1
00:00:00,060 --> 00:00:05,860
So the first thing we're going to discuss is master back up replication, sometimes called master stand

2
00:00:05,910 --> 00:00:06,920
by replication.

3
00:00:07,140 --> 00:00:15,240
And the idea here is you have one leader digging a base node that accepts rights and except grill operations

4
00:00:15,240 --> 00:00:21,360
such as create table or alternative database stuff that you change in the database, create a new database

5
00:00:21,360 --> 00:00:28,350
right versus right, where you're inserting Neuros in a table or updating in your arrow or deleting

6
00:00:28,350 --> 00:00:28,650
rows.

7
00:00:28,680 --> 00:00:29,910
These are called DM's.

8
00:00:30,720 --> 00:00:40,590
So you do all your rights in one master node and there will be a lot of read only stand by Oracle back

9
00:00:40,590 --> 00:00:45,330
up nodes which will receive those rights from the master node.

10
00:00:46,500 --> 00:00:47,220
So that's the idea.

11
00:00:47,250 --> 00:00:50,420
You don't write to those notes yourself as a client.

12
00:00:50,430 --> 00:00:58,400
You always write to the master and the master replicates the data to the standby notes the beautiful.

13
00:00:58,410 --> 00:01:04,380
This is like it's very simple to implement because you don't have conflicts is as one node that takes

14
00:01:04,380 --> 00:01:10,890
rights and that's at nobody else is going to to have rights as well.

15
00:01:10,890 --> 00:01:11,740
Right in the same time.

16
00:01:12,870 --> 00:01:13,800
So here's how it looks like.

17
00:01:13,810 --> 00:01:22,740
So this is the red one is the master node and the green ones are the stand-by replica nodes.

18
00:01:23,070 --> 00:01:29,310
And if I want to insert or update or even create a new table or even to create a database or do anything,

19
00:01:29,700 --> 00:01:40,230
I always do it to the master of the red database instance and that the act of doing that will propagate

20
00:01:40,410 --> 00:01:48,300
those right to the back and nodes because there is a TCP connection between the replica node and the

21
00:01:48,300 --> 00:01:54,140
master node that is like full bidirectional ready with a specific database user as well.

22
00:01:54,450 --> 00:01:55,690
And this is the same thing here.

23
00:01:55,770 --> 00:02:04,800
So anything you do here, this master node will push it to the child nodes or they call stand by nodes.

24
00:02:04,860 --> 00:02:07,770
And and so that's that's beautiful.

25
00:02:08,100 --> 00:02:11,040
So you want to have questions like Hosain does.

26
00:02:11,070 --> 00:02:18,480
Does the client have to wait for these changes to be pushed to the back end or does it immediately if

27
00:02:18,480 --> 00:02:19,770
it's written to the master?

28
00:02:19,780 --> 00:02:22,410
Can I just immediately get back at once?

29
00:02:22,410 --> 00:02:30,400
It's written to the writer said, look, this is another kind of debate and discussion and people are

30
00:02:30,440 --> 00:02:32,430
already doing PhDs.

31
00:02:32,430 --> 00:02:36,750
And what is best this is called synchronous versus asynchronous one I'm going to talk about in the other

32
00:02:36,750 --> 00:02:36,990
room.

33
00:02:37,530 --> 00:02:42,360
So in master backup or application, the beauty of this, if you're doing a read, you can either do

34
00:02:42,360 --> 00:02:44,760
a read against the master.

35
00:02:45,360 --> 00:02:50,730
You're going to get the fresh Aldo's up to date data, but you can always read against the replicas

36
00:02:50,730 --> 00:02:51,920
again, the standby data.

37
00:02:52,050 --> 00:02:53,940
That's absolutely fine.

38
00:02:53,940 --> 00:02:56,490
You can read against those, but you cannot try to those.

39
00:02:56,940 --> 00:03:01,560
The beauty of this is like if this is in a region, a specific region in Asia and this is in a specific

40
00:03:01,560 --> 00:03:05,070
region in Europe and this is in America, you can still do that.

41
00:03:05,640 --> 00:03:07,080
Yeah, the propagation will take time.

42
00:03:07,080 --> 00:03:16,830
But if you if you can take the hit on this consistency, the eventual consistency, and that's what

43
00:03:16,890 --> 00:03:17,790
the model is.

44
00:03:18,420 --> 00:03:23,690
Yes, relational database is half eventual consistency, not just no sequel.

45
00:03:23,940 --> 00:03:27,360
This is a relational database model, yet it has eventual distances.

46
00:03:27,480 --> 00:03:33,330
So those clients will see that that might see the updates a little bit later.

47
00:03:33,600 --> 00:03:35,580
If you can live with that, that's fine.

48
00:03:35,730 --> 00:03:38,970
If you don't, there are ways to solve those, obviously.

49
00:03:39,840 --> 00:03:41,670
But that's the idea here.

50
00:03:41,670 --> 00:03:43,620
So you can you can scale horizontally.

51
00:03:43,620 --> 00:03:44,850
And that's a beautiful this.