1
00:00:00,060 --> 00:00:01,540
‫What is going on, guys?

2
00:00:01,560 --> 00:00:08,850
‫My name is Hussein, and in this video, I want to talk about how database's store information in the

3
00:00:09,090 --> 00:00:13,020
‫disk, and that was a question asked by one of you guys.

4
00:00:13,020 --> 00:00:19,050
‫And it was a very good question because usually we take things for granted.

5
00:00:19,290 --> 00:00:25,980
‫But when you really understand is you feel pity for the computer that does the job for you because.

6
00:00:26,370 --> 00:00:27,300
‫Oh, my God.

7
00:00:27,420 --> 00:00:27,800
‫Right.

8
00:00:27,840 --> 00:00:29,220
‫You appreciate it.

9
00:00:29,310 --> 00:00:29,880
‫Right?

10
00:00:29,880 --> 00:00:33,960
‫When you start washing the dishes, you know how hard that job is, right?

11
00:00:34,380 --> 00:00:39,900
‫Thinking dishes and especially, if you like, cooked an egg and you forgot the plate for a day and

12
00:00:39,900 --> 00:00:46,380
‫then you come the next day and and those that yellow yolk thingy, I don't know what they call them.

13
00:00:46,380 --> 00:00:47,370
‫English, right.

14
00:00:47,490 --> 00:00:52,210
‫The yellow thing, they get stuck in the plate and you need to work really hard then.

15
00:00:52,420 --> 00:00:55,650
‫And you understand how dishwashing is really hard.

16
00:00:55,890 --> 00:00:56,940
‫Same thing with a computer.

17
00:00:56,940 --> 00:01:00,530
‫A computer is not dishwashing per say, but kind of similar.

18
00:01:01,320 --> 00:01:02,190
‫I'm rambling.

19
00:01:02,190 --> 00:01:04,460
‫I know you guys.

20
00:01:04,650 --> 00:01:13,530
‫The idea here is to understand how databases really store information on disk.

21
00:01:14,040 --> 00:01:14,660
‫All right.

22
00:01:14,670 --> 00:01:18,990
‫So it is extremely complicated thing.

23
00:01:19,590 --> 00:01:22,350
‫And let's talk about a little bit this.

24
00:01:22,560 --> 00:01:25,410
‫I'm not going to be adding graphics or anything.

25
00:01:25,410 --> 00:01:27,360
‫Just imagine what you mean, guys.

26
00:01:27,600 --> 00:01:30,780
‫See, follow my hands, because that's what I'm going to use here.

27
00:01:30,930 --> 00:01:37,220
‫I'm going to make another video detailing went with virtual board and all that jazz.

28
00:01:37,230 --> 00:01:39,170
‫But here's the thing, guys.

29
00:01:39,420 --> 00:01:42,410
‫Imagine you have a beautiful table.

30
00:01:42,430 --> 00:01:43,260
‫Has to be beautiful.

31
00:01:43,260 --> 00:01:45,840
‫If it's not a beautiful, then you have a problem.

32
00:01:45,960 --> 00:01:54,630
‫But if it's a table and let's say for simplicity, there is only one field and that field is an integer.

33
00:01:55,560 --> 00:02:03,930
‫Thirty two bit, let's say, right there, that's what I'd bet a bet, which is for whites, not not

34
00:02:03,960 --> 00:02:04,280
‫eight.

35
00:02:04,620 --> 00:02:06,030
‫OK, so.

36
00:02:07,260 --> 00:02:15,360
‫You have an integer right field, so and that's four bytes, so if I insert one rule, any number.

37
00:02:16,370 --> 00:02:21,790
‫Between one and two billion, I guess that's the maximum 32 bit injury you can have, right?

38
00:02:22,070 --> 00:02:27,350
‫It's going to take four bytes of space and we're going to talk about what space really means.

39
00:02:27,770 --> 00:02:34,530
‫So if I insert over here four bytes, it's another four bytes, another four bytes, another four.

40
00:02:34,570 --> 00:02:39,860
‫But how does this really is stored in disk?

41
00:02:39,860 --> 00:02:40,270
‫Right.

42
00:02:40,910 --> 00:02:43,670
‫And I'm going to explain this in two fashion.

43
00:02:44,180 --> 00:02:55,510
‫How is stored in SSD and how is stored in HDD hard disk drive, which is these rotating thingee?

44
00:02:56,270 --> 00:02:56,660
‫All right.

45
00:02:56,660 --> 00:02:58,190
‫How do you like this sound effect guys on.

46
00:02:58,250 --> 00:03:02,810
‫So, yeah, so we have a spinning disk and this is the and there is like a little bit different how

47
00:03:02,810 --> 00:03:04,210
‫they stored information.

48
00:03:04,220 --> 00:03:12,320
‫So let's start with the hard disk drive, which is a spinning disk, the hard disk drive that it is

49
00:03:12,320 --> 00:03:15,080
‫obviously a platter, which is a circle.

50
00:03:15,080 --> 00:03:15,500
‫Right.

51
00:03:15,740 --> 00:03:20,570
‫And there is sectors, pork and each sector.

52
00:03:21,680 --> 00:03:30,890
‫Present information, and each sector has a little bit of luck in it, and these are tagged and identified

53
00:03:30,890 --> 00:03:34,770
‫uniquely for every single sector that is in the hard drive.

54
00:03:34,820 --> 00:03:42,470
‫So you would say, OK, Blätter, number one, track number seven, sector number five and block number

55
00:03:42,560 --> 00:03:43,040
‫two.

56
00:03:43,370 --> 00:03:51,560
‫And that gives you exactly a block, which is a block of disk that corresponds to a number of bytes,

57
00:03:52,010 --> 00:03:53,630
‫let's say five hundred and twelve bytes.

58
00:03:53,760 --> 00:03:55,700
‫OK, so.

59
00:03:56,670 --> 00:04:02,790
‫Go back to our logical representation off the table, if I insert four by this integer, like number

60
00:04:02,790 --> 00:04:04,160
‫seven, which is four by four, right.

61
00:04:04,560 --> 00:04:05,940
‫That's the beautiful thing here.

62
00:04:06,240 --> 00:04:09,000
‫If you're number seven, you're going to reserve four bytes.

63
00:04:09,600 --> 00:04:14,340
‫If you're insert number two million, you're going to reserve for it doesn't matter, right.

64
00:04:14,430 --> 00:04:16,740
‫Because it's old zero zero one zero zero zero zero zero.

65
00:04:16,770 --> 00:04:18,120
‫It's all four bytes.

66
00:04:18,390 --> 00:04:24,780
‫And that four bytes will be stored in that given block.

67
00:04:25,300 --> 00:04:27,060
‫OK, and.

68
00:04:28,300 --> 00:04:34,780
‫Codes, that block is 512 and you start for four bytes, that kind of waste, right, you cannot jump

69
00:04:34,780 --> 00:04:37,210
‫to another block and start another four bytes.

70
00:04:37,660 --> 00:04:38,060
‫Right.

71
00:04:38,320 --> 00:04:39,880
‫You have to insert another row.

72
00:04:40,150 --> 00:04:42,970
‫And here's what the data is smart.

73
00:04:43,270 --> 00:04:51,010
‫The database says, wait a second, this table is reserving this block so far and it's going to put

74
00:04:51,010 --> 00:04:55,840
‫the next four by the next integer next to that four byte.

75
00:04:55,850 --> 00:05:01,300
‫So it's going to be adjacent to it so they can fit in the same block and you keep inserting something

76
00:05:01,300 --> 00:05:04,690
‫in something until that 512 is completely filled.

77
00:05:05,750 --> 00:05:14,570
‫Right, and you might say, wait a second, how do I know if I want to pull row number one to go to

78
00:05:14,570 --> 00:05:15,370
‫that block?

79
00:05:16,220 --> 00:05:20,570
‫That's how the database stores stuff that their base will say.

80
00:05:20,960 --> 00:05:23,320
‫OK, Roe, no.

81
00:05:23,990 --> 00:05:31,700
‫There is a unique identifier for the road and it could be that primary key if that field happened to

82
00:05:31,700 --> 00:05:32,350
‫be primary.

83
00:05:32,660 --> 00:05:36,830
‫So that particular number points.

84
00:05:36,830 --> 00:05:42,680
‫There is metadata in the database that says, hey, this I want, for example, row number one, wrong

85
00:05:42,680 --> 00:05:45,050
‫number one, have some information next to it.

86
00:05:45,320 --> 00:05:49,630
‫Sector seven, block eight, track three.

87
00:05:50,030 --> 00:05:55,520
‫And that information is what the database need to actually go to the disk.

88
00:05:55,520 --> 00:06:02,780
‫That's circle ugly mechanical drive and start spinning and reads.

89
00:06:03,050 --> 00:06:03,590
‫What is it?

90
00:06:03,590 --> 00:06:10,100
‫Read doesn't read four bytes, it doesn't read eight boards and reads the whole God damn block, which

91
00:06:10,100 --> 00:06:13,370
‫is 512 byte of memory.

92
00:06:13,370 --> 00:06:14,690
‫And you put it in memory.

93
00:06:15,290 --> 00:06:17,540
‫And then when we got that, that.

94
00:06:18,690 --> 00:06:25,200
‫Is slow because that's NATO, but the moment I have all this blook 512, I got a wrong number one.

95
00:06:26,350 --> 00:06:32,860
‫And I got no number two as well, by for free, I got to number three and four and five and six and

96
00:06:32,860 --> 00:06:38,970
‫seven until the last straw that is in the book until rule number 128, to be specific.

97
00:06:38,980 --> 00:06:39,280
‫Right.

98
00:06:39,700 --> 00:06:43,780
‫What if I want because 128 is the last block that he can fit.

99
00:06:43,780 --> 00:06:44,650
‫Five hundred and twelve.

100
00:06:45,490 --> 00:06:46,000
‫That's right.

101
00:06:46,270 --> 00:06:52,030
‫So now what if I want to roll 129, for example, which happened to be the value?

102
00:06:53,530 --> 00:06:55,710
‫One of one and a half million light.

103
00:06:57,030 --> 00:07:05,490
‫If you want that or it's not on the block of the block, it's on the block, that's another IIO and

104
00:07:05,500 --> 00:07:08,730
‫they the database knows this information is like, oh, another.

105
00:07:08,730 --> 00:07:11,250
‫I'll go to that location.

106
00:07:11,610 --> 00:07:15,750
‫That one twenty eight to 129 will say all black.

107
00:07:15,760 --> 00:07:19,100
‫Number seven now block number eight, it's the next to it.

108
00:07:19,260 --> 00:07:25,650
‫And then you pull that information back and you might say, are these two blocks nearer to each other.

109
00:07:25,920 --> 00:07:31,670
‫Could be cabinet based on how fragmented your data is, but usually they are next to each other.

110
00:07:31,830 --> 00:07:35,270
‫So that's how mechanical drive works in a nutshell.

111
00:07:35,280 --> 00:07:40,920
‫So how this is how they are stored on desk comes to beautiful.

112
00:07:40,930 --> 00:07:44,010
‫This is this is this these don't have rotating stuff.

113
00:07:44,010 --> 00:07:45,090
‫They don't have sectors.

114
00:07:45,210 --> 00:07:48,060
‫There is no tracks.

115
00:07:48,090 --> 00:07:49,890
‫There is no all the garbage.

116
00:07:50,310 --> 00:07:50,970
‫There is.

117
00:07:50,970 --> 00:07:55,140
‫However, the idea of blocks still and block have a size.

118
00:07:55,360 --> 00:07:55,770
‫All right.

119
00:07:56,640 --> 00:08:00,030
‫512, 1024 whatever.

120
00:08:00,030 --> 00:08:00,360
‫Right.

121
00:08:00,450 --> 00:08:03,540
‫And this blocks of memory have pages.

122
00:08:04,370 --> 00:08:09,410
‫And each page has a size, but it's the same thing, right, and instead the database.

123
00:08:10,490 --> 00:08:16,730
‫Go back to the our representation, logical representation instead of the database storing that track

124
00:08:16,730 --> 00:08:19,760
‫and sector where the actual stuff is for.

125
00:08:19,760 --> 00:08:23,450
‫This is the it just stored the block number and the page number.

126
00:08:23,480 --> 00:08:23,800
‫So.

127
00:08:23,800 --> 00:08:24,260
‫All right.

128
00:08:24,300 --> 00:08:30,430
‫Number one is block number 17 and page number eight.

129
00:08:30,950 --> 00:08:31,330
‫Right.

130
00:08:31,580 --> 00:08:39,470
‫And just pull that information when it pulled that page, let's say 512 bytes and that page, we have

131
00:08:39,470 --> 00:08:45,850
‫that beautiful page we got wrong number one and two and three and four and five and six right up until

132
00:08:45,850 --> 00:08:47,600
‫all 128, same thing.

133
00:08:47,600 --> 00:08:53,810
‫And we didn't talk about rights per say, but I'm going to mention it in that way after a while.

134
00:08:53,810 --> 00:08:54,950
‫But yeah.

135
00:08:54,950 --> 00:08:56,030
‫So that's the same thing.

136
00:08:56,030 --> 00:08:58,550
‫You go back and you read it exactly the same.

137
00:08:58,560 --> 00:09:05,390
‫However, it's 100, 200 to 400 times faster because it's a random access.

138
00:09:05,390 --> 00:09:05,830
‫Right.

139
00:09:05,930 --> 00:09:14,690
‫And instead of spinning wheel, it's it's literally you tell that I controller of us is the go to that

140
00:09:14,690 --> 00:09:18,110
‫particular block, that particular page and pull it.

141
00:09:18,110 --> 00:09:23,050
‫And when you pull a block, you usually pull all the pages with it.

142
00:09:23,060 --> 00:09:26,720
‫That so that's what happens in an SSD.

143
00:09:26,870 --> 00:09:29,860
‫When you pull a block, you pull all the pages with it as well.

144
00:09:30,710 --> 00:09:31,370
‫Let's go through.

145
00:09:31,380 --> 00:09:32,060
‫All right.

146
00:09:32,630 --> 00:09:35,120
‫I want to write to.

147
00:09:36,530 --> 00:09:43,280
‫Paige, we never write to pages, right, as users, we write to Rose is like, oh, I'm going to update

148
00:09:43,280 --> 00:09:48,870
‫this rule number nine to be four instead of one point five million to be two million.

149
00:09:49,550 --> 00:09:50,900
‫So what what do we do here?

150
00:09:50,910 --> 00:09:51,680
‫What does that mean?

151
00:09:52,400 --> 00:09:55,760
‫That translates first of all, that's rule number nine.

152
00:09:55,940 --> 00:09:57,300
‫So pull that role.

153
00:09:57,350 --> 00:09:58,220
‫What does it mean?

154
00:09:58,220 --> 00:09:58,850
‫Where is wrong?

155
00:09:58,850 --> 00:10:07,160
‫Number nine or number nine in the database there is metadata says order number is block number 302 and

156
00:10:07,160 --> 00:10:08,570
‫then pull that.

157
00:10:08,690 --> 00:10:09,530
‫That's the first thing.

158
00:10:09,800 --> 00:10:11,030
‫And then I'm going to update it.

159
00:10:11,030 --> 00:10:12,500
‫And it's one point five million.

160
00:10:13,400 --> 00:10:18,100
‫Update that for White to say two million updated in memory.

161
00:10:18,350 --> 00:10:20,660
‫Now I'm going to write it back.

162
00:10:21,020 --> 00:10:22,760
‫And here's how SSD works.

163
00:10:23,180 --> 00:10:30,480
‫I'm going to talk about how this is the works and I'm going to talk about how the sector works, the

164
00:10:30,500 --> 00:10:31,460
‫mechanical drive.

165
00:10:31,700 --> 00:10:39,920
‫So for the SSD, when you want to write that block, first of all, you write only the pages in the

166
00:10:39,920 --> 00:10:40,720
‫block, right.

167
00:10:40,880 --> 00:10:48,110
‫So that change that one point five million existing in a page number one.

168
00:10:48,110 --> 00:10:48,520
‫Right.

169
00:10:49,220 --> 00:10:51,980
‫But if you write it, you change it to a two million.

170
00:10:51,990 --> 00:10:54,710
‫You have to always create a new page.

171
00:10:54,830 --> 00:10:56,050
‫That's how SSD works.

172
00:10:56,270 --> 00:11:05,560
‫So you write a new page and then you flush back that block as a different page to describe.

173
00:11:06,140 --> 00:11:11,450
‫And this information, reading and writing to the same location that block.

174
00:11:11,450 --> 00:11:16,730
‫And as this idea is extremely Hertz, the SSD feeling.

175
00:11:16,760 --> 00:11:17,120
‫Right.

176
00:11:17,520 --> 00:11:22,310
‫This is these have shelf-life and this block, this block writing and reading.

177
00:11:22,310 --> 00:11:26,390
‫The rewriting, to be specific, has some shelf life.

178
00:11:26,390 --> 00:11:30,620
‫And this property and SSD is called endurance, literally.

179
00:11:30,620 --> 00:11:37,730
‫How many times you can write to the same block before that block is dead and at that block is bed.

180
00:11:37,940 --> 00:11:39,250
‫We cannot reuse it.

181
00:11:39,440 --> 00:11:40,340
‫That means.

182
00:11:41,430 --> 00:11:50,250
‫We just lost space in our society can take that with the hard disk drive, which is the physical thing,

183
00:11:51,130 --> 00:11:55,470
‫you can literally go back and ride to the same location.

184
00:11:55,510 --> 00:11:55,720
‫Right.

185
00:11:55,890 --> 00:12:02,130
‫And you can write to it along time because it's just magnetic right idea.

186
00:12:02,140 --> 00:12:06,750
‫So you can straight right to it without any problem, because when you write to it, you're going to

187
00:12:06,750 --> 00:12:08,060
‫write an entire block back.

188
00:12:08,070 --> 00:12:09,330
‫You cannot just write one bytes.

189
00:12:09,660 --> 00:12:11,850
‫That's that's not how IO controls work.

190
00:12:12,240 --> 00:12:12,600
‫So.

191
00:12:13,530 --> 00:12:18,630
‫So when you do that, when you change the last place, you take that whole block and just flush it back

192
00:12:18,630 --> 00:12:24,890
‫to disk and SSD, there is a limit to how much how much time you can ride to it.

193
00:12:25,080 --> 00:12:30,390
‫But for the hard disk drive, there is no limit to how many times you can write.

194
00:12:30,390 --> 00:12:31,490
‫You can write to the same block.

195
00:12:31,510 --> 00:12:33,000
‫It's OK.

196
00:12:33,210 --> 00:12:38,820
‫Sometimes sectors get corrupt, but that's rarely happens compared to is is the right.

197
00:12:38,970 --> 00:12:42,300
‫So SSD faster reading and writing.

198
00:12:43,490 --> 00:12:50,120
‫But have little short shelf life depends on their endurance, so this is the power, that hard disk

199
00:12:50,120 --> 00:12:50,630
‫drive.

200
00:12:50,660 --> 00:12:52,850
‫They are a little bit slower.

201
00:12:52,880 --> 00:12:58,700
‫They are slow because they are literally abiding by the laws of physics.

202
00:12:58,700 --> 00:12:59,080
‫Right.

203
00:12:59,550 --> 00:13:01,040
‫That's not electro magnetic.

204
00:13:01,040 --> 00:13:04,280
‫It's not magnetic and light and electricity.

205
00:13:04,280 --> 00:13:04,550
‫Right.

206
00:13:04,550 --> 00:13:11,180
‫Is just using literally physical stuff to sneak back to a location and then literally.

207
00:13:11,180 --> 00:13:11,600
‫Right.

208
00:13:11,600 --> 00:13:15,470
‫All these lights, one zero zero zero one zero zero writes all this information to disk.

209
00:13:15,680 --> 00:13:16,050
‫Right.

210
00:13:16,670 --> 00:13:26,240
‫That was a little bit of a brief introduction about how databases store information to disguise how

211
00:13:26,240 --> 00:13:28,280
‫disk drives are different than SSD.

212
00:13:28,440 --> 00:13:31,940
‫You really sometimes need to understand how these internal works.

213
00:13:32,120 --> 00:13:42,530
‫There are papers and people taking CDs just discussing how can we squeeze as much performance from SSD

214
00:13:42,530 --> 00:13:44,000
‫because those guys are the future.

215
00:13:44,000 --> 00:13:46,070
‫Obviously, those those guys are not dead.

216
00:13:46,080 --> 00:13:47,720
‫The mechanical drive we're still using.

217
00:13:47,720 --> 00:13:49,550
‫But these are the performance stuff.

218
00:13:49,550 --> 00:13:57,860
‫But we have noticed with use the data structures that we have built BTD, to be specific.

219
00:13:58,190 --> 00:14:02,090
‫They kill those systems if they are not configured properly.

220
00:14:02,270 --> 00:14:06,710
‫Because what do Beatriz's do better is when you keep inserting stuff.

221
00:14:07,340 --> 00:14:07,640
‫Yeah.

222
00:14:07,640 --> 00:14:11,360
‫And such thing as the love new inserts.

223
00:14:11,540 --> 00:14:16,010
‫If you're inserting new stuff, that's the best thing you can do, because in search of new stuff,

224
00:14:16,010 --> 00:14:21,020
‫you're always taking new blocks and you're writing to new blocks.

225
00:14:21,230 --> 00:14:24,140
‫You're not overriding existing block.

226
00:14:24,140 --> 00:14:28,790
‫So you're not reading something changing and manipulating and writing to the same block.

227
00:14:29,980 --> 00:14:35,350
‫These don't like that because you can only change it so much when I say to them so much, I'm talking

228
00:14:35,350 --> 00:14:41,940
‫about 500, 700, and that's how it increases the cost go up as well.

229
00:14:41,960 --> 00:14:43,900
‫Ride the endurance of that as deep.

230
00:14:44,140 --> 00:14:44,480
‫Right.

231
00:14:44,500 --> 00:14:48,040
‫So Beatriz's when we designed this thing, we didn't we didn't.

232
00:14:48,040 --> 00:14:52,120
‫We designed a way back in the 60s and 70s.

233
00:14:52,340 --> 00:14:53,620
‫I wasn't born back then.

234
00:14:53,620 --> 00:14:59,110
‫But those data structures didn't really think about insurgences.

235
00:14:59,140 --> 00:15:01,360
‫Update is like, OK, we'll just insert stuff.

236
00:15:01,360 --> 00:15:01,630
‫Right.

237
00:15:02,500 --> 00:15:10,150
‫And here's where the problem of misery Belltrees and again, if it's configured incorrectly, if you

238
00:15:10,150 --> 00:15:12,130
‫keep inserting in your tables.

239
00:15:13,100 --> 00:15:20,540
‫You have a B3 three index, and that index has to be large first, not just one into primer keys could

240
00:15:20,540 --> 00:15:23,260
‫be composed of strings and all that stuff.

241
00:15:23,780 --> 00:15:30,760
‫Then as you insert the B tree, that three will try to grow.

242
00:15:30,770 --> 00:15:31,150
‫Right.

243
00:15:31,430 --> 00:15:35,240
‫And when it tries to win, it's about to get imbalanced.

244
00:15:36,560 --> 00:15:41,300
‫The tree will shuffle itself and rebalance itself, and that's that where it kills.

245
00:15:41,330 --> 00:15:44,870
‫This is the when you rebalance, what does it mean?

246
00:15:44,900 --> 00:15:47,000
‫Yes, you are inserting into your logical table.

247
00:15:47,000 --> 00:15:48,400
‫You're just appending data.

248
00:15:48,710 --> 00:15:52,070
‫But the index is restructuring itself, right?

249
00:15:52,270 --> 00:15:52,970
‫Was this.

250
00:15:54,090 --> 00:15:55,470
‫Or is this right?

251
00:15:55,710 --> 00:16:01,880
‫So, yeah, so the the index is updating itself, oh, this is no longer the root.

252
00:16:01,890 --> 00:16:03,350
‫No, this is no, no.

253
00:16:03,390 --> 00:16:06,660
‫Oh, the child or the child have too many nodes update itself.

254
00:16:06,840 --> 00:16:09,420
‫So now it's this is the location it's born to do.

255
00:16:09,660 --> 00:16:17,510
‫So we rebounds and seven rebalancing is nothing but updates when you update in place in SSD.

256
00:16:17,940 --> 00:16:18,870
‫Oh my God.

257
00:16:19,020 --> 00:16:20,610
‫That is the worst thing you can do.

258
00:16:20,640 --> 00:16:20,930
‫Right.

259
00:16:21,030 --> 00:16:23,940
‫And you do this multiple times and uses these gone.

260
00:16:24,240 --> 00:16:29,370
‫And when I say elgan, yeah, that block is gone and that block is gone.

261
00:16:29,700 --> 00:16:37,470
‫Your disk essentially gets smaller because that it can no longer use that block needs to use a new block

262
00:16:37,470 --> 00:16:42,160
‫in memory and memory and desk and that can slow things down guys.

263
00:16:42,270 --> 00:16:49,440
‫So that's the idea of SSD, hard disk drive and how database is storing information there.

264
00:16:49,590 --> 00:16:51,510
‫People still researching this stuff.

265
00:16:51,720 --> 00:16:59,880
‫So they came up with another data structure, Facebook primarily, I believe they didn't come up with

266
00:16:59,880 --> 00:17:00,780
‫this data structure.

267
00:17:01,020 --> 00:17:03,150
‫They said B threes are May.

268
00:17:03,270 --> 00:17:04,380
‫Well, this is D, right?

269
00:17:04,620 --> 00:17:06,540
‫So they came up with their own database engine.

270
00:17:06,540 --> 00:17:08,340
‫They called it Dirac's DB, OK?

271
00:17:08,520 --> 00:17:14,280
‫And they said we have a lower use case where we're writing stuff all the time.

272
00:17:14,400 --> 00:17:19,690
‫And those three balancing batteries are killing our asses, these shelf-life and we're throwing this

273
00:17:19,690 --> 00:17:23,550
‫is d like we don't like there's no tomorrow, so what do we do?

274
00:17:23,760 --> 00:17:32,820
‫So they used an existing data structure called log structure merge three or l e emem right.

275
00:17:33,480 --> 00:17:38,610
‫LSM trees are append only they are great for appends.

276
00:17:38,620 --> 00:17:38,910
‫Right.

277
00:17:38,910 --> 00:17:42,750
‫So they it's, it's, it works and as levels right.

278
00:17:42,990 --> 00:17:48,120
‫I'm not going to go into details but that's essentially then it's also a few levels.

279
00:17:48,130 --> 00:17:52,530
‫It works at a memory level and then level zero, level one, level three and level four.

280
00:17:52,710 --> 00:17:55,380
‫And these levels are always inserts.

281
00:17:55,380 --> 00:18:02,610
‫So the index, even if you build an L symmetry index base, then it always just inserts at the end.

282
00:18:02,610 --> 00:18:04,920
‫And that's really nice for us.

283
00:18:05,130 --> 00:18:10,170
‫Is this the love that obviously you have to have a larger as D2 if you inserting a lot of information.

284
00:18:10,170 --> 00:18:17,640
‫But guys, that was a but a long video talking about how database is store information into to the desk.

285
00:18:17,640 --> 00:18:18,060
‫Right.

286
00:18:18,720 --> 00:18:24,710
‫I like to talk without graphics and stuff like that because I might use this as a podcast and check

287
00:18:24,720 --> 00:18:25,350
‫my podcast.

288
00:18:25,350 --> 00:18:30,330
‫Guys, if you're interested, if you want to listen to this while you're in a train or background or

289
00:18:30,330 --> 00:18:35,460
‫you're working out, you can listen to this information without actually looking at and expecting something

290
00:18:36,450 --> 00:18:37,800
‫graphical on the videos.

291
00:18:37,800 --> 00:18:44,940
‫I do make other videos where I use a virtual board, but sometimes I make this more casual, chatty

292
00:18:45,300 --> 00:18:46,820
‫kind of videos as well.

293
00:18:46,950 --> 00:18:49,280
‫Alright, guys like this video is like it.

294
00:18:49,470 --> 00:18:50,100
‫This like it.

295
00:18:50,100 --> 00:18:53,310
‫If you don't like it, I'm going to see you on the next one.

296
00:18:53,310 --> 00:18:55,350
‫You guys stay awesome.