1
00:00:00,150 --> 00:00:06,750
‫What is going on, guys, my name is Hussein, and as some of you might know, I'm still in vacation

2
00:00:07,350 --> 00:00:16,890
‫and I don't have my regular office and my camera, so I apologize for the audio quality.

3
00:00:17,190 --> 00:00:25,710
‫But I thought I'll make a quick video for the Q&A lecture of the month of December.

4
00:00:25,800 --> 00:00:27,870
‫I started doing this back in October.

5
00:00:28,200 --> 00:00:33,150
‫And you guys seem to enjoy this kind of interaction.

6
00:00:33,620 --> 00:00:41,010
‫Right where I pick a question from the introduction to database and joining Udemy course and elaborate

7
00:00:41,010 --> 00:00:41,430
‫on that.

8
00:00:41,850 --> 00:00:52,110
‫So as you know, I almost answer every single question that you guys ask in the anime section Q&A and

9
00:00:52,110 --> 00:01:00,240
‫but usually, you know, when you when you type things and it's really hard to, you know, I am a horrible

10
00:01:00,240 --> 00:01:00,660
‫person.

11
00:01:00,660 --> 00:01:06,990
‫I like to talk more than I type and my, you know, my writing isn't that great to begin with, but

12
00:01:07,160 --> 00:01:15,630
‫I thought, I'll just make a video to elaborate on one of these answers, and we'll pick one or one

13
00:01:15,630 --> 00:01:18,580
‫or two questions, then elaborate on them.

14
00:01:18,660 --> 00:01:19,890
‫I've always thought so.

15
00:01:20,400 --> 00:01:24,000
‫I thought, I'll start with this question from Shivam.

16
00:01:24,660 --> 00:01:31,870
‫Why will any update and post Chris updates all indexes corresponding to that table?

17
00:01:31,880 --> 00:01:36,690
‫This sort of phrase the question of the brand and explain that one.

18
00:01:36,810 --> 00:01:38,220
‫What exactly is that?

19
00:01:38,910 --> 00:01:41,670
‫So this is referring to the must watch video.

20
00:01:43,700 --> 00:01:49,340
‫A lecturer, actually, of how tables and indexes are a start to this, I really, really recommend

21
00:01:49,340 --> 00:01:59,600
‫you watch this before really starting the course, because this is the fundamental things of the building

22
00:01:59,600 --> 00:02:05,600
‫block that you need to understand and build upon yourself.

23
00:02:05,600 --> 00:02:09,380
‫Once you understand these basics, it's like understanding math.

24
00:02:09,740 --> 00:02:15,590
‫Once you adjust to how ad addition works, you know how multiplication will work.

25
00:02:15,590 --> 00:02:20,750
‫And once you understand how multiplication works, division just comes and naturally, everything is

26
00:02:20,750 --> 00:02:23,750
‫derived from the thing that comes before it.

27
00:02:23,960 --> 00:02:24,290
‫Right?

28
00:02:24,770 --> 00:02:30,140
‫You don't memorize right things usually like in math.

29
00:02:30,140 --> 00:02:31,580
‫You never memorize an equation.

30
00:02:31,580 --> 00:02:34,460
‫You know how it works and you derive it right?

31
00:02:34,470 --> 00:02:37,910
‫You can solve anything by understanding the basic things.

32
00:02:38,150 --> 00:02:42,800
‫That's why we need to do the same thing with any domain.

33
00:02:43,340 --> 00:02:45,500
‫Really, engineering is one of them, right?

34
00:02:45,770 --> 00:02:47,150
‫We tend to memorize.

35
00:02:47,150 --> 00:02:55,530
‫We tend to say, Oh, it's the way it is or my school or or MongoDB is faster is just the way it is.

36
00:02:55,550 --> 00:02:57,280
‫No, it's just at the end of the day.

37
00:02:57,290 --> 00:03:01,700
‫It's just basic building things, building blocks at once.

38
00:03:01,700 --> 00:03:05,780
‫You understand these basic things you can.

39
00:03:05,810 --> 00:03:07,100
‫You don't need anyone.

40
00:03:07,100 --> 00:03:08,420
‫You don't need me anymore.

41
00:03:08,420 --> 00:03:12,200
‫You don't need anyone to tell you, Why is this better than this?

42
00:03:12,380 --> 00:03:16,930
‫You make the decision because at the end of the day, you ask, Is it OK?

43
00:03:16,940 --> 00:03:18,110
‫Why is this OK?

44
00:03:18,110 --> 00:03:19,250
‫How does this work?

45
00:03:19,670 --> 00:03:20,570
‫You know how it works?

46
00:03:20,570 --> 00:03:24,350
‫Because at the end of the day, everything is all comes to databases.

47
00:03:24,830 --> 00:03:32,480
‫And we try to be smart in databases and start, OK, what is happening really right?

48
00:03:32,750 --> 00:03:37,790
‫When you're issuing a sequel state, once you understand what is exactly is happening?

49
00:03:39,120 --> 00:03:46,500
‫You must go to another database and ask the same question, how is this database system doing compared

50
00:03:46,500 --> 00:03:47,430
‫to PostgreSQL?

51
00:03:47,430 --> 00:03:48,440
‫How was mostly called, though?

52
00:03:48,500 --> 00:03:53,820
‫It wasn't because at the end of the day, they have to solve a basic problem, right?

53
00:03:54,690 --> 00:03:59,370
‫How to optimize Io your writing, reading or writing to disk?

54
00:03:59,640 --> 00:04:00,690
‫You write it to memory.

55
00:04:00,690 --> 00:04:04,500
‫You have to flush the disk you you want to solve the durability problem.

56
00:04:04,800 --> 00:04:12,960
‫At the end of the day, it's all, you know, first principles that everybody need to adhere to.

57
00:04:13,320 --> 00:04:14,820
‫There is no exception.

58
00:04:15,300 --> 00:04:23,430
‫And the problem is marketing gets into the way Oh, this is like in memory, whatever the right or this

59
00:04:23,430 --> 00:04:31,080
‫is that faster or this is on the cloud something you know, it just jargon gets into the wing, gets

60
00:04:31,080 --> 00:04:31,890
‫confusing.

61
00:04:31,890 --> 00:04:38,550
‫So that other way, let's get in to the first question, which is why will any update impulse Chris

62
00:04:38,550 --> 00:04:41,100
‫updates indexes correspond to the table?

63
00:04:41,370 --> 00:04:49,890
‫So this is something I mentioned in the in this lecture, and we do need to understand the indexes are

64
00:04:49,890 --> 00:04:51,450
‫there for a reason, right?

65
00:04:52,050 --> 00:05:01,110
‫Indexes are there so that we avoid searching millions and millions and millions of things.

66
00:05:01,380 --> 00:05:05,550
‫Whether these things or documents rows, blobs, who cares, right?

67
00:05:06,360 --> 00:05:10,610
‫Indexes allow us to, you know, take shortcuts.

68
00:05:10,620 --> 00:05:12,480
‫Think of an index like this.

69
00:05:12,720 --> 00:05:15,090
‫You're going to a secretary office.

70
00:05:15,570 --> 00:05:17,640
‫Oh, find me a company, zebra.

71
00:05:17,940 --> 00:05:18,300
‫Right?

72
00:05:18,690 --> 00:05:23,790
‫And there are a lot of, you know, entries and find me.

73
00:05:23,790 --> 00:05:27,030
‫The phone number of this company is called Zipporah, right?

74
00:05:27,570 --> 00:05:32,370
‫You can take the phone book and search from A to Z.

75
00:05:33,420 --> 00:05:37,320
‫But an index says, OK, zebra cells with a Z.

76
00:05:37,650 --> 00:05:41,520
‫Let's only search the ZS or the ZS if you're from the UK.

77
00:05:42,390 --> 00:05:44,880
‫And that's basically what an index is.

78
00:05:44,890 --> 00:05:46,810
‫It just narrows the search space.

79
00:05:47,280 --> 00:05:52,200
‫So like, oh, while researching a b, c d and all that stuff, search only those things you need.

80
00:05:52,860 --> 00:05:54,360
‫That was exactly what the index does.

81
00:05:55,320 --> 00:06:05,100
‫It just smartly, based on the column that you indexed builds a data structure, right, which needs

82
00:06:05,520 --> 00:06:08,570
‫to be stored, needs to be put in memory.

83
00:06:08,590 --> 00:06:09,720
‫Need to be issued.

84
00:06:10,200 --> 00:06:11,820
‫So you still need to manage this thing.

85
00:06:11,940 --> 00:06:12,510
‫It's not free.

86
00:06:12,510 --> 00:06:13,140
‫Nothing is free.

87
00:06:13,620 --> 00:06:13,860
‫Right?

88
00:06:14,160 --> 00:06:15,060
‫So you build that.

89
00:06:15,150 --> 00:06:20,550
‫They just like you and you search it first and then that data sector, that index will tell you, Oh,

90
00:06:20,550 --> 00:06:24,590
‫go to this page and pull this role.

91
00:06:24,600 --> 00:06:25,740
‫It's it's there.

92
00:06:25,740 --> 00:06:32,700
‫It's 100 percent there or some indexes are, you know, proper probabilistic.

93
00:06:32,700 --> 00:06:40,650
‫If I tell you, I I, it's it's it's definitely not there, but it may be there.

94
00:06:40,950 --> 00:06:45,210
‫And that's that's also a useful so-called blue filters that are very useful stuff.

95
00:06:45,360 --> 00:06:51,570
‫So, yeah, so once you understand the indexes or data structures that need to be stored, that need

96
00:06:51,570 --> 00:07:01,350
‫to be maintained because the moment you update your your source tables with the data, you need to keep

97
00:07:01,350 --> 00:07:02,760
‫that index up to date.

98
00:07:03,030 --> 00:07:03,320
‫Right?

99
00:07:03,690 --> 00:07:06,660
‫It's like adding a new echo.

100
00:07:06,870 --> 00:07:07,170
‫Right?

101
00:07:07,470 --> 00:07:16,260
‫And you can't any phone number, you need to have the index right there with the letter have that point

102
00:07:16,260 --> 00:07:18,150
‫to the, you know, the right entry.

103
00:07:18,510 --> 00:07:19,200
‫It's the same thing.

104
00:07:20,740 --> 00:07:21,760
‫Welcome to the question.

105
00:07:22,780 --> 00:07:30,460
‫The question is when you impose Christmas is specifically in politics again, those don't get don't

106
00:07:30,460 --> 00:07:35,140
‫get caught up with definitions or implementation details.

107
00:07:35,530 --> 00:07:39,160
‫Just understand how how each database does stuff.

108
00:07:39,190 --> 00:07:43,580
‫You know, it's just it's this the way they chose to do this, right?

109
00:07:45,100 --> 00:07:48,400
‫All right.

110
00:07:48,910 --> 00:07:52,630
‫Choose a specific design and they stuck with it.

111
00:07:52,960 --> 00:07:55,150
‫It has pros and cons, right?

112
00:07:55,270 --> 00:07:59,620
‫And I'm not going to tell you the pros and cons because you forgot to figure it out yourself with logic

113
00:08:00,310 --> 00:08:00,670
‫and.

114
00:08:02,060 --> 00:08:05,030
‫The index, as I said, the index is.

115
00:08:06,180 --> 00:08:13,530
‫Need to point back to the table, right, which page and which role they need to read, right?

116
00:08:14,630 --> 00:08:19,820
‫And when you do that, what do we point to exactly?

117
00:08:20,210 --> 00:08:24,020
‫We need some sort of identification of the road, right?

118
00:08:24,500 --> 00:08:29,100
‫This is what is called Tibble ID and Postgres at least every year.

119
00:08:29,120 --> 00:08:31,790
‫And so that has a unique tuple ID.

120
00:08:32,430 --> 00:08:32,770
‫Right?

121
00:08:32,990 --> 00:08:37,070
‫And if you updated and specifically in Postgres, you get a new Tupelo ID.

122
00:08:37,550 --> 00:08:37,940
‫Right?

123
00:08:38,720 --> 00:08:40,880
‫But the road is the same logically.

124
00:08:41,150 --> 00:08:46,460
‫So to you or the reader, it's the same road, but it's a different tuple does doesn't make sense.

125
00:08:47,390 --> 00:08:55,070
‫So one logical row could have many, many tuple entries, and each tuple has a unique identifier.

126
00:08:55,370 --> 00:08:58,580
‫That's unique identifier is what the index stores.

127
00:08:58,850 --> 00:09:03,230
‫And this is what effectively does the reading.

128
00:09:03,260 --> 00:09:03,560
‫All right.

129
00:09:03,560 --> 00:09:04,340
‫So where are we?

130
00:09:04,620 --> 00:09:05,420
‫We talked about.

131
00:09:07,650 --> 00:09:20,340
‫Having index data structures that are kind of separated from this large pool of raw data where the table

132
00:09:20,340 --> 00:09:27,000
‫exist, where the documents exist, and these are multiple indexes, you know, because you can create

133
00:09:27,000 --> 00:09:35,820
‫an index on Field A that will create a unique index data structure that will have X amount of bytes

134
00:09:36,330 --> 00:09:41,060
‫that would be stored in this that will need to be fetched and need to be put in memory that will need

135
00:09:41,160 --> 00:09:42,350
‫to be queried.

136
00:09:42,360 --> 00:09:42,740
‫Read.

137
00:09:43,830 --> 00:09:48,330
‫And you can have many, many, many of these indexes, right based on your table.

138
00:09:49,320 --> 00:09:49,530
‫Right.

139
00:09:49,650 --> 00:09:54,300
‫So let's say you have three indexes each for each column.

140
00:09:54,540 --> 00:09:56,940
‫Column A. Have an index.

141
00:09:56,940 --> 00:09:57,990
‫Column B have an index.

142
00:09:57,990 --> 00:10:03,420
‫Column C has an index which points back to the table.

143
00:10:04,410 --> 00:10:08,280
‫Now this is a design choice that the post Christine made.

144
00:10:10,970 --> 00:10:19,250
‫When you search an index, right, you found the value that you're looking for, the value is within

145
00:10:19,250 --> 00:10:21,410
‫us right now.

146
00:10:23,760 --> 00:10:28,710
‫What does the value contain in the index, right?

147
00:10:28,800 --> 00:10:34,230
‫Because you're searching the keys of the index, the value because it's a key value almost right concept,

148
00:10:34,710 --> 00:10:34,990
‫right?

149
00:10:35,250 --> 00:10:40,050
‫The value of the element is basically the tuple idea that we talked about.

150
00:10:40,260 --> 00:10:40,550
‫Right.

151
00:10:42,240 --> 00:10:53,040
‫Means that the tuple I.D., the ephemeral thing that keeps changing is stored on all the indexes.

152
00:10:56,930 --> 00:11:02,590
‫So now think about it, if I update my role, right?

153
00:11:03,440 --> 00:11:14,330
‫It's a unique role and I made an update to a single field there and new to pull it will be created based

154
00:11:14,330 --> 00:11:15,410
‫on Bosca's design.

155
00:11:15,650 --> 00:11:17,060
‫Again, this is only Bosworth.

156
00:11:17,540 --> 00:11:20,030
‫Some other databases do a completely different design.

157
00:11:20,220 --> 00:11:22,160
‫So and this is what I wanted to keep.

158
00:11:22,160 --> 00:11:28,100
‫An open mind is like, Oh, this is not the way or the right way or the wrong way.

159
00:11:28,130 --> 00:11:29,510
‫This is just how it's done.

160
00:11:30,170 --> 00:11:36,050
‫You can come up with something to challenge any of the stuff that is being done today and completely

161
00:11:36,050 --> 00:11:38,510
‫revolutionize database engineering, right?

162
00:11:39,410 --> 00:11:40,790
‫And not on a fluffy way.

163
00:11:40,850 --> 00:11:42,080
‫Like we do it today.

164
00:11:42,080 --> 00:11:49,400
‫Unfortunately, I don't think the fluff these days, nothing is actually legit that is being discussed.

165
00:11:50,330 --> 00:11:59,750
‫So now when I update my roll, call them a rhyme, what do we need to do right?

166
00:12:00,080 --> 00:12:00,770
‫First of all.

167
00:12:02,440 --> 00:12:08,280
‫We changed that content of column, and we didn't touch B or C, right?

168
00:12:09,990 --> 00:12:12,110
‫What happens in post poses create.

169
00:12:12,430 --> 00:12:18,740
‫Exact copy of that tuple raid as a new tuple.

170
00:12:18,790 --> 00:12:27,210
‫You have a new tuple ID, a brand new RAM, but it's effectively logically to you as a user.

171
00:12:27,690 --> 00:12:31,110
‫It's the same row you just updated it, right?

172
00:12:32,370 --> 00:12:34,140
‫So now there is a new tuple idea.

173
00:12:34,440 --> 00:12:39,150
‫We updated Field A, which has an index DNC we don't touch.

174
00:12:40,170 --> 00:12:41,250
‫But here's the problem.

175
00:12:42,570 --> 00:12:48,630
‫If you actually look in the Index B and the index, see.

176
00:12:50,380 --> 00:12:57,670
‫Right, and the index, all of them have entries which point to the old tuple.

177
00:12:58,880 --> 00:12:59,180
‫Right.

178
00:12:59,610 --> 00:13:04,250
‫Because each table again, each double has a unique identifier.

179
00:13:04,850 --> 00:13:07,760
‫So all of these guys are pointing to the old one.

180
00:13:10,090 --> 00:13:10,450
‫Right.

181
00:13:11,020 --> 00:13:15,370
‫So regardless of this old indexes.

182
00:13:17,000 --> 00:13:21,110
‫Need to know about this new to idea.

183
00:13:21,980 --> 00:13:28,770
‫I feel like I talked about this, and I apologize if I did, but I guess it doesn't hurt and I repeat

184
00:13:28,790 --> 00:13:29,880
‫it right.

185
00:13:29,900 --> 00:13:36,980
‫So all these indexes need the knowledge of the news to pull it right, because if someone start searching,

186
00:13:38,360 --> 00:13:46,430
‫let's say, Index C, which we didn't touch for a value that points to that role, they will find the

187
00:13:46,700 --> 00:13:52,820
‫old to pull I.D. and they will get at the old value.

188
00:13:52,830 --> 00:13:54,800
‫Yeah, see, it wasn't touched.

189
00:13:54,800 --> 00:14:02,840
‫But what if they said they'd select a from their bill where sequel law the law was pointed to the old

190
00:14:03,140 --> 00:14:03,940
‫double idea?

191
00:14:04,370 --> 00:14:07,430
‫So now they will get the old value of it, not the one we updated.

192
00:14:07,730 --> 00:14:19,400
‫So regardless, we need to update those indexes right to have the new tuple idea and.

193
00:14:21,900 --> 00:14:29,700
‫And that's technically a problem in ample space, and that's why Ober complained one of the reasons

194
00:14:29,700 --> 00:14:37,370
‫Obama complained about Bush scares about this because now, hey, I updated the column a.

195
00:14:37,620 --> 00:14:40,620
‫Why are you touching indexes B and C?

196
00:14:40,860 --> 00:14:50,170
‫So if you have a lot of indexes right, then what are you going to do is you will end up touching and

197
00:14:50,190 --> 00:14:51,000
‫I say touching.

198
00:14:51,000 --> 00:14:58,640
‫You're actually physically updating, doing I overdoing the rights to these data structures.

199
00:14:58,650 --> 00:15:00,080
‫That's why I make you.

200
00:15:00,210 --> 00:15:03,480
‫I want you just think these are things that are not in the air.

201
00:15:03,480 --> 00:15:07,890
‫These needs to be maintained and the maintenance of this is costly.

202
00:15:08,350 --> 00:15:08,580
‫Right.

203
00:15:08,580 --> 00:15:15,350
‫So you have a lot of all of this indexes then becomes a problem, but post-crisis.

204
00:15:15,360 --> 00:15:16,940
‫I believe that's another cautions.

205
00:15:16,950 --> 00:15:20,790
‫I called into something called the heart optimization

206
00:15:23,250 --> 00:15:32,100
‫equity heap only to and said, All right, we really didn't touch B or C.

207
00:15:32,550 --> 00:15:37,080
‫We only touched a so does have an index if it doesn't have an index.

208
00:15:38,290 --> 00:15:40,920
‫We have to update index, Hey, guys.

209
00:15:41,260 --> 00:15:47,260
‫You change the representation of the row for Column A. There is an index on air.

210
00:15:47,500 --> 00:15:56,650
‫We need to load the index a have the new value of that right because then the B three has to be reshuffled.

211
00:15:56,650 --> 00:15:58,090
‫The B three have to be updated.

212
00:15:58,090 --> 00:15:59,800
‫The keys have to be put into place.

213
00:16:00,220 --> 00:16:02,620
‫We need to do all of this stuff regardless.

214
00:16:02,770 --> 00:16:07,510
‫This is an update to that, a feel which has an index.

215
00:16:07,510 --> 00:16:09,430
‫So you have to update it, feel it.

216
00:16:09,550 --> 00:16:17,770
‫But the performance optimization that the post-arrest team did is like what B and C shouldn't really

217
00:16:17,830 --> 00:16:24,740
‫need to be touched technically because they were they weren't touched, right?

218
00:16:24,760 --> 00:16:25,840
‫Why are you updating them?

219
00:16:26,080 --> 00:16:28,860
‫And that's what's called the heart optimization, which is right here.

220
00:16:28,900 --> 00:16:30,790
‫I'm up against it on my face.

221
00:16:31,510 --> 00:16:32,070
‫It's right here.

222
00:16:32,080 --> 00:16:45,010
‫Hot optimization heap on the table, which effectively keeps the old tuple I.D. in the index right,

223
00:16:45,370 --> 00:16:50,740
‫but updates the heap table, which we are already doing because we updated the roll right.

224
00:16:51,820 --> 00:17:00,100
‫So then we do a pointer in the heap that the old tuple ID actually points to the new tuple I.D. in the

225
00:17:00,100 --> 00:17:05,950
‫hip itself, so that if someone came to the old Apple ID, we let them know her.

226
00:17:06,010 --> 00:17:11,710
‫Hey, by the way, if you want the latest representation of this double, it's has been updated, is

227
00:17:11,710 --> 00:17:12,190
‫not it?

228
00:17:12,400 --> 00:17:20,240
‫There is a new one here and this is done in the same page again, or to this larger, very important

229
00:17:20,240 --> 00:17:25,150
‫to watch this lecture to know what the difference between a page and a hole and all that jazz, right?

230
00:17:25,360 --> 00:17:31,630
‫So the page have many rows and and each row has many tables and this case of PostgreSQL.

231
00:17:32,080 --> 00:17:36,460
‫So now you will have these multiple representation.

232
00:17:36,460 --> 00:17:37,000
‫So it is.

233
00:17:37,270 --> 00:17:45,250
‫They inserted some metadata in the page that says, Oh, tuple ID ID law is actually there is a new

234
00:17:45,250 --> 00:17:45,770
‫version of it.

235
00:17:45,790 --> 00:17:48,220
‫Go and jump to that some metadata.

236
00:17:49,730 --> 00:18:00,350
‫Keep in mind only if the new updated top all happened to be in the same age.

237
00:18:01,650 --> 00:18:02,850
‫As the old couple.

238
00:18:04,960 --> 00:18:10,480
‫So does that mean because it has a fixed size, it's eight Keane, Paul Scholes at 16 in my squad.

239
00:18:10,780 --> 00:18:16,480
‫So the moment the moment you update a row in both squares, we create a new tuple idea.

240
00:18:16,780 --> 00:18:22,030
‫If there is a space left in that age, we're going to insert the new Triple ID in the page itself,

241
00:18:22,780 --> 00:18:25,360
‫which is, as the bureau does, the best thing you can do.

242
00:18:25,720 --> 00:18:32,040
‫Because now if I fit the page all the top of the it is in the speech, the moment you start jumping

243
00:18:32,050 --> 00:18:35,200
‫to another page, that's where you need more oil.

244
00:18:35,710 --> 00:18:43,560
‫Again, it really depends above the disc size and and and the block sizes of the operating system file

245
00:18:43,570 --> 00:18:44,080
‫systems.

246
00:18:44,260 --> 00:18:49,210
‫It's a little bit more complicated than that, but just think of its kind of logically right.

247
00:18:50,080 --> 00:18:54,400
‫If things isn't the same page, then it's more efficient collectively.

248
00:18:55,600 --> 00:19:00,640
‫So a lot of divas couldn't figure out something called the fitting factor.

249
00:19:03,140 --> 00:19:08,960
‫How do you type this, Phil factor Google that do you want to know more about it?

250
00:19:10,980 --> 00:19:17,840
‫And Phil Spector is is a page configuration that says, OK, so this page until 80 percent full, this

251
00:19:17,840 --> 00:19:20,090
‫variant is 50 percent, right?

252
00:19:20,090 --> 00:19:25,030
‫When you start, when you create a new page, only fill it and 50 percent leave.

253
00:19:26,000 --> 00:19:33,440
‫What I understand is leave space for updates and DVDs so that we can use it for later, but instead

254
00:19:33,890 --> 00:19:40,520
‫fill it up until X amount of percentage 80 percent, 70 percent, whatever you configure and then leave

255
00:19:40,520 --> 00:19:41,000
‫it empty.

256
00:19:41,840 --> 00:19:51,020
‫And then when you do that, what will happen is that page will leave some empty space in it for future

257
00:19:51,260 --> 00:19:54,290
‫updates so that tables that have been updated.

258
00:19:55,610 --> 00:19:57,830
‫Gets placed in the same page.

259
00:19:58,120 --> 00:19:58,330
‫Right?

260
00:19:58,370 --> 00:20:03,620
‫They will be reused, and that's that's basically the power of two factors very critical.

261
00:20:04,070 --> 00:20:05,810
‫Very, very critical configuration.

262
00:20:05,810 --> 00:20:09,270
‫When you become a DBA, that's just one of the things you have to know, right?

263
00:20:09,860 --> 00:20:14,870
‫It's like there's configuring this based on the work load is completely something.

264
00:20:15,550 --> 00:20:17,030
‫Yeah, yeah.

265
00:20:17,210 --> 00:20:22,310
‫You gets paid to hand over fist for understanding these low level things.

266
00:20:22,610 --> 00:20:22,910
‫Right?

267
00:20:23,690 --> 00:20:27,830
‫Configuring things really can break or make your architecture.

268
00:20:28,940 --> 00:20:29,240
‫All right.

269
00:20:29,240 --> 00:20:30,410
‫So that's that's basically it.

270
00:20:31,150 --> 00:20:35,480
‫So he points to only works of the tuple is in the same page.

271
00:20:35,480 --> 00:20:37,520
‫If it's not in the same page, all bets are off.

272
00:20:37,850 --> 00:20:40,700
‫We'll get up to the indexes regardless of what we're updating.

273
00:20:40,700 --> 00:20:47,200
‫So that's basically why we're updating this indexes.

274
00:20:47,510 --> 00:20:47,990
‫What's going on?

275
00:20:47,990 --> 00:20:54,110
‫I guess this also answers this question, which is kind of similar.

276
00:20:55,760 --> 00:21:03,470
‫Someone had a similar question here, which is like, I got a I have some confusion regarding the heart

277
00:21:03,470 --> 00:21:06,290
‫approach for that identification problem.

278
00:21:06,770 --> 00:21:10,700
‫And I basically I just answered basically this kind of question as well.

279
00:21:11,690 --> 00:21:18,080
‫And so I guess I didn't talk about write amplification, you know, write amplification is the idea

280
00:21:18,080 --> 00:21:19,460
‫of what we talked about.

281
00:21:19,850 --> 00:21:23,600
‫You do a single object as a user.

282
00:21:24,910 --> 00:21:29,140
‫This is there is a lecture called The Right Amplification, and this course by the other guys took it

283
00:21:29,140 --> 00:21:31,720
‫up, so I didn't object.

284
00:21:32,260 --> 00:21:33,790
‫I touched Field A in the.

285
00:21:34,450 --> 00:21:36,660
‫That's all I did as a user select.

286
00:21:36,820 --> 00:21:41,790
‫I did an update statement and I just pushed this thing set.

287
00:21:42,100 --> 00:21:47,110
‫I just updated that single walk into anything that only that column.

288
00:21:47,110 --> 00:21:47,680
‫I touched it.

289
00:21:49,020 --> 00:21:58,530
‫But the the right that you think you did have an amplification side effects, because that's not really

290
00:21:58,530 --> 00:22:08,670
‫the only thing that you have done technically there, you have indexes and plays, you have pages that

291
00:22:08,670 --> 00:22:10,830
‫need to be sorted and shuffle.

292
00:22:10,830 --> 00:22:17,940
‫If you have a clustered index, that's another story that might be that clustering that needs to be

293
00:22:17,940 --> 00:22:18,660
‫done.

294
00:22:19,140 --> 00:22:28,320
‫So we can all be indexes that needs to be updated regardless of this process, that those indexes might

295
00:22:28,320 --> 00:22:36,900
‫have a three rebalancing, which will need more rights as a result, which results in an amplification

296
00:22:36,900 --> 00:22:37,410
‫of right.

297
00:22:37,410 --> 00:22:39,030
‫So you're doing a single thing.

298
00:22:39,060 --> 00:22:44,850
‫But if you're looking at your own desk, you see a lot of a lot of rights, right?

299
00:22:45,660 --> 00:22:46,360
‫Albers.

300
00:22:46,480 --> 00:22:53,980
‫So this specifically when they were doing replication, which I have a whole section on, they had a

301
00:22:53,980 --> 00:22:54,900
‫data replication.

302
00:22:54,900 --> 00:22:59,010
‫They said, Wait a second, why is the network bandwidth overflowing?

303
00:22:59,550 --> 00:23:01,590
‫We didn't do anything is just a single update.

304
00:23:01,860 --> 00:23:04,020
‫They've seen a flood of rights.

305
00:23:04,530 --> 00:23:05,010
‫Right?

306
00:23:05,190 --> 00:23:11,820
‫So that's when they say, OK, well, we did the other day, but we have 20 indexes that have been updated,

307
00:23:11,820 --> 00:23:12,060
‫right?

308
00:23:12,600 --> 00:23:22,770
‫And Bosca's over where it was, was on an old version of Pulseless that didn't support keep people and

309
00:23:22,770 --> 00:23:26,130
‫little pools, which is hot, which they saw this amplification.

310
00:23:26,130 --> 00:23:36,090
‫As a result, bogus steam try to simplify the heap only eatable problem that that amplification by doing

311
00:23:36,090 --> 00:23:41,250
‫the heart optimization they did that many other optimizations as well.

312
00:23:42,270 --> 00:23:42,810
‫All right, guys.

313
00:23:42,840 --> 00:23:43,950
‫That's it for me today.

314
00:23:43,980 --> 00:23:45,510
‫I'm going to send the next one.

315
00:23:46,140 --> 00:23:47,460
‫Hope you enjoyed this lecture?

316
00:23:48,180 --> 00:23:48,930
‫Talk to you later.

317
00:23:49,380 --> 00:23:49,770
‫Goodbye.