Featured post

Best Machine Learning Book for Beginners

You need a mixof different technologies for Data Science projects. Instead of learning many skills, just learn a few. The four main steps of any project are extracting the data, model development, artificial intelligence, and presentation. Attending interviews with many skills is not so easy. So keep the skills short.
A person with many skills can't perform all the work. You had better learn a few skills like Python, MATLAB, Tableau, and RDBMS. So that you can get a job quickly in the data-science project.
Out of Data Science skills, Machine learning is a new concept. Why because you can learn Python, like any other language. Tableau also the same. Here is the area that needs your 60% effort is Machine learning.  Machine Learning best book to start.

Related Posts How to write multiple IF-conditions in Python Simplified

5 Top features of Columnar Databases (1 of 2 )

The traditional RDBMS - Since the days of punch cards and magnetic tapes, files have been physically contiguous bytes that are accessed from start (open file) to finish (end-of-file flag = TRUE).

Yes, the storage could be split up on a disk and the assorted data pages connected by pointer chains, but it is still the same model. Then the file is broken into records (more physically contiguous bytes), and records are broken into fields (still more physically contiguous bytes).

A file is processed in record by record (read/fetch next) or sequentially navigated in terms of a physical storage location (go to end of file, go back/forward n records, follow a pointer chain, etc.). There is no parallelism in this model. There is also an assumption of a physical ordering of the records within the file and an ordering of fields within the records.
A lot of time and resources have been spent sorting records to make this access practical; you did not do random access on a magnetic tape and you could not do it with a deck of punch cards.

When we got to RDBMS and SQL, this file system model was still the dominant mindset. Even Dr. Codd fell prey to it. He first had to have a PRIMARY KEY in all tables, which corresponded to the sort order of a sequential file.

 Later, he realized that a key is a key and there is no need to make one of them special in RDBMS. However, SQL had already incorporated the old terminology and the early SQL engines were built on existing file systems, so it stuck.

Also read: Part-2 | Part-1
  • The columnar model takes a fundamentally different approach. But it is one that works nicely with SQL and the relational model.
  • In RDBMS, a table is an unordered set of rows that have exactly the same kind of rows. A row is an unordered set of columns all of the same kind, each of which holds scalar values drawn from a known domain. You access the columns by name, not by a physical position in the storage, but you have the "SELECT*" and other shorthand conventions to save typing.
  • The logical model is as follows: a table is a set of rows with one and only one structure; a row is a set of columns; a column is a scalar value drawn from one and only one domain. Storage usually follows this pattern with conventional file systems, using files for tables, records for rows, and fields for columns. But that has nothing to do with physical storage.
  • In the columnar model, we take a table and store each column in its own structure. Rows and tables are reassembled from these rows. Looking at the usual picture of a table, you can see why they are called vertical storage as opposed to horizontal storage models


Popular posts from this blog

Hyperledger Fabric: 20 Real Interview Questions

Best Machine Learning Book for Beginners

Python Assigning Multiple Values at Once