1
00:00:00,580 --> 00:00:01,600
- [Instructor] Next, let's talk about

2
00:00:01,600 --> 00:00:03,619
NoSQL columnar databases,

3
00:00:03,619 --> 00:00:07,770
which are also sometimes called
column-oriented databases.

4
00:00:07,770 --> 00:00:10,875
These are similar in concept
to relational databases.

5
00:00:10,875 --> 00:00:13,030
And in a relational database,

6
00:00:13,030 --> 00:00:15,300
one of the most common things that we do

7
00:00:15,300 --> 00:00:19,500
is using the data that's
organized into rows of tables,

8
00:00:19,500 --> 00:00:21,450
we'll perform select queries

9
00:00:21,450 --> 00:00:26,450
that enable us to pick off
specific columns of information

10
00:00:26,530 --> 00:00:28,520
across either all of the rows

11
00:00:28,520 --> 00:00:33,100
or only the rows that meet
certain search criteria.

12
00:00:33,100 --> 00:00:36,390
So one of the problems
when you're dealing with

13
00:00:36,390 --> 00:00:38,150
data in relational tables

14
00:00:38,150 --> 00:00:41,900
is that such queries can
actually perform really poorly

15
00:00:41,900 --> 00:00:44,810
because you have to get
every single matching row,

16
00:00:44,810 --> 00:00:46,840
find the required column,

17
00:00:46,840 --> 00:00:49,800
and throw away the rest of the row.

18
00:00:49,800 --> 00:00:53,800
So columnar databases
are basically optimized

19
00:00:53,800 --> 00:00:58,667
for searching columns of data very rapidly

20
00:00:59,600 --> 00:01:01,820
in big data applications.

21
00:01:01,820 --> 00:01:03,830
So I provided a couple of links here

22
00:01:03,830 --> 00:01:05,330
where you can learn a little bit more

23
00:01:05,330 --> 00:01:07,220
about column-oriented databases.

24
00:01:07,220 --> 00:01:10,880
But let's just take for
consideration purposes

25
00:01:10,880 --> 00:01:15,280
our authors table from the
earlier examples in this lesson.

26
00:01:15,280 --> 00:01:17,619
And as you can see, we
have some ID numbers

27
00:01:17,619 --> 00:01:19,360
for the different authors.

28
00:01:19,360 --> 00:01:21,490
And we have first names and last names.

29
00:01:21,490 --> 00:01:23,950
So we have rows representing authors,

30
00:01:23,950 --> 00:01:27,640
and columns representing
specific pieces of data

31
00:01:27,640 --> 00:01:28,950
for those authors.

32
00:01:28,950 --> 00:01:33,650
And the ID column was one of
the three columns in our table.

33
00:01:33,650 --> 00:01:36,660
What you're seeing here is
the way it got displayed

34
00:01:36,660 --> 00:01:40,730
as a data frame when
we rendered that table

35
00:01:40,730 --> 00:01:43,400
after querying its total contents.

36
00:01:43,400 --> 00:01:44,960
Now in a columnar database,

37
00:01:44,960 --> 00:01:49,460
rather than storing the data
on an author by author basis,

38
00:01:49,460 --> 00:01:53,250
we store the data based on an
attribute by attribute basis.

39
00:01:53,250 --> 00:01:55,650
So the first column, the ID numbers,

40
00:01:55,650 --> 00:01:58,429
is going to be stored in order by row.

41
00:01:58,429 --> 00:02:02,900
The second set of values
will be the first names

42
00:02:02,900 --> 00:02:04,030
in the same order.

43
00:02:04,030 --> 00:02:05,677
And then we'll have the last names

44
00:02:05,677 --> 00:02:07,172
in the same order as well.

45
00:02:07,172 --> 00:02:12,172
So because the column elements
are maintained in row order,

46
00:02:12,210 --> 00:02:14,860
when you get the value
at a particular index

47
00:02:14,860 --> 00:02:16,440
from one of the columns,

48
00:02:16,440 --> 00:02:19,780
you can get the
corresponding pieces of data

49
00:02:19,780 --> 00:02:21,518
from the other columns if you need them

50
00:02:21,518 --> 00:02:26,370
by accessing the same index
number in those columns as well.

51
00:02:26,370 --> 00:02:29,200
And a couple of the
popular columnar databases

52
00:02:29,200 --> 00:02:32,993
are MariaDB's ColumnStore and HBase.