1
00:00:00,000 --> 00:00:08,469
foreign

2
00:00:00,500 --> 00:00:08,469
[Music]

3
00:00:11,219 --> 00:00:18,060
welcome back everybody now this morning

4
00:00:13,980 --> 00:00:20,240
we have got Andy here and he started

5
00:00:18,060 --> 00:00:22,920
hacking as a teenager with micro

6
00:00:20,240 --> 00:00:24,539
microprocessors were first available you

7
00:00:22,920 --> 00:00:26,820
had to build your own personal computer

8
00:00:24,539 --> 00:00:29,820
which was Quite a feat back then

9
00:00:26,820 --> 00:00:31,260
since the start of 2016 Andy's developed

10
00:00:29,820 --> 00:00:34,260
machine learning applications for

11
00:00:31,260 --> 00:00:36,480
Robotics and drones today they'll be

12
00:00:34,260 --> 00:00:38,760
talking about an open source distributed

13
00:00:36,480 --> 00:00:41,460
embedded framework that consolidates

14
00:00:38,760 --> 00:00:44,340
artificial intelligence robotics The

15
00:00:41,460 --> 00:00:46,440
Internet of Things media streaming and

16
00:00:44,340 --> 00:00:48,860
machine learning pipelines into a single

17
00:00:46,440 --> 00:00:50,850
cohesive platform

18
00:00:48,860 --> 00:00:52,820
thank you

19
00:00:50,850 --> 00:00:55,199
[Applause]

20
00:00:52,820 --> 00:00:56,760
hello everyone it's really fantastic to

21
00:00:55,199 --> 00:00:58,440
see everyone sort of it's a bit dark

22
00:00:56,760 --> 00:01:00,020
although I really wish I was sitting

23
00:00:58,440 --> 00:01:02,280
with you not not up here

24
00:01:00,020 --> 00:01:03,780
listening to the title author and I

25
00:01:02,280 --> 00:01:06,060
thought gosh that sounds a bit arrogant

26
00:01:03,780 --> 00:01:09,060
and ambitious but um well having having

27
00:01:06,060 --> 00:01:11,460
a good work at it I think my slides are

28
00:01:09,060 --> 00:01:13,500
probably a little bit text dense usual

29
00:01:11,460 --> 00:01:15,960
engineering fault um that URL at the

30
00:01:13,500 --> 00:01:18,600
bottom in yellow that's the slide so if

31
00:01:15,960 --> 00:01:21,479
you have a mobile device now jump on

32
00:01:18,600 --> 00:01:23,400
jump on that Tony URL and uh you might

33
00:01:21,479 --> 00:01:25,560
be able to see the sides close up

34
00:01:23,400 --> 00:01:27,900
so I've been working on this project

35
00:01:25,560 --> 00:01:31,560
with several people for a long long time

36
00:01:27,900 --> 00:01:33,840
now and one of the aspirations was to uh

37
00:01:31,560 --> 00:01:35,460
be able to have it used commercially

38
00:01:33,840 --> 00:01:37,259
because it's so hard to do

39
00:01:35,460 --> 00:01:39,540
um you know open source development you

40
00:01:37,259 --> 00:01:41,640
know just in nights and weekends and so

41
00:01:39,540 --> 00:01:44,340
um well the good news is uh last Friday

42
00:01:41,640 --> 00:01:46,860
the company I worked for soap on agreed

43
00:01:44,340 --> 00:01:48,119
to a big open source so you know if you

44
00:01:46,860 --> 00:01:49,979
do like what you're here today you'll be

45
00:01:48,119 --> 00:01:52,439
able to have a bit of a play

46
00:01:49,979 --> 00:01:54,119
but before I start should um welcome to

47
00:01:52,439 --> 00:01:56,700
the country I begin today by

48
00:01:54,119 --> 00:01:58,740
acknowledging the jury Warren people of

49
00:01:56,700 --> 00:02:00,180
the cooling Nation the traditional

50
00:01:58,740 --> 00:02:02,460
custodians of land on which We Gather

51
00:02:00,180 --> 00:02:04,320
today I pay my respects to their Elders

52
00:02:02,460 --> 00:02:05,759
past and present I extend their respect

53
00:02:04,320 --> 00:02:08,600
to Aboriginal and Torres Strait Islander

54
00:02:05,759 --> 00:02:08,600
pupils here today

55
00:02:09,360 --> 00:02:14,819
so uh just uh taking Hughes lead

56
00:02:12,239 --> 00:02:16,920
um uh Back in the Day live playing a bit

57
00:02:14,819 --> 00:02:18,900
Micro Control microprocessors and still

58
00:02:16,920 --> 00:02:21,959
do now so this is one of my background

59
00:02:18,900 --> 00:02:24,239
projects it's a 6502

60
00:02:21,959 --> 00:02:25,800
um but rather than an original

61
00:02:24,239 --> 00:02:27,900
um chip that you can get these days and

62
00:02:25,800 --> 00:02:29,760
get 40 megahertz chips um you know throw

63
00:02:27,900 --> 00:02:31,560
a lot of ram at it up on the top left

64
00:02:29,760 --> 00:02:34,379
there's a little digital oscilloscope so

65
00:02:31,560 --> 00:02:37,379
I can look at the clocks and um and also

66
00:02:34,379 --> 00:02:39,900
using a Arduino from from John that uh

67
00:02:37,379 --> 00:02:43,500
basically can load the Ram with programs

68
00:02:39,900 --> 00:02:45,780
I can move the clock from from zero up

69
00:02:43,500 --> 00:02:47,519
to a few megahertz and also monitor the

70
00:02:45,780 --> 00:02:49,019
address bus so just so much easier today

71
00:02:47,519 --> 00:02:51,300
you know to play with this stuff

72
00:02:49,019 --> 00:02:52,980
compared to where it used to be and uh

73
00:02:51,300 --> 00:02:54,660
yeah and we're on C compile it's got a

74
00:02:52,980 --> 00:02:56,280
little lisp interpreter it's a lot of

75
00:02:54,660 --> 00:02:58,620
fun

76
00:02:56,280 --> 00:03:00,060
and uh and this is how I guess many of

77
00:02:58,620 --> 00:03:02,099
us in the room room started you could

78
00:03:00,060 --> 00:03:03,959
really um it's very tangible you could

79
00:03:02,099 --> 00:03:05,400
um you write code and also play with the

80
00:03:03,959 --> 00:03:07,080
hardware and uh I think these days

81
00:03:05,400 --> 00:03:09,780
there's many layers of attraction that

82
00:03:07,080 --> 00:03:11,400
sort of keep us away from that

83
00:03:09,780 --> 00:03:13,560
um mother Mother's or hobby is uh

84
00:03:11,400 --> 00:03:16,260
playing playing the Astra and uh and

85
00:03:13,560 --> 00:03:17,580
dogs sorry robot dogs

86
00:03:16,260 --> 00:03:19,019
um so this is very much a work in

87
00:03:17,580 --> 00:03:20,760
progress but heading heading towards

88
00:03:19,019 --> 00:03:23,760
these things to bring this Confluence of

89
00:03:20,760 --> 00:03:26,640
um our iot devices plus robotics plus

90
00:03:23,760 --> 00:03:28,260
plus video nml so um just yeah just to

91
00:03:26,640 --> 00:03:29,580
play games at home and also actually

92
00:03:28,260 --> 00:03:31,560
hopefully do something valuable in the

93
00:03:29,580 --> 00:03:34,980
world

94
00:03:31,560 --> 00:03:36,360
okay so here's this morning to um show

95
00:03:34,980 --> 00:03:38,099
off this piece of work called Ico

96
00:03:36,360 --> 00:03:40,379
services this framework

97
00:03:38,099 --> 00:03:42,480
um I I did cheat and use chat CPT just

98
00:03:40,379 --> 00:03:44,340
just to pick a three-word title for this

99
00:03:42,480 --> 00:03:45,659
slide and that was that was it

100
00:03:44,340 --> 00:03:47,159
um but but I've got to keep up with the

101
00:03:45,659 --> 00:03:49,220
cool kids

102
00:03:47,159 --> 00:03:49,220
um

103
00:03:50,760 --> 00:03:53,940
um I work I work at Silver Pond which is

104
00:03:52,560 --> 00:03:55,680
fantastic place

105
00:03:53,940 --> 00:03:58,140
um the business team and the engineer

106
00:03:55,680 --> 00:04:00,780
Engineers are great we basically make

107
00:03:58,140 --> 00:04:02,760
machine learning products and I'm a

108
00:04:00,780 --> 00:04:04,440
systems engineer and I basically do the

109
00:04:02,760 --> 00:04:08,280
infrastructure of support and those sort

110
00:04:04,440 --> 00:04:10,580
of things and as often as I can hang out

111
00:04:08,280 --> 00:04:12,659
with friends at the hackerspace and

112
00:04:10,580 --> 00:04:14,519
basically play of devices and robots

113
00:04:12,659 --> 00:04:17,639
there in my spare time

114
00:04:14,519 --> 00:04:18,900
uh so today we'll trust this embedded

115
00:04:17,639 --> 00:04:21,540
framework I'm trying to give you some

116
00:04:18,900 --> 00:04:23,280
context I was thinking at first I'll

117
00:04:21,540 --> 00:04:24,600
just you know just big lead up and the

118
00:04:23,280 --> 00:04:26,340
big demonstration at the end I thought

119
00:04:24,600 --> 00:04:27,900
no that'd be the worst why does give you

120
00:04:26,340 --> 00:04:29,820
the give the demonstration the start and

121
00:04:27,900 --> 00:04:31,860
if you like it you can stay

122
00:04:29,820 --> 00:04:33,120
um we're gonna Crash and Burn early

123
00:04:31,860 --> 00:04:35,820
um try and cover what this framework

124
00:04:33,120 --> 00:04:37,620
might be good for how it differs try and

125
00:04:35,820 --> 00:04:40,139
give some technical details but not no

126
00:04:37,620 --> 00:04:41,220
not try to boil your death

127
00:04:40,139 --> 00:04:43,259
um although I couldn't tell if you're

128
00:04:41,220 --> 00:04:45,240
sleeping or not and then I also just

129
00:04:43,259 --> 00:04:46,860
covered some of the aspects of uh of

130
00:04:45,240 --> 00:04:48,540
trying to get a Synergy or the best

131
00:04:46,860 --> 00:04:50,880
outcome of working both commercially and

132
00:04:48,540 --> 00:04:52,080
and as a community developer and then

133
00:04:50,880 --> 00:04:53,699
maybe discuss a little bit about what

134
00:04:52,080 --> 00:04:55,440
next

135
00:04:53,699 --> 00:04:58,139
Okay so

136
00:04:55,440 --> 00:05:00,720
you know it goes about saying it's been

137
00:04:58,139 --> 00:05:03,360
an extraordinary time over the last 60

138
00:05:00,720 --> 00:05:04,740
years 30 years 10 years

139
00:05:03,360 --> 00:05:06,720
um you know all the software could

140
00:05:04,740 --> 00:05:07,919
possibly want just there at a touch of a

141
00:05:06,720 --> 00:05:09,600
button

142
00:05:07,919 --> 00:05:10,740
um you know or and for those fortunate

143
00:05:09,600 --> 00:05:12,600
enough we'd say basically have you know

144
00:05:10,740 --> 00:05:14,400
we're connected all the time out of our

145
00:05:12,600 --> 00:05:15,600
devices in our home in our pockets and

146
00:05:14,400 --> 00:05:17,100
so on

147
00:05:15,600 --> 00:05:19,560
um the hardware we have now is amazing

148
00:05:17,100 --> 00:05:20,820
now just looking at the numbers

149
00:05:19,560 --> 00:05:22,919
um you know the transistor count so

150
00:05:20,820 --> 00:05:24,840
looking back to that first slide that

151
00:05:22,919 --> 00:05:25,800
microprocessors knows today's like 6502

152
00:05:24,840 --> 00:05:27,419
about three and a half thousand

153
00:05:25,800 --> 00:05:29,580
transistors

154
00:05:27,419 --> 00:05:32,400
um a clock rate of megahertz and uh yeah

155
00:05:29,580 --> 00:05:34,199
one core uh these days

156
00:05:32,400 --> 00:05:36,419
um the clock the transistor count is

157
00:05:34,199 --> 00:05:37,979
about six orders magnitude greater the

158
00:05:36,419 --> 00:05:39,600
clock three sort of actually greater and

159
00:05:37,979 --> 00:05:41,940
the the cores you know maybe a magnet

160
00:05:39,600 --> 00:05:44,639
about a thousand times fold so it comes

161
00:05:41,940 --> 00:05:47,039
to about 10 to the 12 increase in uh in

162
00:05:44,639 --> 00:05:50,100
I guess raw horsepower it's pretty crazy

163
00:05:47,039 --> 00:05:53,520
but that's that's what's enabling uh AI

164
00:05:50,100 --> 00:05:56,220
with its um insatiable need for our data

165
00:05:53,520 --> 00:05:58,380
and and compute and it goes about saying

166
00:05:56,220 --> 00:06:01,800
that last year was um pretty amazing in

167
00:05:58,380 --> 00:06:04,020
the way that just from a website pretty

168
00:06:01,800 --> 00:06:05,340
much anyone who's connected Can can play

169
00:06:04,020 --> 00:06:07,259
with these tools now and Achieve

170
00:06:05,340 --> 00:06:10,620
extraordinary results and uh just

171
00:06:07,259 --> 00:06:13,080
yesterday open AI I did a case a bit of

172
00:06:10,620 --> 00:06:15,600
a preview on gpt4 which might be out for

173
00:06:13,080 --> 00:06:16,740
a few months it sounds like but yeah

174
00:06:15,600 --> 00:06:18,300
there's a link there for anyone who's

175
00:06:16,740 --> 00:06:21,360
interested it's just um just a short

176
00:06:18,300 --> 00:06:22,860
demo by this the co-founder and

177
00:06:21,360 --> 00:06:25,800
basically they just lifted the game on

178
00:06:22,860 --> 00:06:30,539
most mostly what it does

179
00:06:25,800 --> 00:06:31,680
okay so you hear that um if uh if you

180
00:06:30,539 --> 00:06:33,120
sort of can describe the problem you're

181
00:06:31,680 --> 00:06:34,440
halfway to a solution

182
00:06:33,120 --> 00:06:36,360
um and sometimes it feels like that but

183
00:06:34,440 --> 00:06:38,340
other times just go no it still seems

184
00:06:36,360 --> 00:06:40,979
like a ton of work to do but anyone

185
00:06:38,340 --> 00:06:42,900
who's um Works any of these domains of

186
00:06:40,979 --> 00:06:44,100
innovative things I might work with

187
00:06:42,900 --> 00:06:45,539
media

188
00:06:44,100 --> 00:06:47,340
um yeah you're in there more space or

189
00:06:45,539 --> 00:06:48,660
robotics each of these things pretty

190
00:06:47,340 --> 00:06:50,100
much you know more than a career's worth

191
00:06:48,660 --> 00:06:52,860
and they will have different Frameworks

192
00:06:50,100 --> 00:06:55,500
and tooling and and it's um you know it

193
00:06:52,860 --> 00:06:56,460
can be quite a difficult you want to

194
00:06:55,500 --> 00:06:59,940
pull your hair out and try to integrate

195
00:06:56,460 --> 00:07:01,680
these Technologies and so um yeah the

196
00:06:59,940 --> 00:07:02,880
hardware is different the uh you know

197
00:07:01,680 --> 00:07:04,199
your concerns when doing media are

198
00:07:02,880 --> 00:07:06,600
different from when you're when you're

199
00:07:04,199 --> 00:07:07,979
doing machine learning and uh and and

200
00:07:06,600 --> 00:07:09,180
yet if you want to build these um some

201
00:07:07,979 --> 00:07:11,699
Modern applications you need to bring

202
00:07:09,180 --> 00:07:13,080
all these things together coherently and

203
00:07:11,699 --> 00:07:15,139
on top of that there's also just all the

204
00:07:13,080 --> 00:07:16,979
other concerns you have in terms of um

205
00:07:15,139 --> 00:07:18,240
failure and security

206
00:07:16,979 --> 00:07:19,199
um you know how to configure how do you

207
00:07:18,240 --> 00:07:20,220
persist

208
00:07:19,199 --> 00:07:22,319
um you know if you've got multiple

209
00:07:20,220 --> 00:07:23,880
programming languages so it's pretty

210
00:07:22,319 --> 00:07:25,500
challenging to um you know to Google

211
00:07:23,880 --> 00:07:28,139
systems these days whether they're

212
00:07:25,500 --> 00:07:29,460
reasonably reliable robust secure and so

213
00:07:28,139 --> 00:07:31,800
on

214
00:07:29,460 --> 00:07:34,740
let's check to check

215
00:07:31,800 --> 00:07:36,599
cool all right so yeah so I'm just going

216
00:07:34,740 --> 00:07:39,419
to head off start off with the demo

217
00:07:36,599 --> 00:07:40,919
um uh I was very kindly allowed to to

218
00:07:39,419 --> 00:07:42,599
use this even though you know business

219
00:07:40,919 --> 00:07:44,220
people know are very very concerned

220
00:07:42,599 --> 00:07:45,780
about how an engineer presents stuff

221
00:07:44,220 --> 00:07:48,240
because you know we don't we don't mind

222
00:07:45,780 --> 00:07:50,099
things being a bit ratty and daggy and

223
00:07:48,240 --> 00:07:52,940
this is really just here to show off the

224
00:07:50,099 --> 00:07:55,860
The Open Source framework

225
00:07:52,940 --> 00:07:57,660
uh yeah so I said I worked for a local

226
00:07:55,860 --> 00:08:00,599
machine Learning Company

227
00:07:57,660 --> 00:08:02,639
um and we've done the classic pivot from

228
00:08:00,599 --> 00:08:04,979
Consulting over the years to to try and

229
00:08:02,639 --> 00:08:07,979
productize what we do and that's uh it

230
00:08:04,979 --> 00:08:09,900
can be quite challenging and and the

231
00:08:07,979 --> 00:08:11,460
actual product itself is much much more

232
00:08:09,900 --> 00:08:14,759
than uh the the piece of work I'm

233
00:08:11,460 --> 00:08:16,199
showing you today uh but uh but this is

234
00:08:14,759 --> 00:08:19,020
probably the best demonstration of what

235
00:08:16,199 --> 00:08:21,780
I can do in terms of um the uh you know

236
00:08:19,020 --> 00:08:24,479
using webrtc for video

237
00:08:21,780 --> 00:08:26,699
um capturing data from a device either a

238
00:08:24,479 --> 00:08:28,199
camera or in this case a little Ikea

239
00:08:26,699 --> 00:08:29,099
train

240
00:08:28,199 --> 00:08:31,259
um

241
00:08:29,099 --> 00:08:33,959
and uh yeah so I have to have to show

242
00:08:31,259 --> 00:08:35,459
and also when you um when you're doing

243
00:08:33,959 --> 00:08:37,440
commercial work you you're confronted

244
00:08:35,459 --> 00:08:39,599
with problems that um you know you may

245
00:08:37,440 --> 00:08:42,180
not get when just doing your hobbyist

246
00:08:39,599 --> 00:08:43,260
work uh because when you're um I know

247
00:08:42,180 --> 00:08:44,459
when I'm doing my own projects I

248
00:08:43,260 --> 00:08:45,779
particularly try and do the things I I

249
00:08:44,459 --> 00:08:47,459
know I think it will I'll be able to get

250
00:08:45,779 --> 00:08:49,320
away with you know basically tail of my

251
00:08:47,459 --> 00:08:51,860
my efforts to something I think will be

252
00:08:49,320 --> 00:08:54,060
doable where it's often cut clients or

253
00:08:51,860 --> 00:08:56,880
give you such a situation which you go

254
00:08:54,060 --> 00:08:58,019
oh man it's so hard to do something a

255
00:08:56,880 --> 00:08:59,459
little bit easier

256
00:08:58,019 --> 00:09:00,540
um but it is a good opportunity to

257
00:08:59,459 --> 00:09:03,060
really make sure your work is Battle

258
00:09:00,540 --> 00:09:04,920
tested and uh and I believe very

259
00:09:03,060 --> 00:09:07,200
strongly that um there's good synergies

260
00:09:04,920 --> 00:09:08,940
between commercial and community that um

261
00:09:07,200 --> 00:09:10,560
you know we just have to look around you

262
00:09:08,940 --> 00:09:13,019
today to know how the quality of people

263
00:09:10,560 --> 00:09:14,700
uh they're out there in in our community

264
00:09:13,019 --> 00:09:17,160
and around the world

265
00:09:14,700 --> 00:09:19,140
um and uh and also but but you also know

266
00:09:17,160 --> 00:09:21,120
we're always struggling for funds and uh

267
00:09:19,140 --> 00:09:24,360
and an opportunity of people finding

268
00:09:21,120 --> 00:09:27,380
work and so on so I think um they bring

269
00:09:24,360 --> 00:09:29,519
these two two different um I guess

270
00:09:27,380 --> 00:09:31,980
organizations you know the

271
00:09:29,519 --> 00:09:33,200
the community or commercial is really

272
00:09:31,980 --> 00:09:36,000
important

273
00:09:33,200 --> 00:09:38,279
so what I found is uh I've been working

274
00:09:36,000 --> 00:09:41,040
as a very very long time and actually

275
00:09:38,279 --> 00:09:42,660
conceptually so it's been used on nearly

276
00:09:41,040 --> 00:09:45,480
four projects now through my career just

277
00:09:42,660 --> 00:09:48,540
as I've gone from from job to job and uh

278
00:09:45,480 --> 00:09:49,440
and it's gradually got better and uh but

279
00:09:48,540 --> 00:09:51,959
one one thing that sort of happened

280
00:09:49,440 --> 00:09:53,339
during lockdown was um yeah we normally

281
00:09:51,959 --> 00:09:56,160
try and Achieve work-life balance you

282
00:09:53,339 --> 00:09:57,899
know be social and family and stay

283
00:09:56,160 --> 00:09:59,339
healthy and so on and just for

284
00:09:57,899 --> 00:10:01,580
enrollment at least for at least about a

285
00:09:59,339 --> 00:10:04,380
year and a half is like nup

286
00:10:01,580 --> 00:10:06,000
uh there's a very unfortunate situation

287
00:10:04,380 --> 00:10:07,500
but I was thinking to myself well if I

288
00:10:06,000 --> 00:10:08,940
can't get this project done now when I'm

289
00:10:07,500 --> 00:10:11,640
not allowed to leave the house

290
00:10:08,940 --> 00:10:13,740
and and that's the stuff I will I'll

291
00:10:11,640 --> 00:10:14,880
never get it done so um I thought you

292
00:10:13,740 --> 00:10:16,440
know this might take you know six months

293
00:10:14,880 --> 00:10:18,000
to a year and here we are three and a

294
00:10:16,440 --> 00:10:20,519
half years later and uh yeah it's still

295
00:10:18,000 --> 00:10:21,660
a lot to do but but it's true if it

296
00:10:20,519 --> 00:10:24,300
wasn't for

297
00:10:21,660 --> 00:10:25,560
um being locked away in my own home for

298
00:10:24,300 --> 00:10:26,760
for many months that would never

299
00:10:25,560 --> 00:10:29,279
happened

300
00:10:26,760 --> 00:10:31,680
okay so and then just to give some

301
00:10:29,279 --> 00:10:33,540
context uh this is a little bit of

302
00:10:31,680 --> 00:10:35,700
framework basically performs uh sort of

303
00:10:33,540 --> 00:10:36,839
the infrastructure for um a bigger

304
00:10:35,700 --> 00:10:38,220
product at work at least in terms of

305
00:10:36,839 --> 00:10:41,459
some of the back end and some of the

306
00:10:38,220 --> 00:10:43,620
distributed computing aspects and um and

307
00:10:41,459 --> 00:10:45,180
uh what this it's been often when you're

308
00:10:43,620 --> 00:10:47,220
um work in your organization have to do

309
00:10:45,180 --> 00:10:48,240
that thing today yeah you don't have two

310
00:10:47,220 --> 00:10:49,680
weeks to get it right you have to

311
00:10:48,240 --> 00:10:51,240
basically you know do something for a

312
00:10:49,680 --> 00:10:52,800
client that you know that you um have to

313
00:10:51,240 --> 00:10:54,600
immediately about give me all the

314
00:10:52,800 --> 00:10:57,120
thought it needs whereas um sometimes

315
00:10:54,600 --> 00:10:58,380
we've um you know without open source

316
00:10:57,120 --> 00:11:00,540
working here we can take you can take

317
00:10:58,380 --> 00:11:02,459
the time overnight or weekend to um

318
00:11:00,540 --> 00:11:03,800
to think through that design or

319
00:11:02,459 --> 00:11:05,820
basically

320
00:11:03,800 --> 00:11:07,079
improve some of that Tech debt and how

321
00:11:05,820 --> 00:11:09,240
it really sucks but you'll never have

322
00:11:07,079 --> 00:11:12,060
the priority to do on Monday

323
00:11:09,240 --> 00:11:14,760
okay so this is a bit of a conceptual

324
00:11:12,060 --> 00:11:16,560
view of the overall a product basically

325
00:11:14,760 --> 00:11:17,339
the idea is to take uh data from the

326
00:11:16,560 --> 00:11:20,519
field

327
00:11:17,339 --> 00:11:22,560
um it could be uh devices and things out

328
00:11:20,519 --> 00:11:24,240
you know basically maybe cameras in the

329
00:11:22,560 --> 00:11:26,100
city like looking after the road system

330
00:11:24,240 --> 00:11:29,279
or could be in rural basically looking

331
00:11:26,100 --> 00:11:31,800
after um cows surprisingly cockroaches I

332
00:11:29,279 --> 00:11:33,720
think but yeah but basically um where's

333
00:11:31,800 --> 00:11:36,060
your livestock or basically Resource

334
00:11:33,720 --> 00:11:37,560
Management all the oceans so one project

335
00:11:36,060 --> 00:11:39,120
we've worked on is basically fish fish

336
00:11:37,560 --> 00:11:41,160
counting to make sure that the state of

337
00:11:39,120 --> 00:11:43,260
the fish stock and helping marine

338
00:11:41,160 --> 00:11:45,200
biologists so that comes in as a series

339
00:11:43,260 --> 00:11:47,700
of Rich data and streams it could be um

340
00:11:45,200 --> 00:11:48,720
GPS location it could be lighter it

341
00:11:47,700 --> 00:11:50,579
could be video it could be hyper

342
00:11:48,720 --> 00:11:51,899
spectral and then you basically want to

343
00:11:50,579 --> 00:11:54,360
um send that through data center

344
00:11:51,899 --> 00:11:55,200
pipelines and one one of the advantages

345
00:11:54,360 --> 00:11:57,839
of

346
00:11:55,200 --> 00:11:59,339
um doing things the data center is uh is

347
00:11:57,839 --> 00:12:00,660
that you get the information from all of

348
00:11:59,339 --> 00:12:04,019
your devices in the field so basically

349
00:12:00,660 --> 00:12:05,459
get this more um of omnisone sort of

350
00:12:04,019 --> 00:12:06,779
overview and make decisions that you

351
00:12:05,459 --> 00:12:07,740
can't make as an individual device on

352
00:12:06,779 --> 00:12:10,380
the edge

353
00:12:07,740 --> 00:12:11,820
and then that goes through we basically

354
00:12:10,380 --> 00:12:13,920
integrate to our other people's

355
00:12:11,820 --> 00:12:17,700
Enterprise systems they can which they

356
00:12:13,920 --> 00:12:21,180
can uh make sense today themselves and

357
00:12:17,700 --> 00:12:21,839
also basically use ontology or taxonomy

358
00:12:21,180 --> 00:12:22,459
to

359
00:12:21,839 --> 00:12:25,019
um

360
00:12:22,459 --> 00:12:25,980
break down the uh I guess the

361
00:12:25,019 --> 00:12:27,540
relationships between the things you're

362
00:12:25,980 --> 00:12:29,160
seeing in the world so that's sort of

363
00:12:27,540 --> 00:12:31,200
what it's saying to do

364
00:12:29,160 --> 00:12:34,079
okay so what about shows um this is a

365
00:12:31,200 --> 00:12:36,680
bit more of a complex view the

366
00:12:34,079 --> 00:12:39,240
the blue bits are basically about webrtc

367
00:12:36,680 --> 00:12:41,640
webrtc is pretty amazing and basically

368
00:12:39,240 --> 00:12:43,200
allows you to have bi-directional audio

369
00:12:41,640 --> 00:12:44,480
video and data that gets through

370
00:12:43,200 --> 00:12:47,639
firewalls

371
00:12:44,480 --> 00:12:50,160
and anyone's using it as a result of

372
00:12:47,639 --> 00:12:54,180
covert on the earlier on Zoom meetings

373
00:12:50,160 --> 00:12:55,019
so so there's a a part of this race you

374
00:12:54,180 --> 00:12:56,820
have to set up the web artistic

375
00:12:55,019 --> 00:12:58,500
connection all the white bits they're

376
00:12:56,820 --> 00:13:00,360
all they're all services that are where

377
00:12:58,500 --> 00:13:02,459
we basically sit on this framework of

378
00:13:00,360 --> 00:13:04,800
sending messages to each other and the

379
00:13:02,459 --> 00:13:06,600
uh because the uh the green the green

380
00:13:04,800 --> 00:13:08,880
bits is just the underlying um mosquito

381
00:13:06,600 --> 00:13:10,380
so what we end up with is a a web

382
00:13:08,880 --> 00:13:13,320
browser is going to

383
00:13:10,380 --> 00:13:14,579
um over websockets talk to the back end

384
00:13:13,320 --> 00:13:15,720
to the registrar find out what sort of

385
00:13:14,579 --> 00:13:17,339
services they are they're particularly

386
00:13:15,720 --> 00:13:19,320
interest in the data sources like like

387
00:13:17,339 --> 00:13:21,180
the cam the camera from the train an

388
00:13:19,320 --> 00:13:23,279
rtsp camera in our office which I should

389
00:13:21,180 --> 00:13:26,040
have checked before I'm presenting today

390
00:13:23,279 --> 00:13:28,740
and also I'd like to see a camera on my

391
00:13:26,040 --> 00:13:30,480
laptop uh it's the selected data source

392
00:13:28,740 --> 00:13:32,700
which that allows the web browser to go

393
00:13:30,480 --> 00:13:35,040
to the signaling server uh to set up the

394
00:13:32,700 --> 00:13:36,959
webrtc connection uh once that's good at

395
00:13:35,040 --> 00:13:38,160
the web server can then talk sorry the

396
00:13:36,959 --> 00:13:40,560
web browsing and talk to the back end

397
00:13:38,160 --> 00:13:43,260
webrtc manager to then set up a session

398
00:13:40,560 --> 00:13:44,940
client and then once that's all good the

399
00:13:43,260 --> 00:13:48,120
data stream soon ends up going to the ml

400
00:13:44,940 --> 00:13:50,220
model then back to back to your laptop

401
00:13:48,120 --> 00:13:51,839
um photos is worth mentioning

402
00:13:50,220 --> 00:13:54,000
um oh and also interestingly enough um

403
00:13:51,839 --> 00:13:55,560
the it's what the web browser workers

404
00:13:54,000 --> 00:13:57,420
all in go

405
00:13:55,560 --> 00:13:58,980
um on top of azim which is very very

406
00:13:57,420 --> 00:14:00,959
cool so basically the same run the same

407
00:13:58,980 --> 00:14:03,360
code in the browser as the back end okay

408
00:14:00,959 --> 00:14:04,680
well let's see if I can bring this

409
00:14:03,360 --> 00:14:06,959
together

410
00:14:04,680 --> 00:14:09,300
so

411
00:14:06,959 --> 00:14:11,480
I'll just just refresh this

412
00:14:09,300 --> 00:14:14,639
so this is the web browser view

413
00:14:11,480 --> 00:14:17,639
what you'll notice is up in the top left

414
00:14:14,639 --> 00:14:19,980
is a number of services the the mock

415
00:14:17,639 --> 00:14:22,139
train the live office camera and face

416
00:14:19,980 --> 00:14:24,120
detection in the middle is the chart so

417
00:14:22,139 --> 00:14:26,060
having found the services so I'm just a

418
00:14:24,120 --> 00:14:29,700
quick look at one

419
00:14:26,060 --> 00:14:31,920
so using eventual consistency of over um

420
00:14:29,700 --> 00:14:33,779
effectively over mqt all the data about

421
00:14:31,920 --> 00:14:36,720
that train is coming coming back to here

422
00:14:33,779 --> 00:14:38,339
like it's location at speed and also it

423
00:14:36,720 --> 00:14:40,079
also provides UI hints it says look go

424
00:14:38,339 --> 00:14:42,660
and put the put my GPS coordinates on a

425
00:14:40,079 --> 00:14:44,279
map put my speed on a chart so just off

426
00:14:42,660 --> 00:14:46,560
the screen here there's some chart

427
00:14:44,279 --> 00:14:49,820
information for all those devices

428
00:14:46,560 --> 00:14:51,839
and just to show you on the back end

429
00:14:49,820 --> 00:14:52,920
you can see that's the actual devices

430
00:14:51,839 --> 00:14:54,180
they've got a little dashboard which

431
00:14:52,920 --> 00:14:55,500
I'll talk more about later and if we

432
00:14:54,180 --> 00:14:57,779
look at that train we can actually see

433
00:14:55,500 --> 00:14:59,220
the same GPS coordinates here in that

434
00:14:57,779 --> 00:15:01,440
middle View

435
00:14:59,220 --> 00:15:03,720
so so one of the goals the system is to

436
00:15:01,440 --> 00:15:05,040
make um complex systems be able to

437
00:15:03,720 --> 00:15:06,300
inspect any part of them and see what's

438
00:15:05,040 --> 00:15:08,699
going on

439
00:15:06,300 --> 00:15:11,120
so let's uh let's see if this works

440
00:15:08,699 --> 00:15:11,120
um right

441
00:15:11,519 --> 00:15:15,060
it's my presentation here's the real

442
00:15:13,380 --> 00:15:17,160
thing

443
00:15:15,060 --> 00:15:19,019
okay

444
00:15:17,160 --> 00:15:20,940
so this is basically taking a

445
00:15:19,019 --> 00:15:23,490
pre-recorded video file because um the

446
00:15:20,940 --> 00:15:26,559
Train the train is here today

447
00:15:23,490 --> 00:15:26,559
[Music]

448
00:15:27,959 --> 00:15:33,740
like what I can do is I can check the

449
00:15:30,000 --> 00:15:33,740
check the back end see why it's

450
00:15:35,519 --> 00:15:38,420
a refresh

451
00:15:38,940 --> 00:15:43,440
oh here we go yeah your team

452
00:15:41,220 --> 00:15:44,880
so that that's video that's being sent

453
00:15:43,440 --> 00:15:46,920
uh sent through the back end to the

454
00:15:44,880 --> 00:15:48,959
front end with webrtc

455
00:15:46,920 --> 00:15:51,720
um on normal stances so be going through

456
00:15:48,959 --> 00:15:54,720
the ml model and uh and if it checks uh

457
00:15:51,720 --> 00:15:57,000
signs or buildings or or um trees that

458
00:15:54,720 --> 00:16:00,120
let you know so that's sort of the first

459
00:15:57,000 --> 00:16:02,100
thing now the second thing I'd do is um

460
00:16:00,120 --> 00:16:03,360
and go up to here and I can send a

461
00:16:02,100 --> 00:16:05,160
command through to the train from the

462
00:16:03,360 --> 00:16:07,500
web browser just saying just saying stop

463
00:16:05,160 --> 00:16:09,000
and so we can see that we don't have

464
00:16:07,500 --> 00:16:10,920
video stop and so it's one of the trains

465
00:16:09,000 --> 00:16:12,959
and if we look at the back if we look at

466
00:16:10,920 --> 00:16:14,519
the back end we can see the the trains

467
00:16:12,959 --> 00:16:16,920
also stopped in its actual state has

468
00:16:14,519 --> 00:16:18,839
gone from from running to false

469
00:16:16,920 --> 00:16:19,500
there we go and if I wanted to I could

470
00:16:18,839 --> 00:16:23,100
um

471
00:16:19,500 --> 00:16:25,800
just from this dashboard I can go and uh

472
00:16:23,100 --> 00:16:27,540
they make that true

473
00:16:25,800 --> 00:16:30,060
and hopefully hip trains going again

474
00:16:27,540 --> 00:16:32,820
fantastic and soft up but the video

475
00:16:30,060 --> 00:16:34,339
isn't I think I know why I forgot to

476
00:16:32,820 --> 00:16:37,980
then

477
00:16:34,339 --> 00:16:40,259
use a slot user in Facebook that should

478
00:16:37,980 --> 00:16:41,759
that should start the video up again

479
00:16:40,259 --> 00:16:43,380
the second thing I'm going to do now

480
00:16:41,759 --> 00:16:45,480
I'll just have to do a refresh is to

481
00:16:43,380 --> 00:16:48,019
just see if I can show the the camera's

482
00:16:45,480 --> 00:16:48,019
live stream

483
00:16:50,220 --> 00:16:52,940
the office camera

484
00:16:54,959 --> 00:16:57,540
so this is in a building it's actually

485
00:16:56,339 --> 00:16:59,820
looking over the river just up the

486
00:16:57,540 --> 00:17:01,620
street from us and so you could almost

487
00:16:59,820 --> 00:17:04,140
but not quite see where we are there's

488
00:17:01,620 --> 00:17:06,559
the that's that bridge on the far right

489
00:17:04,140 --> 00:17:10,260
is just uh just across the river from us

490
00:17:06,559 --> 00:17:12,480
so that's an rtsp camera in office and

491
00:17:10,260 --> 00:17:13,679
I'm hoping to do the with the ml

492
00:17:12,480 --> 00:17:16,439
Engineers is

493
00:17:13,679 --> 00:17:18,600
um is a trainer model that recognizes

494
00:17:16,439 --> 00:17:20,880
you know boats and trains and and things

495
00:17:18,600 --> 00:17:24,240
like that just just count traffic

496
00:17:20,880 --> 00:17:25,620
basically to demonstrate

497
00:17:24,240 --> 00:17:27,179
um I was actually running this last

498
00:17:25,620 --> 00:17:28,140
night just in case I've had a failure

499
00:17:27,179 --> 00:17:31,799
today

500
00:17:28,140 --> 00:17:34,260
um and every hour the across the casinos

501
00:17:31,799 --> 00:17:37,380
got these um little gas towers and they

502
00:17:34,260 --> 00:17:39,059
um and the Senate little gas balls um

503
00:17:37,380 --> 00:17:40,799
it's global warming

504
00:17:39,059 --> 00:17:43,440
um but uh

505
00:17:40,799 --> 00:17:45,480
um it's been doing that for decades uh

506
00:17:43,440 --> 00:17:48,720
lots of fried pigeon I believe

507
00:17:45,480 --> 00:17:49,980
um and uh I don't know I might take a

508
00:17:48,720 --> 00:17:51,419
few seconds and say oh maybe a bit later

509
00:17:49,980 --> 00:17:53,160
to show you the view because they've got

510
00:17:51,419 --> 00:17:54,840
these massive gas ports over the river

511
00:17:53,160 --> 00:17:57,179
which was pretty impressive in black and

512
00:17:54,840 --> 00:17:58,740
white okay the last thing I'll show you

513
00:17:57,179 --> 00:18:00,780
while we're on a roll

514
00:17:58,740 --> 00:18:02,340
is um

515
00:18:00,780 --> 00:18:04,320
so in some ways this is not that

516
00:18:02,340 --> 00:18:06,059
exciting because we see video and stuff

517
00:18:04,320 --> 00:18:07,919
going to browsers all day every day

518
00:18:06,059 --> 00:18:09,299
these days um but it's really more the

519
00:18:07,919 --> 00:18:10,200
architecture that was put in place to

520
00:18:09,299 --> 00:18:11,640
make hopefully make this a little bit

521
00:18:10,200 --> 00:18:12,299
more moderate and easier I guess is the

522
00:18:11,640 --> 00:18:14,100
point

523
00:18:12,299 --> 00:18:17,160
oh hang on so I've opened and showed the

524
00:18:14,100 --> 00:18:19,940
same same thing again sorry

525
00:18:17,160 --> 00:18:22,559
nice fast internet fantastic

526
00:18:19,940 --> 00:18:25,799
thanks there thanks Steve

527
00:18:22,559 --> 00:18:27,780
he's uh the uh all the volunteers are

528
00:18:25,799 --> 00:18:29,400
awesome as always now this takes a

529
00:18:27,780 --> 00:18:32,100
little bit longer because um the ml

530
00:18:29,400 --> 00:18:34,580
model takes about 10 seconds to um to

531
00:18:32,100 --> 00:18:36,660
load up maybe a bit longer and

532
00:18:34,580 --> 00:18:37,980
there's a little problem because I have

533
00:18:36,660 --> 00:18:39,419
distributed leases in the lease time was

534
00:18:37,980 --> 00:18:40,679
also about 10 seconds saying if we don't

535
00:18:39,419 --> 00:18:41,580
hear back

536
00:18:40,679 --> 00:18:44,160
um

537
00:18:41,580 --> 00:18:46,140
uh assume there's a problem or not but I

538
00:18:44,160 --> 00:18:47,880
have actually extend the lease out so I

539
00:18:46,140 --> 00:18:49,559
should be that should be up now here we

540
00:18:47,880 --> 00:18:51,059
go no hang on

541
00:18:49,559 --> 00:18:52,500
oh sorry it's I keep going back to My

542
00:18:51,059 --> 00:18:54,660
Demo page oh here we go that's actually

543
00:18:52,500 --> 00:18:58,320
working reasonably well uh yeah so this

544
00:18:54,660 --> 00:18:59,820
is um just round trip from the camera

545
00:18:58,320 --> 00:19:01,140
does everyone can see it okay

546
00:18:59,820 --> 00:19:02,880
um yeah

547
00:19:01,140 --> 00:19:04,320
and the nice thing about we're about to

548
00:19:02,880 --> 00:19:07,980
see we should be able to get roughly um

549
00:19:04,320 --> 00:19:09,660
less than a second latency so from from

550
00:19:07,980 --> 00:19:12,780
this camera up to the data center and

551
00:19:09,660 --> 00:19:14,340
then back and back down again so wow

552
00:19:12,780 --> 00:19:16,760
the time I worked we can all go we can

553
00:19:14,340 --> 00:19:16,760
all go home

554
00:19:18,720 --> 00:19:21,740
Okay so

555
00:19:22,440 --> 00:19:25,460
it's actually going to

556
00:19:25,860 --> 00:19:31,260
that's what I want all right so um I

557
00:19:29,280 --> 00:19:33,179
have to speed this up a bit so there's

558
00:19:31,260 --> 00:19:35,400
been lots of influences so one is the

559
00:19:33,179 --> 00:19:38,039
the actor model it's a model for doing

560
00:19:35,400 --> 00:19:39,600
concurrent computation with distributed

561
00:19:38,039 --> 00:19:43,500
systems you know everything is an actor

562
00:19:39,600 --> 00:19:46,080
it was created back in 74 and Guy Steele

563
00:19:43,500 --> 00:19:47,880
was curious about whether it could

564
00:19:46,080 --> 00:19:50,160
actually be not just Theory but work in

565
00:19:47,880 --> 00:19:51,600
the invented scheme specifically and

566
00:19:50,160 --> 00:19:53,520
that for other reasons or to test out

567
00:19:51,600 --> 00:19:55,440
the actor model and along the way

568
00:19:53,520 --> 00:19:56,580
invented call by continuations and I

569
00:19:55,440 --> 00:19:57,720
think he's probably like 21 or something

570
00:19:56,580 --> 00:20:00,840
like that

571
00:19:57,720 --> 00:20:02,660
a bit scary but also erlang is probably

572
00:20:00,840 --> 00:20:06,840
the most famous use of the actor model

573
00:20:02,660 --> 00:20:07,980
the erlang language uh powers or did

574
00:20:06,840 --> 00:20:09,480
power a lot of the world's

575
00:20:07,980 --> 00:20:11,220
telecommunication networks very very

576
00:20:09,480 --> 00:20:13,440
reliably and also the links used in

577
00:20:11,220 --> 00:20:15,179
WhatsApp and WeChat who knew

578
00:20:13,440 --> 00:20:16,679
um so the actor models um maybe not

579
00:20:15,179 --> 00:20:18,960
everyone's over it's pretty successful

580
00:20:16,679 --> 00:20:20,700
another influence for me was the uh it

581
00:20:18,960 --> 00:20:22,679
was a genie which was a stupid system

582
00:20:20,700 --> 00:20:25,620
that um a little Breakaway group in

583
00:20:22,679 --> 00:20:27,299
Sunrise which I'm super well designed by

584
00:20:25,620 --> 00:20:29,820
some very very smart people for

585
00:20:27,299 --> 00:20:31,980
experienced fully asynchronous service

586
00:20:29,820 --> 00:20:33,539
based they can just use Java interfaces

587
00:20:31,980 --> 00:20:35,360
so you didn't have to worry about any

588
00:20:33,539 --> 00:20:38,100
stubs or

589
00:20:35,360 --> 00:20:39,480
marching or marshalling but wow the out

590
00:20:38,100 --> 00:20:40,919
of box experience was just really really

591
00:20:39,480 --> 00:20:42,179
tough it was very challenging to use so

592
00:20:40,919 --> 00:20:44,160
unfortunately it's the function now but

593
00:20:42,179 --> 00:20:47,960
it was a great great thing to learn from

594
00:20:44,160 --> 00:20:50,340
but just just briefly on actors

595
00:20:47,960 --> 00:20:52,580
actors can do three things they can

596
00:20:50,340 --> 00:20:55,140
basically send messages to other actors

597
00:20:52,580 --> 00:20:57,419
they can create other actors and they

598
00:20:55,140 --> 00:20:58,919
can also change their state which means

599
00:20:57,419 --> 00:21:01,200
that there their behavior on the next

600
00:20:58,919 --> 00:21:03,419
message will change and the great thing

601
00:21:01,200 --> 00:21:05,760
about actors compared to other just

602
00:21:03,419 --> 00:21:07,980
having a whole Myriad of services it's

603
00:21:05,760 --> 00:21:09,419
just easy to to um reason about them

604
00:21:07,980 --> 00:21:11,179
about your network you know just

605
00:21:09,419 --> 00:21:14,100
simplifies things

606
00:21:11,179 --> 00:21:15,419
but without compromising being able to

607
00:21:14,100 --> 00:21:17,460
get stuff done

608
00:21:15,419 --> 00:21:20,940
so some more inferences uh yeah everyone

609
00:21:17,460 --> 00:21:22,260
knows I love lisp you know um and uh in

610
00:21:20,940 --> 00:21:23,340
this system we've actually used list for

611
00:21:22,260 --> 00:21:24,419
the for the messages I think if

612
00:21:23,340 --> 00:21:25,980
everything as being a function call

613
00:21:24,419 --> 00:21:28,679
which can just translate directly to a

614
00:21:25,980 --> 00:21:30,059
lisp s expression and what's great is

615
00:21:28,679 --> 00:21:31,620
that allows you to do mobile code you

616
00:21:30,059 --> 00:21:32,760
know it could that s expression could be

617
00:21:31,620 --> 00:21:33,900
data structure but also could be a

618
00:21:32,760 --> 00:21:36,480
filter so you basically say to another

619
00:21:33,900 --> 00:21:37,860
another notary Network or another actor

620
00:21:36,480 --> 00:21:39,240
in a network here's this filter we're

621
00:21:37,860 --> 00:21:41,280
going to apply so I can make very easily

622
00:21:39,240 --> 00:21:42,840
move your computation around the network

623
00:21:41,280 --> 00:21:44,640
um of course you know keeping security

624
00:21:42,840 --> 00:21:46,260
concerns in mind

625
00:21:44,640 --> 00:21:47,700
um but great thing about Bliss machines

626
00:21:46,260 --> 00:21:49,200
back in the day was everything was

627
00:21:47,700 --> 00:21:51,240
enlisted have they had Hardware

628
00:21:49,200 --> 00:21:52,799
instructions uh the operating system GUI

629
00:21:51,240 --> 00:21:54,120
applications it was just Turtles all the

630
00:21:52,799 --> 00:21:56,700
way down it just showed you what could

631
00:21:54,120 --> 00:21:58,679
be done the other um inspiration is

632
00:21:56,700 --> 00:22:00,360
Small Talk which is

633
00:21:58,679 --> 00:22:02,580
um uh Alan K say look you always

634
00:22:00,360 --> 00:22:04,679
intended that uh it really should be

635
00:22:02,580 --> 00:22:06,539
message oriented not object oriented um

636
00:22:04,679 --> 00:22:08,940
but yeah basically about small talk is

637
00:22:06,539 --> 00:22:10,440
even the language itself is where you

638
00:22:08,940 --> 00:22:11,820
basically send the messages to other

639
00:22:10,440 --> 00:22:13,500
parts so your if statement is actually

640
00:22:11,820 --> 00:22:16,080
it's performed through sending messages

641
00:22:13,500 --> 00:22:19,440
to the the true block or the the false

642
00:22:16,080 --> 00:22:20,760
Block it's um and if you've seen a small

643
00:22:19,440 --> 00:22:22,020
talk um or if you haven't actually

644
00:22:20,760 --> 00:22:24,780
haven't seen small 2 you can download

645
00:22:22,020 --> 00:22:27,179
squeak and it runs on Modern up your

646
00:22:24,780 --> 00:22:28,620
operating systems and you basically see

647
00:22:27,179 --> 00:22:30,179
what an amazing system this was where

648
00:22:28,620 --> 00:22:31,500
you can you can click on a window and

649
00:22:30,179 --> 00:22:33,000
it'll take you to the source code

650
00:22:31,500 --> 00:22:34,440
browser and you can change the behavior

651
00:22:33,000 --> 00:22:38,039
in the system on the fly without without

652
00:22:34,440 --> 00:22:39,539
rebooting in fact The Story Goes which

653
00:22:38,039 --> 00:22:41,820
is true because in CCTV jobs talking

654
00:22:39,539 --> 00:22:43,799
about this on YouTube when um the small

655
00:22:41,820 --> 00:22:45,480
talk team invited a number of companies

656
00:22:43,799 --> 00:22:46,860
in to see what they've done because no

657
00:22:45,480 --> 00:22:50,100
one had ever seen no one had really ever

658
00:22:46,860 --> 00:22:51,480
seen a graphics using a face before with

659
00:22:50,100 --> 00:22:54,559
a Windows and mys except a few

660
00:22:51,480 --> 00:22:58,140
researchers and angle bar of course

661
00:22:54,559 --> 00:22:59,520
and uh they they they showed they showed

662
00:22:58,140 --> 00:23:00,360
that they showed him about six things he

663
00:22:59,520 --> 00:23:02,520
said look

664
00:23:00,360 --> 00:23:03,539
um I was so blown away by you know what

665
00:23:02,520 --> 00:23:05,940
the first thing I saw which is the

666
00:23:03,539 --> 00:23:07,440
windows which even obviously took back

667
00:23:05,940 --> 00:23:10,080
to Apple he totally forgot about what

668
00:23:07,440 --> 00:23:12,840
those other things were ethernet laser

669
00:23:10,080 --> 00:23:15,179
printing yeah but the thing that

670
00:23:12,840 --> 00:23:18,419
captures imagination was according to

671
00:23:15,179 --> 00:23:19,980
the Xerox research was how they did

672
00:23:18,419 --> 00:23:22,020
select was really ugly it basically just

673
00:23:19,980 --> 00:23:24,539
blocked out the whole text and it was

674
00:23:22,020 --> 00:23:25,620
ugly and and Steve asked him well that's

675
00:23:24,539 --> 00:23:27,000
just terrible you know what can you do

676
00:23:25,620 --> 00:23:28,860
about it and then the engineer said look

677
00:23:27,000 --> 00:23:31,919
hang on give me a sec I found that you

678
00:23:28,860 --> 00:23:33,780
found the uh the small talk method

679
00:23:31,919 --> 00:23:35,220
um I said look if we subtract this

680
00:23:33,780 --> 00:23:37,320
rectangle from this rectangle basically

681
00:23:35,220 --> 00:23:39,000
this nice outline did that in about a

682
00:23:37,320 --> 00:23:40,200
minute uh didn't stop the system and

683
00:23:39,000 --> 00:23:43,260
then he basically had like almost like

684
00:23:40,200 --> 00:23:45,900
moving Arts like type thing

685
00:23:43,260 --> 00:23:48,360
um yeah that's what that's right Mr jobs

686
00:23:45,900 --> 00:23:50,460
forgot about all the other things anyway

687
00:23:48,360 --> 00:23:52,500
um moving on because that's right so

688
00:23:50,460 --> 00:23:54,299
some overall design principles

689
00:23:52,500 --> 00:23:55,620
um I've completely stands through

690
00:23:54,299 --> 00:23:56,700
messages

691
00:23:55,620 --> 00:23:58,020
um I want these things to be very

692
00:23:56,700 --> 00:24:00,419
lightweight so they can run on embedded

693
00:23:58,020 --> 00:24:02,039
systems and I really want to think about

694
00:24:00,419 --> 00:24:02,760
in terms of things as function calls you

695
00:24:02,039 --> 00:24:04,320
know

696
00:24:02,760 --> 00:24:06,360
um so basically you say I'm just calling

697
00:24:04,320 --> 00:24:07,799
functions between my um actors locally

698
00:24:06,360 --> 00:24:09,179
or remotely and not have to worry about

699
00:24:07,799 --> 00:24:12,000
marshalling on marshalling all those

700
00:24:09,179 --> 00:24:13,679
things and something about just simple

701
00:24:12,000 --> 00:24:16,080
human readable messages is so easy to

702
00:24:13,679 --> 00:24:18,360
debug especially if using mqtt basically

703
00:24:16,080 --> 00:24:19,980
just do a mosquito Pub to basically do

704
00:24:18,360 --> 00:24:22,140
some injection testing to go hey let's

705
00:24:19,980 --> 00:24:24,179
um just exercise that actor or device or

706
00:24:22,140 --> 00:24:25,260
whatever and you can just also subscribe

707
00:24:24,179 --> 00:24:27,419
and see what's going on in your system

708
00:24:25,260 --> 00:24:28,740
so it's really great but of course that

709
00:24:27,419 --> 00:24:30,480
does collapse a bit when you've got lots

710
00:24:28,740 --> 00:24:32,100
and lots of things going on

711
00:24:30,480 --> 00:24:34,320
um so in the system pretty much trying

712
00:24:32,100 --> 00:24:36,059
to use a lisp and NPC everywhere but uh

713
00:24:34,320 --> 00:24:38,520
the applications if you want to

714
00:24:36,059 --> 00:24:39,659
um you know this often have good reasons

715
00:24:38,520 --> 00:24:42,539
or pressure to basically you know use

716
00:24:39,659 --> 00:24:43,980
use Json and Avro for the foundation or

717
00:24:42,539 --> 00:24:45,480
protobufs or whatever that's that's fine

718
00:24:43,980 --> 00:24:47,940
too

719
00:24:45,480 --> 00:24:49,919
um think about uh when you have a fully

720
00:24:47,940 --> 00:24:51,720
estimated system is you can't really do

721
00:24:49,919 --> 00:24:53,640
blocking calls you know to like do RPC

722
00:24:51,720 --> 00:24:54,960
we make a call a request and expect back

723
00:24:53,640 --> 00:24:56,640
a response

724
00:24:54,960 --> 00:24:58,620
um so pretty simple rule is that

725
00:24:56,640 --> 00:25:00,240
functions don't return values

726
00:24:58,620 --> 00:25:01,919
um which makes uh much easier to do a

727
00:25:00,240 --> 00:25:03,720
consumer system but also makes it a

728
00:25:01,919 --> 00:25:05,280
little bit more difficult to program in

729
00:25:03,720 --> 00:25:06,600
terms of you know do you do callbacks uh

730
00:25:05,280 --> 00:25:08,280
what if you're going to create something

731
00:25:06,600 --> 00:25:09,900
and you want to get great get its

732
00:25:08,280 --> 00:25:12,000
reference back that becomes really hard

733
00:25:09,900 --> 00:25:13,980
but there are ways around that for

734
00:25:12,000 --> 00:25:17,159
example using eventual consisting cases

735
00:25:13,980 --> 00:25:19,740
you can basically say uh uh I'm

736
00:25:17,159 --> 00:25:21,419
interested in this uh in this state uh

737
00:25:19,740 --> 00:25:24,000
over the network like this reference or

738
00:25:21,419 --> 00:25:25,260
it's computation or result and that just

739
00:25:24,000 --> 00:25:28,860
lands an occasion we're seeing somebody

740
00:25:25,260 --> 00:25:30,600
just do a local get that value so the

741
00:25:28,860 --> 00:25:31,919
problem is if you're only doing if

742
00:25:30,600 --> 00:25:33,779
you're fundamentally got a request

743
00:25:31,919 --> 00:25:35,520
response system based around HTTP or

744
00:25:33,779 --> 00:25:36,960
rest or all the graphical all these

745
00:25:35,520 --> 00:25:38,400
things is you just go all the students

746
00:25:36,960 --> 00:25:40,260
what is asynchronous system things will

747
00:25:38,400 --> 00:25:42,360
need speed it becomes a special hard

748
00:25:40,260 --> 00:25:43,559
case if you tackle it early on and say

749
00:25:42,360 --> 00:25:45,360
no we'll just make or just try and make

750
00:25:43,559 --> 00:25:47,340
it easy from start while those problems

751
00:25:45,360 --> 00:25:49,200
go away the other thing we use is um

752
00:25:47,340 --> 00:25:50,820
Dynamics proxy objects so we can have

753
00:25:49,200 --> 00:25:52,679
layers of abstraction so we can abstract

754
00:25:50,820 --> 00:25:54,360
out different um different third-party

755
00:25:52,679 --> 00:25:59,940
things so we can have you know amputes

756
00:25:54,360 --> 00:26:02,400
of Transport or Roz or um or or a hard

757
00:25:59,940 --> 00:26:04,919
graphical going over HTTP so or we can

758
00:26:02,400 --> 00:26:07,860
um swap in different security models

759
00:26:04,919 --> 00:26:10,080
um and so on by by doing that and uh

760
00:26:07,860 --> 00:26:12,900
just as an aside there's a great talk by

761
00:26:10,080 --> 00:26:15,600
Kevin Greer around the phone group at

762
00:26:12,900 --> 00:26:17,820
Google and uh and he just is comparing

763
00:26:15,600 --> 00:26:18,960
uh the advantages that Unix had over

764
00:26:17,820 --> 00:26:20,460
mortex because they did what's called

765
00:26:18,960 --> 00:26:22,440
coding the perimeter they basically

766
00:26:20,460 --> 00:26:24,059
looked at their cross-cutting concerns

767
00:26:22,440 --> 00:26:26,279
and you basically say we've got

768
00:26:24,059 --> 00:26:28,679
functionality on one axis and we've got

769
00:26:26,279 --> 00:26:30,000
um uh I guess I think things must be

770
00:26:28,679 --> 00:26:32,100
like persistence and configuration or

771
00:26:30,000 --> 00:26:32,940
communication on the other axis instead

772
00:26:32,100 --> 00:26:34,260
of basically trying to for each

773
00:26:32,940 --> 00:26:36,000
individual program and figure those

774
00:26:34,260 --> 00:26:36,900
things out over and over again in Unix

775
00:26:36,000 --> 00:26:38,940
because we're having you know we

776
00:26:36,900 --> 00:26:40,620
construct things of pipes and so on and

777
00:26:38,940 --> 00:26:43,799
small composable pieces

778
00:26:40,620 --> 00:26:45,120
um Dennis Ritchie and uh and Ken

779
00:26:43,799 --> 00:26:47,220
Thompson were able to move much faster

780
00:26:45,120 --> 00:26:50,299
because they were on the perimeter they

781
00:26:47,220 --> 00:26:50,299
won't try to fill in the whole area

782
00:26:51,419 --> 00:26:56,820
um so just about how that looks um very

783
00:26:54,900 --> 00:26:59,159
quickly is I've got a couple of actors

784
00:26:56,820 --> 00:27:00,600
here so actor zero wants to create actor

785
00:26:59,159 --> 00:27:02,820
one so it's going to talk to a livestock

786
00:27:00,600 --> 00:27:04,140
manager do that so the the blue bits the

787
00:27:02,820 --> 00:27:05,880
blue message is around

788
00:27:04,140 --> 00:27:07,919
um creating the actor so active zero

789
00:27:05,880 --> 00:27:09,240
says look create an act of the livestock

790
00:27:07,919 --> 00:27:10,500
manager the livestock manager will

791
00:27:09,240 --> 00:27:12,539
create that

792
00:27:10,500 --> 00:27:13,980
um that actor actor one that that

793
00:27:12,539 --> 00:27:15,659
reference gets back to actor zero and

794
00:27:13,980 --> 00:27:18,419
then back to Zurich and then say Okay

795
00:27:15,659 --> 00:27:20,340
pause run delete whatever you want and

796
00:27:18,419 --> 00:27:22,260
so the main thing I want to show here is

797
00:27:20,340 --> 00:27:25,200
that the in the within the actors you're

798
00:27:22,260 --> 00:27:27,539
just using function calls uh whether

799
00:27:25,200 --> 00:27:28,980
using the dynamic proxy we can just turn

800
00:27:27,539 --> 00:27:30,419
those into into best expressions and

801
00:27:28,980 --> 00:27:33,840
send those across the network and ask

802
00:27:30,419 --> 00:27:35,340
them for you and uh and so you'd have to

803
00:27:33,840 --> 00:27:36,960
worry about all the machinery and the

804
00:27:35,340 --> 00:27:38,400
other thing is the the uh the green bit

805
00:27:36,960 --> 00:27:39,779
that's the case that's the caching we're

806
00:27:38,400 --> 00:27:42,720
basically using eventual consistency so

807
00:27:39,779 --> 00:27:45,600
just one one uh one effectively class of

808
00:27:42,720 --> 00:27:47,520
uh in station for a producer just wrap

809
00:27:45,600 --> 00:27:49,380
your state and the same thing on the on

810
00:27:47,520 --> 00:27:51,179
the other side we just have a consumer

811
00:27:49,380 --> 00:27:52,380
that says look um this is the state I

812
00:27:51,179 --> 00:27:53,760
want you can filter that state so you

813
00:27:52,380 --> 00:27:55,140
don't get all of the state just just a

814
00:27:53,760 --> 00:27:57,059
bit you need and then what you just do

815
00:27:55,140 --> 00:27:59,940
is you step in into your local cache and

816
00:27:57,059 --> 00:28:01,679
uh and and get the most recent value so

817
00:27:59,940 --> 00:28:03,179
hopefully that gives a bit of conceptual

818
00:28:01,679 --> 00:28:04,799
View

819
00:28:03,179 --> 00:28:05,880
all right so some distributed design

820
00:28:04,799 --> 00:28:07,320
principles

821
00:28:05,880 --> 00:28:09,659
um

822
00:28:07,320 --> 00:28:11,640
uh back in the 90s um at some

823
00:28:09,659 --> 00:28:13,740
Microsystems uh there was this list of

824
00:28:11,640 --> 00:28:15,299
things that are basically fallacies if

825
00:28:13,740 --> 00:28:16,320
you believe that um the network is

826
00:28:15,299 --> 00:28:18,000
viable

827
00:28:16,320 --> 00:28:19,320
um yeah your application is not going to

828
00:28:18,000 --> 00:28:20,400
not going to be reliable and they had a

829
00:28:19,320 --> 00:28:24,059
list of seven of these and then James

830
00:28:20,400 --> 00:28:26,159
Gosling had the out of the eighth um uh

831
00:28:24,059 --> 00:28:27,659
even though uh Java entered back the

832
00:28:26,159 --> 00:28:30,000
network looked a bit homogeneous back in

833
00:28:27,659 --> 00:28:31,620
the 90s for a while it's not true all

834
00:28:30,000 --> 00:28:34,440
the devices we have these days are quite

835
00:28:31,620 --> 00:28:35,880
different so these are things if you

836
00:28:34,440 --> 00:28:38,100
make any of these assumptions

837
00:28:35,880 --> 00:28:40,860
um in your in your architectural design

838
00:28:38,100 --> 00:28:42,440
you're going to um you know come kind of

839
00:28:40,860 --> 00:28:44,820
hit some potholes for sure

840
00:28:42,440 --> 00:28:46,860
uh and I'm not trying to basically try

841
00:28:44,820 --> 00:28:48,720
and cover all Distribution Systems uh

842
00:28:46,860 --> 00:28:50,100
theory in a few minutes but sort of or

843
00:28:48,720 --> 00:28:51,720
if you mentioned a few other things so

844
00:28:50,100 --> 00:28:53,820
it's cap theorem which basically says

845
00:28:51,720 --> 00:28:55,700
three things you can only have two which

846
00:28:53,820 --> 00:28:57,659
is um

847
00:28:55,700 --> 00:28:59,940
partition part tolerance is like in

848
00:28:57,659 --> 00:29:01,200
handle Network failure uh consistency do

849
00:28:59,940 --> 00:29:02,520
you want when you I mean we do rights

850
00:29:01,200 --> 00:29:04,740
and reads does everything have the same

851
00:29:02,520 --> 00:29:06,960
value and availability you know can you

852
00:29:04,740 --> 00:29:09,059
see can you see the that piece of

853
00:29:06,960 --> 00:29:12,360
information from a from a from another

854
00:29:09,059 --> 00:29:15,179
part of the system things like databases

855
00:29:12,360 --> 00:29:16,440
which do acid they really work hard to

856
00:29:15,179 --> 00:29:18,179
make sure that no matter what you have

857
00:29:16,440 --> 00:29:19,980
availability consistency

858
00:29:18,179 --> 00:29:22,559
um but the problem with that is you've

859
00:29:19,980 --> 00:29:24,419
got to do a network failure always so

860
00:29:22,559 --> 00:29:26,100
that was about 20 odd years ago as the

861
00:29:24,419 --> 00:29:28,440
cap theorem in the last dozen or so

862
00:29:26,100 --> 00:29:30,360
years there's what's called Pac ELC

863
00:29:28,440 --> 00:29:32,880
which says not only do you have to worry

864
00:29:30,360 --> 00:29:34,740
about cap which is basically you choose

865
00:29:32,880 --> 00:29:36,600
to have partition tolerance and then

866
00:29:34,740 --> 00:29:38,580
worry about about consistency even if

867
00:29:36,600 --> 00:29:40,020
even if that was it was good you still

868
00:29:38,580 --> 00:29:42,600
have to choose between whether you want

869
00:29:40,020 --> 00:29:43,620
to have um good latency or you want

870
00:29:42,600 --> 00:29:45,600
everything to be consistent because

871
00:29:43,620 --> 00:29:48,659
there's a bit of a trade-off there so um

872
00:29:45,600 --> 00:29:50,779
in um uh I go Services what I've done is

873
00:29:48,659 --> 00:29:53,580
I said look you have to worry about uh

874
00:29:50,779 --> 00:29:56,220
Network partition and what to do when

875
00:29:53,580 --> 00:29:57,960
the network fails but then after that

876
00:29:56,220 --> 00:29:59,100
um basically using a form of a or

877
00:29:57,960 --> 00:30:00,960
slightly stronger performance

878
00:29:59,100 --> 00:30:02,399
consistency which is effectively called

879
00:30:00,960 --> 00:30:04,140
sequential consistency where you still

880
00:30:02,399 --> 00:30:06,059
preserve the right order and you and

881
00:30:04,140 --> 00:30:07,320
basically the the data will be available

882
00:30:06,059 --> 00:30:08,520
eventually

883
00:30:07,320 --> 00:30:10,320
the other thing we use is something

884
00:30:08,520 --> 00:30:13,440
called distributed leases

885
00:30:10,320 --> 00:30:15,179
um so when if you basically uh done some

886
00:30:13,440 --> 00:30:16,980
operation that requires basically causes

887
00:30:15,179 --> 00:30:19,799
another part of the system to consider

888
00:30:16,980 --> 00:30:21,240
to acquire some resources maybe made a

889
00:30:19,799 --> 00:30:23,880
database connection you've now got a

890
00:30:21,240 --> 00:30:25,200
connection there how long should This

891
00:30:23,880 --> 00:30:26,700
Server hold on to that and if you

892
00:30:25,200 --> 00:30:28,799
haven't agreed up front in the network

893
00:30:26,700 --> 00:30:30,240
files you've got no way of um carrying

894
00:30:28,799 --> 00:30:31,860
on the conversation like closing out

895
00:30:30,240 --> 00:30:33,419
properly so the thing about leases you

896
00:30:31,860 --> 00:30:35,059
can say look I'm going to list for five

897
00:30:33,419 --> 00:30:37,500
minutes and if you don't hear from me

898
00:30:35,059 --> 00:30:40,380
both sides can assume it's done and over

899
00:30:37,500 --> 00:30:41,940
so even a network partition uh the two

900
00:30:40,380 --> 00:30:43,200
parties can will know what the situation

901
00:30:41,940 --> 00:30:44,460
is you know five minutes went past you

902
00:30:43,200 --> 00:30:46,740
haven't heard from each other that's all

903
00:30:44,460 --> 00:30:48,659
it's all over very powerful very simple

904
00:30:46,740 --> 00:30:52,860
but very powerful technique

905
00:30:48,659 --> 00:30:53,460
okay so um people who've come on to the

906
00:30:52,860 --> 00:30:55,440
um

907
00:30:53,460 --> 00:30:56,640
otmc have know that we've used sync

908
00:30:55,440 --> 00:30:58,380
called the Arco engine which is a micro

909
00:30:56,640 --> 00:31:00,179
bit of micropython code for the you know

910
00:30:58,380 --> 00:31:01,320
like the swag badge and other things so

911
00:31:00,179 --> 00:31:03,779
what's the difference between services

912
00:31:01,320 --> 00:31:05,460
and the engine basically the engine was

913
00:31:03,779 --> 00:31:07,460
really about providing a little bit of

914
00:31:05,460 --> 00:31:09,720
an event based model some little

915
00:31:07,460 --> 00:31:11,880
peripheral drivers for um you know

916
00:31:09,720 --> 00:31:13,740
buttons and screens and networking and

917
00:31:11,880 --> 00:31:15,960
so on uh just purely about the device

918
00:31:13,740 --> 00:31:17,159
but the services framework I'm

919
00:31:15,960 --> 00:31:19,020
discussing today so that's sort of the

920
00:31:17,159 --> 00:31:20,640
difference and uh so the services

921
00:31:19,020 --> 00:31:21,840
basically says here's the definition of

922
00:31:20,640 --> 00:31:25,020
all the protocols provides a reference

923
00:31:21,840 --> 00:31:27,360
implementation and then the engine

924
00:31:25,020 --> 00:31:29,640
um uh can I do everything you need on

925
00:31:27,360 --> 00:31:31,260
the device or wherever and and then says

926
00:31:29,640 --> 00:31:33,320
oh and here's how we can use services to

927
00:31:31,260 --> 00:31:35,700
interconnect

928
00:31:33,320 --> 00:31:38,419
up now

929
00:31:35,700 --> 00:31:38,419
okay just check out

930
00:31:41,520 --> 00:31:46,460
okay so this this is um how archive

931
00:31:44,820 --> 00:31:49,980
servers look from from a very high level

932
00:31:46,460 --> 00:31:51,779
uh the um there's the main the main

933
00:31:49,980 --> 00:31:54,120
Services the registrar so the registrar

934
00:31:51,779 --> 00:31:55,740
allows you to um discover join and

935
00:31:54,120 --> 00:31:57,480
discover all the other services so in

936
00:31:55,740 --> 00:31:59,340
this case we just have two services on

937
00:31:57,480 --> 00:32:01,440
the bottom one and two they both add

938
00:31:59,340 --> 00:32:02,640
themselves to the registrar with an its

939
00:32:01,440 --> 00:32:04,919
expression which is basically just add

940
00:32:02,640 --> 00:32:06,539
and a bit of detail a few details so and

941
00:32:04,919 --> 00:32:08,460
then um and then it allows to find each

942
00:32:06,539 --> 00:32:10,559
other by querying the Registrar of

943
00:32:08,460 --> 00:32:12,000
course and then then uh service one

944
00:32:10,559 --> 00:32:13,340
confined service to and do output so

945
00:32:12,000 --> 00:32:16,200
it's not nothing nothing too

946
00:32:13,340 --> 00:32:18,539
revolutionary there

947
00:32:16,200 --> 00:32:20,220
um right so what is the service

948
00:32:18,539 --> 00:32:23,100
um well pretty much like I've asked a

949
00:32:20,220 --> 00:32:24,960
definite service so a service can um Can

950
00:32:23,100 --> 00:32:28,140
can join the network can be discovered

951
00:32:24,960 --> 00:32:29,460
by other components uh it can uh then uh

952
00:32:28,140 --> 00:32:30,899
receive receive messages through

953
00:32:29,460 --> 00:32:33,000
subscription they can publish some

954
00:32:30,899 --> 00:32:36,240
submissions but uh that's what I've done

955
00:32:33,000 --> 00:32:37,620
it in Ico is um is also to find some

956
00:32:36,240 --> 00:32:40,200
additional things the Surfers have most

957
00:32:37,620 --> 00:32:41,340
importantly is topic prefix so in name

958
00:32:40,200 --> 00:32:42,600
contain other messaging systems you

959
00:32:41,340 --> 00:32:44,159
always choose topics and often you do

960
00:32:42,600 --> 00:32:46,380
that by saying oh what's the temperature

961
00:32:44,159 --> 00:32:48,720
inside the topic will be kitchen slash

962
00:32:46,380 --> 00:32:50,760
temperature in this in this case our

963
00:32:48,720 --> 00:32:52,679
coach Auto generates a topic which is

964
00:32:50,760 --> 00:32:54,059
based on your namespace so it can

965
00:32:52,679 --> 00:32:55,980
basically partition things for reasons

966
00:32:54,059 --> 00:32:57,779
for security or testing whatever so you

967
00:32:55,980 --> 00:33:00,299
get these different name spaces the host

968
00:32:57,779 --> 00:33:01,679
that you're on the processity on the

969
00:33:00,299 --> 00:33:03,779
service IDs those things give you a

970
00:33:01,679 --> 00:33:05,520
unique topic path which is also useful

971
00:33:03,779 --> 00:33:06,779
because I say oh what host is my service

972
00:33:05,520 --> 00:33:07,919
running on or what process if you want

973
00:33:06,779 --> 00:33:09,720
if you want to kill it so it's going to

974
00:33:07,919 --> 00:33:11,640
be useful that way and it also means

975
00:33:09,720 --> 00:33:13,740
there's no collisions and what the the

976
00:33:11,640 --> 00:33:16,799
service or device can do is it can then

977
00:33:13,740 --> 00:33:18,480
just append additional initial strings

978
00:33:16,799 --> 00:33:21,360
saying I want an input for an apple

979
00:33:18,480 --> 00:33:23,279
topic or or logging or whatever I also

980
00:33:21,360 --> 00:33:25,140
can name the services the protocol

981
00:33:23,279 --> 00:33:26,580
defines what what Implement what

982
00:33:25,140 --> 00:33:28,380
interfaces the service Implement so it

983
00:33:26,580 --> 00:33:30,659
could be okay it's always going to be a

984
00:33:28,380 --> 00:33:32,100
service possibly an actor who could also

985
00:33:30,659 --> 00:33:34,679
be a temperature sensor or it could be

986
00:33:32,100 --> 00:33:36,120
something it's a category of group of

987
00:33:34,679 --> 00:33:37,860
other things so the interfaces to say

988
00:33:36,120 --> 00:33:39,419
well all those things that you um you

989
00:33:37,860 --> 00:33:40,740
are which is also another way of finding

990
00:33:39,419 --> 00:33:42,779
something so basically could say I find

991
00:33:40,740 --> 00:33:45,480
all the um you know find all the

992
00:33:42,779 --> 00:33:47,940
persistence stores you know or databases

993
00:33:45,480 --> 00:33:49,559
we also have the transport which means

994
00:33:47,940 --> 00:33:52,380
we can swap in other transports like

995
00:33:49,559 --> 00:33:53,700
rock Ross not CBT and the tags are

996
00:33:52,380 --> 00:33:55,019
really how you can find things like

997
00:33:53,700 --> 00:33:57,360
likely with an email you can basically

998
00:33:55,019 --> 00:34:00,120
Tag Services and say you know create new

999
00:33:57,360 --> 00:34:01,440
keywords and give them values so right

1000
00:34:00,120 --> 00:34:03,539
now using that for eventual consistency

1001
00:34:01,440 --> 00:34:05,519
you can say e s equals two ten minutes

1002
00:34:03,539 --> 00:34:07,620
is cool I better speed this up

1003
00:34:05,519 --> 00:34:10,379
um to to um to basically say you support

1004
00:34:07,620 --> 00:34:11,820
Venture consistency UI for UI hints and

1005
00:34:10,379 --> 00:34:14,220
and so on

1006
00:34:11,820 --> 00:34:15,659
okay so the register is

1007
00:34:14,220 --> 00:34:16,800
um we support primary and secondary

1008
00:34:15,659 --> 00:34:18,540
registrars

1009
00:34:16,800 --> 00:34:20,760
um and that's just the messages that you

1010
00:34:18,540 --> 00:34:24,000
can send to a registrar so add add

1011
00:34:20,760 --> 00:34:26,580
remove uh hyper failure share allows you

1012
00:34:24,000 --> 00:34:28,379
to say look um I want to have a filter

1013
00:34:26,580 --> 00:34:30,300
that's all the stars are wild cards for

1014
00:34:28,379 --> 00:34:31,440
the various um service Fields here's the

1015
00:34:30,300 --> 00:34:33,540
service Fields I want you to filter on

1016
00:34:31,440 --> 00:34:36,480
and um

1017
00:34:33,540 --> 00:34:38,099
uh and that those messages those

1018
00:34:36,480 --> 00:34:39,839
missions come back into your case

1019
00:34:38,099 --> 00:34:42,480
um so you can initiate that the other

1020
00:34:39,839 --> 00:34:43,919
thing we do is use by the Dashboard is

1021
00:34:42,480 --> 00:34:45,060
um it's also get the history so the red

1022
00:34:43,919 --> 00:34:46,080
strike keeps the history of services

1023
00:34:45,060 --> 00:34:47,820
have gone away so you might have a

1024
00:34:46,080 --> 00:34:49,080
service that starts up dies and goes

1025
00:34:47,820 --> 00:34:50,580
away you want to be able to at least

1026
00:34:49,080 --> 00:34:52,560
know that it ran and also be able to

1027
00:34:50,580 --> 00:34:53,820
look at its logs

1028
00:34:52,560 --> 00:34:55,500
okay so I'll quickly show you the

1029
00:34:53,820 --> 00:34:56,700
dashboard it's a bit like so you saw

1030
00:34:55,500 --> 00:34:58,680
briefly before it's a bit like top

1031
00:34:56,700 --> 00:35:00,240
interacts the registrar shows you all

1032
00:34:58,680 --> 00:35:01,740
your services

1033
00:35:00,240 --> 00:35:04,020
um it allows you to interrogate and

1034
00:35:01,740 --> 00:35:05,460
check all the variables and logging and

1035
00:35:04,020 --> 00:35:08,060
so on so let's see if I can just quickly

1036
00:35:05,460 --> 00:35:08,060
show that again

1037
00:35:16,560 --> 00:35:21,300
all right cool this is just running uh

1038
00:35:18,960 --> 00:35:22,740
on my my Home Server that's uh that's so

1039
00:35:21,300 --> 00:35:25,920
we've got the register up there we've

1040
00:35:22,740 --> 00:35:28,380
got two um uh swag badges

1041
00:35:25,920 --> 00:35:29,880
um set csp32s but I could just look

1042
00:35:28,380 --> 00:35:34,400
because I can just add another

1043
00:35:29,880 --> 00:35:34,400
um another just another service if I uh

1044
00:35:40,859 --> 00:35:46,260
samples

1045
00:35:43,680 --> 00:35:49,460
so this is the equivalent hello world

1046
00:35:46,260 --> 00:35:49,460
hopefully that's going to work for me

1047
00:35:50,640 --> 00:35:54,480
fun of doing things ah actually ah I

1048
00:35:53,220 --> 00:35:57,480
know what's happened there wrong screen

1049
00:35:54,480 --> 00:36:00,000
this adds up my backend server

1050
00:35:57,480 --> 00:36:01,260
whoops

1051
00:36:00,000 --> 00:36:04,260
um let me just do it on the local

1052
00:36:01,260 --> 00:36:07,260
machine now yeah so

1053
00:36:04,260 --> 00:36:09,740
it's like SSH but worse

1054
00:36:07,260 --> 00:36:09,740
hang on

1055
00:36:12,839 --> 00:36:16,560
okay so let's have a little about much

1056
00:36:14,339 --> 00:36:18,599
faster yeah 18.

1057
00:36:16,560 --> 00:36:19,980
um and so we can see the various

1058
00:36:18,599 --> 00:36:21,839
variables like it's life cycle State

1059
00:36:19,980 --> 00:36:23,700
it's ready you can change your log level

1060
00:36:21,839 --> 00:36:25,740
we can change that log level to to debug

1061
00:36:23,700 --> 00:36:28,220
for example does that sound you can now

1062
00:36:25,740 --> 00:36:30,720
change the debug

1063
00:36:28,220 --> 00:36:32,040
dictionaries and so on another one I do

1064
00:36:30,720 --> 00:36:33,900
is have the version number and the

1065
00:36:32,040 --> 00:36:35,339
source code for the service they uh so

1066
00:36:33,900 --> 00:36:36,599
the audio just click on that will take

1067
00:36:35,339 --> 00:36:38,640
you the source code for that that

1068
00:36:36,599 --> 00:36:40,980
service and you also see the tags so

1069
00:36:38,640 --> 00:36:42,420
this um this service supports eventual

1070
00:36:40,980 --> 00:36:43,760
consistency

1071
00:36:42,420 --> 00:36:46,260
um and what should be able to do is um

1072
00:36:43,760 --> 00:36:48,300
go to a log you'll be able to see be

1073
00:36:46,260 --> 00:36:50,040
able to see the logs as well which um

1074
00:36:48,300 --> 00:36:51,839
I've probably got time to show that but

1075
00:36:50,040 --> 00:36:55,140
I just have a dashboard allows you to

1076
00:36:51,839 --> 00:36:59,820
inspect pretty much inspect everything

1077
00:36:55,140 --> 00:37:02,280
all right um about 20 I'm 24 so what's

1078
00:36:59,820 --> 00:37:04,680
the speed up bit so um

1079
00:37:02,280 --> 00:37:06,359
so mqt is great it's um support wildly

1080
00:37:04,680 --> 00:37:09,480
support it's very lightweight um but we

1081
00:37:06,359 --> 00:37:12,119
can swap it out for those transports

1082
00:37:09,480 --> 00:37:13,260
um Services versus actors so swap back

1083
00:37:12,119 --> 00:37:14,820
into controls between the two but

1084
00:37:13,260 --> 00:37:16,260
Services pretty much all you do anything

1085
00:37:14,820 --> 00:37:18,480
like this it can be multi-thread there's

1086
00:37:16,260 --> 00:37:20,280
no constraints whereas actors um because

1087
00:37:18,480 --> 00:37:21,960
of the actor model have got constraints

1088
00:37:20,280 --> 00:37:23,760
that make them um you know easier to

1089
00:37:21,960 --> 00:37:25,020
reason with they have an inbox of

1090
00:37:23,760 --> 00:37:27,780
incoming messages which they walk

1091
00:37:25,020 --> 00:37:31,140
through one at a time and then and that

1092
00:37:27,780 --> 00:37:33,619
prevents um Sr race conditions and so on

1093
00:37:31,140 --> 00:37:35,760
so you can um have um

1094
00:37:33,619 --> 00:37:37,440
basically very lightweight in scale we'd

1095
00:37:35,760 --> 00:37:38,700
have one service per process which can

1096
00:37:37,440 --> 00:37:40,560
be quite heavy weight or you can have

1097
00:37:38,700 --> 00:37:42,960
multiple Services per process so just to

1098
00:37:40,560 --> 00:37:44,160
show you that very quickly

1099
00:37:42,960 --> 00:37:48,300
um I'll just go back to the hello world

1100
00:37:44,160 --> 00:37:50,460
example if I say make 20 of these

1101
00:37:48,300 --> 00:37:52,140
okay we've now got 20 hello world set

1102
00:37:50,460 --> 00:37:54,180
they're all in the same services in the

1103
00:37:52,140 --> 00:37:56,160
same process so my aim is to basically

1104
00:37:54,180 --> 00:37:58,380
be able to scale up to a million

1105
00:37:56,160 --> 00:37:59,339
services on a single machine

1106
00:37:58,380 --> 00:38:00,900
um the other thing I might be able to

1107
00:37:59,339 --> 00:38:02,099
quickly show is if you've got lifecycle

1108
00:38:00,900 --> 00:38:04,619
manager

1109
00:38:02,099 --> 00:38:06,599
site in that's that's just using the

1110
00:38:04,619 --> 00:38:07,980
livestock manager take create 10 clients

1111
00:38:06,599 --> 00:38:09,720
and you can see that they're on

1112
00:38:07,980 --> 00:38:12,000
different processes because their their

1113
00:38:09,720 --> 00:38:13,619
process ID is different so yeah so we

1114
00:38:12,000 --> 00:38:15,900
support those sort of things and these

1115
00:38:13,619 --> 00:38:17,280
these features are a little bit small

1116
00:38:15,900 --> 00:38:18,540
components so you can you can compose

1117
00:38:17,280 --> 00:38:20,700
into your service so we don't have to

1118
00:38:18,540 --> 00:38:22,440
write these caches and uh these are

1119
00:38:20,700 --> 00:38:25,140
creating other actors all the time just

1120
00:38:22,440 --> 00:38:26,780
uh it's all being done uh so finally

1121
00:38:25,140 --> 00:38:29,040
we've got uh pipelines

1122
00:38:26,780 --> 00:38:30,240
pipelines allow us to create data flow

1123
00:38:29,040 --> 00:38:32,820
architectures that's what we use for the

1124
00:38:30,240 --> 00:38:36,240
machine machine learning and also for

1125
00:38:32,820 --> 00:38:39,720
our gstream media media pipelines and

1126
00:38:36,240 --> 00:38:40,560
the pipelines are just actors so you

1127
00:38:39,720 --> 00:38:43,020
just then you just link them together

1128
00:38:40,560 --> 00:38:44,640
through the topics and also the uh the

1129
00:38:43,020 --> 00:38:46,020
pythons themselves are pipeline elements

1130
00:38:44,640 --> 00:38:48,000
and so you basically you can have a

1131
00:38:46,020 --> 00:38:50,099
hierarchy of pipelines that consist of

1132
00:38:48,000 --> 00:38:52,079
other pipelines and so on

1133
00:38:50,099 --> 00:38:53,460
so it ends up looking a little bit like

1134
00:38:52,079 --> 00:38:54,780
this so up the top left you've got a

1135
00:38:53,460 --> 00:38:56,820
process manager which can use those

1136
00:38:54,780 --> 00:38:58,320
green lines to create all the actors and

1137
00:38:56,820 --> 00:38:59,339
then they all they all they all then

1138
00:38:58,320 --> 00:39:01,500
discover them so they ought to go and

1139
00:38:59,339 --> 00:39:04,520
join the registrar and the pipeline

1140
00:39:01,500 --> 00:39:04,520
managing and hook them all together

1141
00:39:04,560 --> 00:39:08,460
or four thanks very much okay so

1142
00:39:06,900 --> 00:39:10,079
hopefully let's give you a bit of

1143
00:39:08,460 --> 00:39:12,540
insight to how I hope this Tech this

1144
00:39:10,079 --> 00:39:14,339
framework can um uh bring all these

1145
00:39:12,540 --> 00:39:17,280
different domains into the site into the

1146
00:39:14,339 --> 00:39:19,800
site that make them consistent to use as

1147
00:39:17,280 --> 00:39:23,099
a application developer so devices are

1148
00:39:19,800 --> 00:39:24,960
Services uh pipelines are linked

1149
00:39:23,099 --> 00:39:27,420
basically data flow of services Samsung

1150
00:39:24,960 --> 00:39:28,980
for the ml models and and other

1151
00:39:27,420 --> 00:39:30,420
Frameworks like rise and you'll probably

1152
00:39:28,980 --> 00:39:32,339
have done this for the play framework it

1153
00:39:30,420 --> 00:39:36,359
can just be abstractions uh transport

1154
00:39:32,339 --> 00:39:37,800
extraction so um we've hired that away

1155
00:39:36,359 --> 00:39:41,040
um so I was going to talk a little bit

1156
00:39:37,800 --> 00:39:43,560
about uh why our buy versus build

1157
00:39:41,040 --> 00:39:45,540
um it's often comes up as Engineers we

1158
00:39:43,560 --> 00:39:47,460
want to build we want to build new stuff

1159
00:39:45,540 --> 00:39:49,619
but you want to make sure that um what

1160
00:39:47,460 --> 00:39:51,420
you're doing is um you know there isn't

1161
00:39:49,619 --> 00:39:52,740
already an answer out there already that

1162
00:39:51,420 --> 00:39:54,480
um that the existing Solutions are

1163
00:39:52,740 --> 00:39:56,460
definitely not good enough and also

1164
00:39:54,480 --> 00:39:58,020
internally have um good support from

1165
00:39:56,460 --> 00:39:59,040
your business engineering team and

1166
00:39:58,020 --> 00:40:01,079
you're able to support yourself long

1167
00:39:59,040 --> 00:40:02,280
term for um you know supporting the

1168
00:40:01,079 --> 00:40:05,280
community and so on and you're actually

1169
00:40:02,280 --> 00:40:07,020
doing an advantage hopefully hopefully

1170
00:40:05,280 --> 00:40:09,420
we'll try to compensate it internally

1171
00:40:07,020 --> 00:40:11,940
here the other is open source that's um

1172
00:40:09,420 --> 00:40:14,339
quite a challenge to convince your um

1173
00:40:11,940 --> 00:40:15,240
the business people to um allow you to

1174
00:40:14,339 --> 00:40:16,680
do this

1175
00:40:15,240 --> 00:40:18,000
um there are there are benefits but also

1176
00:40:16,680 --> 00:40:19,740
concerns around our giving away

1177
00:40:18,000 --> 00:40:21,960
competitive advantage and losing control

1178
00:40:19,740 --> 00:40:24,240
of the terminology and so on but really

1179
00:40:21,960 --> 00:40:26,820
I believe it's a virtuous circle that um

1180
00:40:24,240 --> 00:40:28,800
they actually put things that businesses

1181
00:40:26,820 --> 00:40:30,540
pretty much have entirely depending upon

1182
00:40:28,800 --> 00:40:33,660
open source and if they don't contribute

1183
00:40:30,540 --> 00:40:35,280
back that's obviously not a good thing

1184
00:40:33,660 --> 00:40:38,160
so here's a quick list of what comes up

1185
00:40:35,280 --> 00:40:40,200
next so um incorporating Roz actually

1186
00:40:38,160 --> 00:40:41,760
documentation examples um there's few

1187
00:40:40,200 --> 00:40:43,980
people like Thomas figaroid himself

1188
00:40:41,760 --> 00:40:45,240
without documentation but go Thomas yeah

1189
00:40:43,980 --> 00:40:46,980
I was just one night okay hang on

1190
00:40:45,240 --> 00:40:48,660
there's a device I don't recognize that

1191
00:40:46,980 --> 00:40:52,920
device on my network because I'm

1192
00:40:48,660 --> 00:40:55,619
Thomas's swag patch however an Adelaide

1193
00:40:52,920 --> 00:40:57,000
um and uh just quickly um I've been

1194
00:40:55,619 --> 00:40:58,740
chatting to John

1195
00:40:57,000 --> 00:41:00,119
um about whether we should finally

1196
00:40:58,740 --> 00:41:02,280
tackle our open Hardware conference so

1197
00:41:00,119 --> 00:41:03,720
if anyone thinks this is a not a totally

1198
00:41:02,280 --> 00:41:05,040
insane idea please come and come and

1199
00:41:03,720 --> 00:41:07,800
chat to us I'm going to be heading over

1200
00:41:05,040 --> 00:41:09,660
to the open Harbor Summit and see how

1201
00:41:07,800 --> 00:41:11,280
how they do it in the other side on the

1202
00:41:09,660 --> 00:41:13,140
outside of the planet

1203
00:41:11,280 --> 00:41:14,820
uh anyway but it'll also be bringing in

1204
00:41:13,140 --> 00:41:16,380
other languages and uh and also

1205
00:41:14,820 --> 00:41:19,079
incorporate bringing in some of our

1206
00:41:16,380 --> 00:41:21,720
previous origin C projects

1207
00:41:19,079 --> 00:41:24,599
uh support your local hacker space

1208
00:41:21,720 --> 00:41:26,460
um they're great places and uh we lost

1209
00:41:24,599 --> 00:41:28,560
the future in during covert but I'm the

1210
00:41:26,460 --> 00:41:30,240
ones that have survived are fantastic so

1211
00:41:28,560 --> 00:41:32,099
there's probably one near you or you

1212
00:41:30,240 --> 00:41:33,420
should perhaps start one

1213
00:41:32,099 --> 00:41:34,619
um and just like to thank everyone you

1214
00:41:33,420 --> 00:41:36,599
know people have the people in the

1215
00:41:34,619 --> 00:41:38,400
community uh trying to join around at

1216
00:41:36,599 --> 00:41:40,020
work um you know likes Australia and all

1217
00:41:38,400 --> 00:41:41,220
your volunteers I think give you guys

1218
00:41:40,020 --> 00:41:43,320
some of the followers because you put us

1219
00:41:41,220 --> 00:41:46,440
all back together in

1220
00:41:43,320 --> 00:41:48,000
um yeah and also everyone here and of

1221
00:41:46,440 --> 00:41:50,640
course John

1222
00:41:48,000 --> 00:41:52,339
um yeah that's that's kind of it rush

1223
00:41:50,640 --> 00:41:56,400
through it at the end but um I got there

1224
00:41:52,339 --> 00:41:58,560
amazing so it's back in Brisbane I hope

1225
00:41:56,400 --> 00:41:59,700
you have John doing his uh Jedi Mind

1226
00:41:58,560 --> 00:42:01,260
Tricks

1227
00:41:59,700 --> 00:42:03,900
the

1228
00:42:01,260 --> 00:42:05,040
yeah so um I think it was about only a

1229
00:42:03,900 --> 00:42:06,540
couple months before the connect the

1230
00:42:05,040 --> 00:42:08,760
Microsoft connected come out and also

1231
00:42:06,540 --> 00:42:11,400
the uh the parrot drone one of the very

1232
00:42:08,760 --> 00:42:13,020
early um consumer drones and uh I think

1233
00:42:11,400 --> 00:42:14,579
jonos we were just chatting one day I

1234
00:42:13,020 --> 00:42:16,200
said what can we do for Brisbane and

1235
00:42:14,579 --> 00:42:17,820
it's like hey we'll smack these two new

1236
00:42:16,200 --> 00:42:18,720
things together

1237
00:42:17,820 --> 00:42:20,339
um

1238
00:42:18,720 --> 00:42:22,500
which uh it was actually part of John

1239
00:42:20,339 --> 00:42:24,780
John's broader broader vision for um

1240
00:42:22,500 --> 00:42:27,060
tough another one was um this is us

1241
00:42:24,780 --> 00:42:29,160
doing um presentation preparation the

1242
00:42:27,060 --> 00:42:31,320
night before in the hotel room

1243
00:42:29,160 --> 00:42:33,119
um also another chat typical I like

1244
00:42:31,320 --> 00:42:34,680
volunteers Hall of Brisbane gets trashed

1245
00:42:33,119 --> 00:42:37,260
by floods and in modern days they

1246
00:42:34,680 --> 00:42:40,760
completely redo a conference I think

1247
00:42:37,260 --> 00:42:40,760
this is to prank Lana Brindley Maybe

1248
00:42:41,720 --> 00:42:44,720
all right

1249
00:42:51,240 --> 00:42:55,800
but it was very tasty

1250
00:42:53,119 --> 00:42:57,839
all right unfortunately we don't have

1251
00:42:55,800 --> 00:43:00,420
time for questions I'm sure you'll be

1252
00:42:57,839 --> 00:43:01,560
able to catch Angie around and about I

1253
00:43:00,420 --> 00:43:02,720
just want to say thank you very much

1254
00:43:01,560 --> 00:43:07,099
thank you

1255
00:43:02,720 --> 00:43:07,099
for coming and talk to us thank you

1256
00:43:08,790 --> 00:43:12,829
[Applause]