future!
Here is my questions:
Thanks.
Even though structured databases have become big enough and qualify as big data, true big data is mostly unstructured is the opinion of some of the Big Data Ninjas - meaning, if the data is structured, it not truly Big Data! (let's skip this aspect for the time being!)
The question of SQL or NoSQL always comes up in this discussion and this is subjective to the kind of data that is being handled. The issue is if you have a constantly changing data and the database requires quick scalability, NoSQL is ideal.
The big data is unstructured NoSQL, and the data warehouse queries this database and creates a structured data for storage in a static place. This serves as our point of analysis. The queries can be run as often as necessary.
The data required for analysis can be stored in SQL. So NoSQL and SQL can and must be used in tandem for best benefits. NoSQL the benefit is quick scalability ... SQL is structured and standardized and can be scaled (may be not as fast as NoSQL but depends). So currently most industry experts prefer to work with both as the need requires.
Regards,
Dr. Akilesh. R
India
Even though structured databases have become big enough and qualify as big data, true big data is mostly unstructured is the opinion of some of the Big Data Ninjas - meaning, if the data is structured, it not truly Big Data! (let's skip this aspect for the time being!)
The question of SQL or NoSQL always comes up in this discussion and this is subjective to the kind of data that is being handled. The issue is if you have a constantly changing data and the database requires quick scalability, NoSQL is ideal.
The big data is unstructured NoSQL, and the data warehouse queries this database and creates a structured data for storage in a static place. This serves as our point of analysis. The queries can be run as often as necessary.
The data required for analysis can be stored in SQL. So NoSQL and SQL can and must be used in tandem for best benefits. NoSQL the benefit is quick scalability ... SQL is structured and standardized and can be scaled (may be not as fast as NoSQL but depends). So currently most industry experts prefer to work with both as the need requires.
Regards,
Dr. Akilesh. R
India
There is third possibility - NewSQL databases.
NewSQL is a category of SQL database products that address the performance and scalability issues posed by traditional online transaction processing relational database management systems.
NewSQL databases aim to achieve the scalability of NoSQL while still providing the relational data model, full SQL query support, ACID transactions ensured by traditional relational databases.
In Big Data and databases the idea is the data is dynamic and to intense to persist, so rather the concept is to persist the queries, since the questions we seek from Big Data are relatively static in comparison to the Big Data. Essentially these queries are long running, continuously returning results from Big Data. This result may be persisted as it is usually a very small fraction of the Big Data.
SQL is language used to Query relational database systems also flavor SQL is used even Non Rellational databases such as HIVE.
If your Question is to use Relational or Non Relational databases, Traditional rellational databases is meant for ACID Properties, Non relational databases is compromised on atomicity of data. At this movement Non relational databases like HADOOP, Cassandra .... will not be better choice if your data present in multiple related tables because of atomocity level you may want .
Coming to your question of Facebook Likes as user base in billions it always better to store in single table, however Non relational databases very quick in scanning table with millions and millions of entries as they do indexed search.
There are no "SQL" and "NoSQL" but several data models implemented in different DBMSs that may be appropriated to OLAP (online analytical processing). The "classic" architecture is relational datawarehouse as main storage with multiples datamarts (usually, relational too) and multidimensional cubes as end-user databases.
This figure and article by Prashanth Jayaram may help:
https://dzone.com/articles/sql-vs-nosql
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Analysis
- Data Analysis
- Dataset
- Big Data