1
00:00:07,100 --> 00:00:10,680
- Now let's review an
introduction to Amazon Neptune.

2
00:00:10,680 --> 00:00:13,410
Now with Amazon Neptune, we gain access

3
00:00:13,410 --> 00:00:16,410
to a fully-managed graph database.

4
00:00:16,410 --> 00:00:19,570
And this graph database is optimized

5
00:00:19,570 --> 00:00:21,150
for storing relationships.

6
00:00:21,150 --> 00:00:24,410
Potentially billions of relationships.

7
00:00:24,410 --> 00:00:26,080
Now when we talk about relationships,

8
00:00:26,080 --> 00:00:27,150
it is important to understand

9
00:00:27,150 --> 00:00:30,440
that we're not necessarily
talking about relational data.

10
00:00:30,440 --> 00:00:33,610
If we were to compare this
to a relational database,

11
00:00:33,610 --> 00:00:36,230
where we have relational data.

12
00:00:36,230 --> 00:00:38,240
For example, in a relational database,

13
00:00:38,240 --> 00:00:41,490
you may have a user record

14
00:00:41,490 --> 00:00:44,950
and then a list of orders, right?

15
00:00:44,950 --> 00:00:46,100
You may have a one-to-many.

16
00:00:46,100 --> 00:00:48,530
A list of orders that that user placed

17
00:00:48,530 --> 00:00:50,400
in an E-commerce kind of situation.

18
00:00:50,400 --> 00:00:53,060
And then each order may have a one-to-many

19
00:00:53,060 --> 00:00:56,340
with the list of items
that were in that order.

20
00:00:56,340 --> 00:00:57,173
Right?

21
00:00:57,173 --> 00:01:00,490
And so, you would, in order
to form relationships,

22
00:01:00,490 --> 00:01:02,420
in order to form a relationship

23
00:01:02,420 --> 00:01:06,210
between a user and a
product that they purchase,

24
00:01:06,210 --> 00:01:09,620
you have to derive that
from that relational data.

25
00:01:09,620 --> 00:01:12,110
So relational data is
not quite the same thing

26
00:01:12,110 --> 00:01:15,363
as relationship data
is in a graph database.

27
00:01:16,292 --> 00:01:19,543
And so a graph database
more easily allows us

28
00:01:21,970 --> 00:01:26,410
to form a direct
relationship between a user

29
00:01:26,410 --> 00:01:28,100
and a product that they purchase.

30
00:01:28,100 --> 00:01:30,300
A direct relationship between a user

31
00:01:30,300 --> 00:01:32,580
and a movie that they watch, right?

32
00:01:32,580 --> 00:01:37,200
And so, when you think about
those kinds of data sets,

33
00:01:37,200 --> 00:01:40,140
you think about shopping on Amazon.com.

34
00:01:40,140 --> 00:01:42,690
There's a relationship between myself

35
00:01:42,690 --> 00:01:45,730
and the products that
I've actually purchased.

36
00:01:45,730 --> 00:01:47,510
There's a relationship between myself

37
00:01:47,510 --> 00:01:49,700
and maybe my friend.

38
00:01:49,700 --> 00:01:52,934
And then there's relationship
between other people

39
00:01:52,934 --> 00:01:56,090
of the similar demographic
and the products

40
00:01:56,090 --> 00:01:57,570
that they bought.

41
00:01:57,570 --> 00:02:00,420
If we take this kind of thinking

42
00:02:00,420 --> 00:02:04,020
to something like Netflix
or a social graph,

43
00:02:04,020 --> 00:02:06,380
you can see that there's relationships

44
00:02:06,380 --> 00:02:07,860
between myself and my friends,

45
00:02:07,860 --> 00:02:11,110
between myself and the
movies that I've watched,

46
00:02:11,110 --> 00:02:13,570
and then a relationship between my friends

47
00:02:13,570 --> 00:02:15,230
and the movies that
they've watched, right?

48
00:02:15,230 --> 00:02:18,270
So we can take this kind
of relationship data

49
00:02:18,270 --> 00:02:20,830
and one of the classic use cases for it

50
00:02:20,830 --> 00:02:23,600
is creating recommendation engines.

51
00:02:23,600 --> 00:02:25,758
We can use relationships to see.

52
00:02:25,758 --> 00:02:29,740
Well, if I have a relationship
between this movie,

53
00:02:29,740 --> 00:02:32,520
and a friend has a relationship
between this movie,

54
00:02:32,520 --> 00:02:35,170
then perhaps I will like
the movie that they liked,

55
00:02:35,170 --> 00:02:37,300
and maybe they would
like one that I liked.

56
00:02:37,300 --> 00:02:40,000
That's just one classic example.

57
00:02:40,000 --> 00:02:44,100
And so, with Amazon Neptune,
when you have a use case

58
00:02:44,100 --> 00:02:46,040
for relationship type of data,

59
00:02:46,040 --> 00:02:48,410
and you want to be able to transverse

60
00:02:48,410 --> 00:02:52,238
that graph of relationships
and derive meaning from that,

61
00:02:52,238 --> 00:02:55,160
then, of course, we need a
model for storing that data.

62
00:02:55,160 --> 00:02:57,610
And Amazon Neptune supports

63
00:02:57,610 --> 00:02:59,160
a number of different graph models,

64
00:02:59,160 --> 00:03:02,170
including Property Graph and RDF.

65
00:03:02,170 --> 00:03:06,473
It also supports a couple
of different query languages

66
00:03:08,130 --> 00:03:11,380
including Apache TinkerPop,
Gremlin, and SPARQL.

67
00:03:11,380 --> 00:03:14,470
So when we actually want to
program against that database

68
00:03:14,470 --> 00:03:16,770
and write a language that allows us

69
00:03:16,770 --> 00:03:19,690
to traverse the relationship graph,

70
00:03:19,690 --> 00:03:21,740
we can write that and express that

71
00:03:21,740 --> 00:03:24,163
in TinkerPop, Gremlin, or SPARQL.

72
00:03:25,100 --> 00:03:26,670
And of course, like other services,

73
00:03:26,670 --> 00:03:31,293
similar to DynamoDB, and
S3, and Aurora, and so on,

74
00:03:32,220 --> 00:03:34,270
Amazon Neptune is also the data

75
00:03:34,270 --> 00:03:37,610
that we write to these
tables is also replicated

76
00:03:37,610 --> 00:03:39,710
across multiple AZs.

77
00:03:39,710 --> 00:03:42,930
So as you can see,
durability, fault tolerance,

78
00:03:42,930 --> 00:03:45,723
high availability is already built in.

79
00:03:47,440 --> 00:03:51,810
Neptune also performs a
continuous backup to S3.

80
00:03:51,810 --> 00:03:54,410
So, unlike RDS, where we have to specify

81
00:03:54,410 --> 00:03:59,410
a maintenance window, backups
are done continuously to S3.

82
00:04:01,000 --> 00:04:03,910
Neptune also supports,
very much like Aurora,

83
00:04:03,910 --> 00:04:07,660
Neptune also supports
up to 15 read replicas.

84
00:04:07,660 --> 00:04:11,670
So again, if we have a
very read-heavy application

85
00:04:11,670 --> 00:04:16,570
that's constantly pulling and
querying these relationships,

86
00:04:16,570 --> 00:04:20,240
then we can offload that
to the read replicas

87
00:04:20,240 --> 00:04:22,990
and allow the master of
the focused-on rights.

88
00:04:22,990 --> 00:04:26,460
And by doing that, we can
increase the performance

89
00:04:26,460 --> 00:04:28,670
for the system as a whole.

90
00:04:28,670 --> 00:04:31,460
So again, when you think about performing

91
00:04:31,460 --> 00:04:35,480
or modeling and querying
direct relationships

92
00:04:35,480 --> 00:04:38,600
between persons and products
and persons and movies,

93
00:04:38,600 --> 00:04:42,360
or between different people,
for example, in a social graph,

94
00:04:42,360 --> 00:04:44,173
think about Amazon Neptune.