1
00:00:06,320 --> 00:00:11,499
[Music]

2
00:00:16,800 --> 00:00:20,560
welcome back to the final talk today in

3
00:00:18,880 --> 00:00:23,359
the yuma theater

4
00:00:20,560 --> 00:00:25,840
this talk will be on nornet a unix and

5
00:00:23,359 --> 00:00:28,720
open source software-based international

6
00:00:25,840 --> 00:00:30,640
platform for networking research

7
00:00:28,720 --> 00:00:32,800
the talk will be given by thomas

8
00:00:30,640 --> 00:00:35,200
dropholtz

9
00:00:32,800 --> 00:00:37,600
uh who works as a chief research

10
00:00:35,200 --> 00:00:39,840
engineer for the similar metropolitan

11
00:00:37,600 --> 00:00:42,879
center for digital engineering simula

12
00:00:39,840 --> 00:00:45,039
met in oslo norway

13
00:00:42,879 --> 00:00:47,680
thomas will endeavor to allow time for

14
00:00:45,039 --> 00:00:51,199
questions at the end so please enter any

15
00:00:47,680 --> 00:00:54,239
questions in the tab in venulis

16
00:00:51,199 --> 00:00:56,640
take it away thomas

17
00:00:54,239 --> 00:01:00,320
thank you for the introduction then

18
00:00:56,640 --> 00:01:02,079
welcome to my presentation about our new

19
00:01:00,320 --> 00:01:05,040
net infrastructure

20
00:01:02,079 --> 00:01:07,439
first i will shortly introduce you the

21
00:01:05,040 --> 00:01:09,439
motivation of our work

22
00:01:07,439 --> 00:01:12,320
which is around multi-homing and

23
00:01:09,439 --> 00:01:13,280
multi-pass transport and then i come to

24
00:01:12,320 --> 00:01:16,080
some

25
00:01:13,280 --> 00:01:18,400
interesting details about our testbed

26
00:01:16,080 --> 00:01:21,119
infrastructure particularly with regard

27
00:01:18,400 --> 00:01:23,600
to virtual machines containers and

28
00:01:21,119 --> 00:01:25,759
multihold networking

29
00:01:23,600 --> 00:01:27,360
so what's the motivation of our work

30
00:01:25,759 --> 00:01:30,240
well let us look

31
00:01:27,360 --> 00:01:33,200
first at classic internet communication

32
00:01:30,240 --> 00:01:34,560
in the most basic property example

33
00:01:33,200 --> 00:01:36,880
a browsing

34
00:01:34,560 --> 00:01:39,840
when you use server browsing well you

35
00:01:36,880 --> 00:01:42,159
have a pc as a web browser having an

36
00:01:39,840 --> 00:01:44,880
ipv4 address

37
00:01:42,159 --> 00:01:47,360
this browser connects to app server

38
00:01:44,880 --> 00:01:49,840
which also has an ipv4 address

39
00:01:47,360 --> 00:01:51,520
and then you have over the tcp protocol

40
00:01:49,840 --> 00:01:54,720
with the 3bn check establish a

41
00:01:51,520 --> 00:01:56,799
connection with 3bn ventrac and

42
00:01:54,720 --> 00:01:58,000
we have classic client server

43
00:01:56,799 --> 00:02:01,600
communication

44
00:01:58,000 --> 00:02:04,000
this is the way most applications

45
00:02:01,600 --> 00:02:05,600
in the internet

46
00:02:04,000 --> 00:02:08,479
are working today

47
00:02:05,600 --> 00:02:10,879
what is a quite old

48
00:02:08,479 --> 00:02:12,000
way of communication

49
00:02:10,879 --> 00:02:14,800
it is

50
00:02:12,000 --> 00:02:17,120
likely it will work this way also for

51
00:02:14,800 --> 00:02:19,360
many other applications in the future

52
00:02:17,120 --> 00:02:20,400
but nowadays of course we have a bit

53
00:02:19,360 --> 00:02:22,319
more than

54
00:02:20,400 --> 00:02:23,840
such simple client server applications

55
00:02:22,319 --> 00:02:25,760
we have internet of things

56
00:02:23,840 --> 00:02:28,400
virtualization sensor networks cloud

57
00:02:25,760 --> 00:02:29,599
computing vehicle networks and many

58
00:02:28,400 --> 00:02:31,519
other things

59
00:02:29,599 --> 00:02:32,800
and these have of course

60
00:02:31,519 --> 00:02:35,680
more

61
00:02:32,800 --> 00:02:38,480
strict requirements on the communication

62
00:02:35,680 --> 00:02:41,599
particularly nowadays we not only have

63
00:02:38,480 --> 00:02:42,959
the ipv4 protocol we have ipv6 and with

64
00:02:41,599 --> 00:02:46,319
rpv6

65
00:02:42,959 --> 00:02:49,760
usually we have multiple addresses for

66
00:02:46,319 --> 00:02:52,640
interface and even multiple interfaces

67
00:02:49,760 --> 00:02:56,160
and usually today of course we have ipv4

68
00:02:52,640 --> 00:02:58,959
ipv6 dual stack devices so many

69
00:02:56,160 --> 00:03:01,120
computers today have both ipv4 address

70
00:02:58,959 --> 00:03:03,760
and ipv6 unless

71
00:03:01,120 --> 00:03:05,200
and today of course we have also

72
00:03:03,760 --> 00:03:08,000
mobility

73
00:03:05,200 --> 00:03:11,280
so probably everybody of you has a

74
00:03:08,000 --> 00:03:13,120
smartphone in his pocket

75
00:03:11,280 --> 00:03:16,000
it's quite convenient device so if you

76
00:03:13,120 --> 00:03:19,519
are at home you're using wi-fi

77
00:03:16,000 --> 00:03:20,239
if you go outside the device changes

78
00:03:19,519 --> 00:03:22,720
to

79
00:03:20,239 --> 00:03:23,519
for example here in oslo it is uh five

80
00:03:22,720 --> 00:03:24,879
three

81
00:03:23,519 --> 00:03:28,560
maybe um

82
00:03:24,879 --> 00:03:30,640
also 4g 3g depending on where you are

83
00:03:28,560 --> 00:03:32,400
and of course the smartphone would even

84
00:03:30,640 --> 00:03:34,080
have third interface

85
00:03:32,400 --> 00:03:36,480
or bluetooth

86
00:03:34,080 --> 00:03:39,200
same as classically

87
00:03:36,480 --> 00:03:41,840
a router because the router classically

88
00:03:39,200 --> 00:03:44,319
had these two interfaces

89
00:03:41,840 --> 00:03:46,799
and of course also depending on what

90
00:03:44,319 --> 00:03:49,760
kind of laptop computer you have we have

91
00:03:46,799 --> 00:03:52,879
also multiple interfaces like ethernet

92
00:03:49,760 --> 00:03:54,480
wi-fi and maybe even mobile broadband

93
00:03:52,879 --> 00:03:56,720
built-in

94
00:03:54,480 --> 00:03:58,159
so what can you do with multiple

95
00:03:56,720 --> 00:04:00,959
interfaces

96
00:03:58,159 --> 00:04:02,560
well you have the property that you are

97
00:04:00,959 --> 00:04:04,239
multi-homed

98
00:04:02,560 --> 00:04:06,319
if you have multiple other resistances

99
00:04:04,239 --> 00:04:08,080
you are connected to multiple networks

100
00:04:06,319 --> 00:04:11,040
so particularly

101
00:04:08,080 --> 00:04:13,120
you can connect to

102
00:04:11,040 --> 00:04:15,519
more than one internet service provider

103
00:04:13,120 --> 00:04:18,320
to have redundancy

104
00:04:15,519 --> 00:04:20,639
so look at this uh

105
00:04:18,320 --> 00:04:23,199
user of material here

106
00:04:20,639 --> 00:04:26,600
looks not that convinced about this

107
00:04:23,199 --> 00:04:28,720
redundancy what's the problem with

108
00:04:26,600 --> 00:04:31,759
redundancy is nice if something goes

109
00:04:28,720 --> 00:04:34,479
wrong then you still have connectivity

110
00:04:31,759 --> 00:04:37,600
but if nothing goes wrong

111
00:04:34,479 --> 00:04:40,000
well if you are connected to two isps

112
00:04:37,600 --> 00:04:42,240
each month you receive two internet

113
00:04:40,000 --> 00:04:44,080
bills

114
00:04:42,240 --> 00:04:46,240
for example in australia maybe an

115
00:04:44,080 --> 00:04:48,240
internet bill from optos another one

116
00:04:46,240 --> 00:04:49,360
from telstra

117
00:04:48,240 --> 00:04:50,320
well

118
00:04:49,360 --> 00:04:52,479
paying

119
00:04:50,320 --> 00:04:54,000
double the amount necessary is not

120
00:04:52,479 --> 00:04:55,440
really nice

121
00:04:54,000 --> 00:04:59,520
obviously

122
00:04:55,440 --> 00:05:00,479
so um in 99.9 something percent of the

123
00:04:59,520 --> 00:05:02,400
time

124
00:05:00,479 --> 00:05:04,400
nothing goes wrong and you do not

125
00:05:02,400 --> 00:05:05,840
appreciate redundancy

126
00:05:04,400 --> 00:05:08,639
so can you do something with a

127
00:05:05,840 --> 00:05:08,639
redundancy

128
00:05:08,800 --> 00:05:13,440
uh when everything is okay

129
00:05:12,000 --> 00:05:16,080
there is the possibility to do

130
00:05:13,440 --> 00:05:17,360
multi-pass transport if you have two rsp

131
00:05:16,080 --> 00:05:19,680
connections

132
00:05:17,360 --> 00:05:21,120
why not trust utilizing them both

133
00:05:19,680 --> 00:05:24,160
simultaneously

134
00:05:21,120 --> 00:05:25,840
then at least if everything is okay

135
00:05:24,160 --> 00:05:28,560
usually

136
00:05:25,840 --> 00:05:30,800
you get a higher bandwidth

137
00:05:28,560 --> 00:05:31,600
this is of course nice

138
00:05:30,800 --> 00:05:33,840
so

139
00:05:31,600 --> 00:05:35,039
with tcp of course classical tcp

140
00:05:33,840 --> 00:05:37,199
communication

141
00:05:35,039 --> 00:05:39,440
is established between two ipv4

142
00:05:37,199 --> 00:05:42,240
addresses this does not work but there

143
00:05:39,440 --> 00:05:44,320
is an extension multi-pass sap

144
00:05:42,240 --> 00:05:46,880
which supports multi-homing and

145
00:05:44,320 --> 00:05:48,720
particularly multi-pass transport

146
00:05:46,880 --> 00:05:51,280
and then it works

147
00:05:48,720 --> 00:05:53,759
also if you are using the sctp protocol

148
00:05:51,280 --> 00:05:55,120
which already has built-in support for

149
00:05:53,759 --> 00:05:57,840
multi-homing

150
00:05:55,120 --> 00:06:01,520
you can add an extension from current

151
00:05:57,840 --> 00:06:04,479
multi-path transfer for satp cmp-sdp

152
00:06:01,520 --> 00:06:07,280
to also have multi-path transport

153
00:06:04,479 --> 00:06:08,479
so in either case you get multi-pass

154
00:06:07,280 --> 00:06:11,120
transport

155
00:06:08,479 --> 00:06:13,280
so you can utilize multiple network

156
00:06:11,120 --> 00:06:15,360
connections simultaneously which is a

157
00:06:13,280 --> 00:06:18,000
quite nice thing and therefore it is

158
00:06:15,360 --> 00:06:20,720
also a topic in research and

159
00:06:18,000 --> 00:06:22,240
standardization

160
00:06:20,720 --> 00:06:24,880
so

161
00:06:22,240 --> 00:06:27,120
let me shortly go into a little bit more

162
00:06:24,880 --> 00:06:31,759
detail about multi-parts transport

163
00:06:27,120 --> 00:06:33,840
particularly with mpdcp or cmkcp how

164
00:06:31,759 --> 00:06:35,840
does it work

165
00:06:33,840 --> 00:06:37,840
looking at the

166
00:06:35,840 --> 00:06:39,759
osa motor stack

167
00:06:37,840 --> 00:06:41,120
you have the application

168
00:06:39,759 --> 00:06:44,160
protocol

169
00:06:41,120 --> 00:06:47,039
in the application layer on top

170
00:06:44,160 --> 00:06:49,039
of of course the kernel

171
00:06:47,039 --> 00:06:50,319
which implements the transport layer

172
00:06:49,039 --> 00:06:53,919
that is

173
00:06:50,319 --> 00:06:56,800
mptcp or cmt-sctp as part of the

174
00:06:53,919 --> 00:06:58,800
implementation of dcp or sctp

175
00:06:56,800 --> 00:07:01,360
and application protocol communicates

176
00:06:58,800 --> 00:07:02,720
with the transport layer over the socket

177
00:07:01,360 --> 00:07:04,720
api

178
00:07:02,720 --> 00:07:07,120
that is

179
00:07:04,720 --> 00:07:08,400
the user space communicates with the

180
00:07:07,120 --> 00:07:10,479
kernel

181
00:07:08,400 --> 00:07:12,160
over this socket api

182
00:07:10,479 --> 00:07:15,919
and network layer of course you have

183
00:07:12,160 --> 00:07:16,960
ipv4 as well as rpv6 so as i've said

184
00:07:15,919 --> 00:07:20,240
usually

185
00:07:16,960 --> 00:07:22,160
you have a dual stack today

186
00:07:20,240 --> 00:07:25,680
so now the transfer protocol can

187
00:07:22,160 --> 00:07:28,160
establish paths or so-called subfloors

188
00:07:25,680 --> 00:07:31,599
over the underlying network protocols

189
00:07:28,160 --> 00:07:33,680
that is subfloors over ipv4 and ipv6 you

190
00:07:31,599 --> 00:07:36,800
can even mix

191
00:07:33,680 --> 00:07:38,080
some flows with ipv4 and rpv6 so you can

192
00:07:36,800 --> 00:07:39,599
have

193
00:07:38,080 --> 00:07:41,280
uh

194
00:07:39,599 --> 00:07:44,479
mp3 connection

195
00:07:41,280 --> 00:07:45,840
both using ipv4 and ipv6 in the most

196
00:07:44,479 --> 00:07:47,919
extreme case

197
00:07:45,840 --> 00:07:51,199
you can even start a connection in a

198
00:07:47,919 --> 00:07:52,400
pure ipv4 network later at ipv6

199
00:07:51,199 --> 00:07:54,319
addresses

200
00:07:52,400 --> 00:07:57,520
and finally remove

201
00:07:54,319 --> 00:08:00,800
the rpv4 addresses so you can upgrade

202
00:07:57,520 --> 00:08:03,360
your system from ipv4 to fpv6

203
00:08:00,800 --> 00:08:05,759
without breaking connections

204
00:08:03,360 --> 00:08:08,879
it's of course quite extreme case but at

205
00:08:05,759 --> 00:08:12,560
least it is possible of course

206
00:08:08,879 --> 00:08:15,280
so now we have subflows but how to

207
00:08:12,560 --> 00:08:18,080
put user data payload data onto the

208
00:08:15,280 --> 00:08:19,680
surface and this decision is made by the

209
00:08:18,080 --> 00:08:23,120
scheduler

210
00:08:19,680 --> 00:08:26,319
also far so simple but they of course

211
00:08:23,120 --> 00:08:29,120
are challenges for multi-pass transport

212
00:08:26,319 --> 00:08:30,160
one of the shelves are particularly that

213
00:08:29,120 --> 00:08:32,479
parts

214
00:08:30,160 --> 00:08:33,599
in the network some frozen network may

215
00:08:32,479 --> 00:08:35,839
overlap

216
00:08:33,599 --> 00:08:37,360
particularly if they share the same

217
00:08:35,839 --> 00:08:38,719
bottleneck link somewhere in the

218
00:08:37,360 --> 00:08:41,120
internet

219
00:08:38,719 --> 00:08:43,360
if we have multiple subfloors running

220
00:08:41,120 --> 00:08:46,399
over the same share bottleneck

221
00:08:43,360 --> 00:08:48,240
compared to a concurrent tcp connection

222
00:08:46,399 --> 00:08:50,560
you get of course more bandwidth which

223
00:08:48,240 --> 00:08:52,839
is unfair so fairness is one challenge

224
00:08:50,560 --> 00:08:54,399
and the other challenge is of course

225
00:08:52,839 --> 00:08:56,320
scheduling

226
00:08:54,399 --> 00:08:57,839
when you think of subfloors maybe you

227
00:08:56,320 --> 00:09:00,160
think of

228
00:08:57,839 --> 00:09:03,200
subfloors having very similar

229
00:09:00,160 --> 00:09:05,440
characteristics regarding bandwidth

230
00:09:03,200 --> 00:09:07,680
delay and packet loss

231
00:09:05,440 --> 00:09:10,320
but of course usually in the internet

232
00:09:07,680 --> 00:09:12,640
such sub-rows are quite dissimilar

233
00:09:10,320 --> 00:09:14,240
and then it makes sense to very

234
00:09:12,640 --> 00:09:16,240
carefully think about

235
00:09:14,240 --> 00:09:18,640
which subfloors to use and which

236
00:09:16,240 --> 00:09:21,120
sufferers do not use particularly to get

237
00:09:18,640 --> 00:09:25,200
a good performance performance maybe

238
00:09:21,120 --> 00:09:27,360
bandwidth but also some applications are

239
00:09:25,200 --> 00:09:30,240
latency sensitive

240
00:09:27,360 --> 00:09:31,760
so maybe then it's not useful to use a

241
00:09:30,240 --> 00:09:33,920
subfloor where you can get the highest

242
00:09:31,760 --> 00:09:35,600
low port but where you have the lowest

243
00:09:33,920 --> 00:09:37,440
contraption

244
00:09:35,600 --> 00:09:40,160
so a lot of challenges

245
00:09:37,440 --> 00:09:41,279
and of course the research for this is

246
00:09:40,160 --> 00:09:43,680
necessary

247
00:09:41,279 --> 00:09:46,240
but let me first introduce you how to

248
00:09:43,680 --> 00:09:50,000
use mptcp

249
00:09:46,240 --> 00:09:51,920
well it's defined as an rc but with some

250
00:09:50,000 --> 00:09:55,360
open topics still

251
00:09:51,920 --> 00:09:57,839
but with mptcp you can make a system

252
00:09:55,360 --> 00:10:00,000
capable of multi-pass transport

253
00:09:57,839 --> 00:10:02,240
by extending the

254
00:10:00,000 --> 00:10:05,600
old tcp protocol

255
00:10:02,240 --> 00:10:08,399
and particularly nice feature of mptcp

256
00:10:05,600 --> 00:10:11,360
is that is quite backwards compatible so

257
00:10:08,399 --> 00:10:13,120
on the wire communications and btcp

258
00:10:11,360 --> 00:10:14,720
look like

259
00:10:13,120 --> 00:10:17,920
normal

260
00:10:14,720 --> 00:10:21,839
tcp connections so the subfloors look

261
00:10:17,920 --> 00:10:25,120
like regular tcp connections that is

262
00:10:21,839 --> 00:10:28,000
mptcp will also work usually over middle

263
00:10:25,120 --> 00:10:29,920
boxes that particularly means network

264
00:10:28,000 --> 00:10:31,760
and portables installation as well as

265
00:10:29,920 --> 00:10:34,000
firewalls and such

266
00:10:31,760 --> 00:10:38,079
so it's quite nice because of this

267
00:10:34,000 --> 00:10:40,079
compatibility so it is easy to use

268
00:10:38,079 --> 00:10:41,680
in existing networks

269
00:10:40,079 --> 00:10:44,320
and the nice thing is there is a quite

270
00:10:41,680 --> 00:10:47,200
good implementation for linux available

271
00:10:44,320 --> 00:10:48,640
from university catholic dilemma in

272
00:10:47,200 --> 00:10:52,240
belgium

273
00:10:48,640 --> 00:10:54,959
and also swinburne university has some

274
00:10:52,240 --> 00:10:56,560
years ago started an implementation for

275
00:10:54,959 --> 00:10:59,040
freebase day

276
00:10:56,560 --> 00:11:01,519
so there are two open source

277
00:10:59,040 --> 00:11:03,519
implementations available

278
00:11:01,519 --> 00:11:05,600
so um

279
00:11:03,519 --> 00:11:07,680
how can you now use

280
00:11:05,600 --> 00:11:10,480
it particularly on the linux because we

281
00:11:07,680 --> 00:11:14,320
are here at the linux once

282
00:11:10,480 --> 00:11:17,200
well um obviously you need an empty dcp

283
00:11:14,320 --> 00:11:19,839
capable curl but is not a big deal you

284
00:11:17,200 --> 00:11:22,000
can just compile one

285
00:11:19,839 --> 00:11:23,839
so let me give an example

286
00:11:22,000 --> 00:11:27,440
we have a computer which is connected to

287
00:11:23,839 --> 00:11:30,640
the internet over an internet

288
00:11:27,440 --> 00:11:32,480
open standard interface eth0

289
00:11:30,640 --> 00:11:34,880
and obviously we want multi-homing and

290
00:11:32,480 --> 00:11:35,920
multi-pass transport so we make a second

291
00:11:34,880 --> 00:11:39,680
connection

292
00:11:35,920 --> 00:11:43,279
over eth1 to a second isp

293
00:11:39,680 --> 00:11:45,200
so does it now work multi-pass transport

294
00:11:43,279 --> 00:11:47,680
with mptcp

295
00:11:45,200 --> 00:11:50,399
no there is a bit missing yet

296
00:11:47,680 --> 00:11:52,639
the first problem is of course if we

297
00:11:50,399 --> 00:11:54,959
have such a setup we have

298
00:11:52,639 --> 00:11:56,560
probably a default route over one of the

299
00:11:54,959 --> 00:11:57,920
rsps

300
00:11:56,560 --> 00:12:01,040
and

301
00:11:57,920 --> 00:12:04,480
of course then the linux roding stack

302
00:12:01,040 --> 00:12:07,360
uses the route which is lowest metric

303
00:12:04,480 --> 00:12:10,560
which in this case would be over isp1 so

304
00:12:07,360 --> 00:12:14,639
all traffic would go over isp1

305
00:12:10,560 --> 00:12:16,880
while sp2 would not be utilized

306
00:12:14,639 --> 00:12:18,959
of course not what we want we have to do

307
00:12:16,880 --> 00:12:21,440
something in addition

308
00:12:18,959 --> 00:12:22,320
and the solution under linux is routing

309
00:12:21,440 --> 00:12:25,440
holes

310
00:12:22,320 --> 00:12:28,320
so we in fact we have to define separate

311
00:12:25,440 --> 00:12:30,240
routing tables for each isp

312
00:12:28,320 --> 00:12:32,639
sb1 and the isp2

313
00:12:30,240 --> 00:12:35,120
and then select for an outgoing packet

314
00:12:32,639 --> 00:12:37,200
according to the source address

315
00:12:35,120 --> 00:12:40,240
which routing table to use

316
00:12:37,200 --> 00:12:42,160
so let me explain this in a

317
00:12:40,240 --> 00:12:44,480
small example

318
00:12:42,160 --> 00:12:47,120
if again our computer and we now

319
00:12:44,480 --> 00:12:49,920
configure an rp address

320
00:12:47,120 --> 00:12:54,240
for ethernet 0

321
00:12:49,920 --> 00:12:57,600
10 1 142 as slash 24 network

322
00:12:54,240 --> 00:12:59,519
and of course we define a default route

323
00:12:57,600 --> 00:13:01,360
and this goes of course in the main

324
00:12:59,519 --> 00:13:04,720
routing table

325
00:13:01,360 --> 00:13:08,240
this is as classic network configuration

326
00:13:04,720 --> 00:13:11,279
under is but now as i've said we want an

327
00:13:08,240 --> 00:13:13,040
additional routing table for isb1

328
00:13:11,279 --> 00:13:14,880
so we define

329
00:13:13,040 --> 00:13:17,360
a new table

330
00:13:14,880 --> 00:13:20,880
and since the new table number one is

331
00:13:17,360 --> 00:13:23,920
empty we have also to define routing for

332
00:13:20,880 --> 00:13:27,600
the local scope

333
00:13:23,920 --> 00:13:30,639
and of course then at the default route

334
00:13:27,600 --> 00:13:32,880
and the same we can do for ethernet one

335
00:13:30,639 --> 00:13:38,079
for the second rsp in this case we use

336
00:13:32,880 --> 00:13:40,399
the address 172 16 142 as again as

337
00:13:38,079 --> 00:13:42,800
slash 24 network

338
00:13:40,399 --> 00:13:46,000
and create a second table table number

339
00:13:42,800 --> 00:13:47,519
two with the local scope as well as our

340
00:13:46,000 --> 00:13:49,760
default route

341
00:13:47,519 --> 00:13:51,279
so now we have in addition to the main

342
00:13:49,760 --> 00:13:53,360
routing table

343
00:13:51,279 --> 00:13:58,079
two additional tables table one and

344
00:13:53,360 --> 00:13:58,079
table two for isb1 and the iceberg two

345
00:13:58,560 --> 00:14:03,440
so now missing is just a selector so we

346
00:14:02,560 --> 00:14:06,399
add

347
00:14:03,440 --> 00:14:10,079
two routing holes so if a packet is

348
00:14:06,399 --> 00:14:12,160
coming from the source address 10 142 we

349
00:14:10,079 --> 00:14:16,160
use table number one

350
00:14:12,160 --> 00:14:20,800
and if the packet is coming from 172 16

351
00:14:16,160 --> 00:14:20,800
142 we use table number two

352
00:14:21,120 --> 00:14:24,160
table number one and terminal number two

353
00:14:22,800 --> 00:14:27,600
are numbers

354
00:14:24,160 --> 00:14:30,320
if you like you can also use etc upo 2

355
00:14:27,600 --> 00:14:32,480
arctic tables to give these tables names

356
00:14:30,320 --> 00:14:36,160
for example if you are connected

357
00:14:32,480 --> 00:14:40,560
to optos and telstra you can

358
00:14:36,160 --> 00:14:40,560
name the tables optos and telstra

359
00:14:43,040 --> 00:14:48,320
us look yeah we have configured now we

360
00:14:45,600 --> 00:14:51,040
have the routing holds particularly the

361
00:14:48,320 --> 00:14:52,480
lookup for table number one and table

362
00:14:51,040 --> 00:14:53,920
number two according to the source

363
00:14:52,480 --> 00:14:56,320
address

364
00:14:53,920 --> 00:14:59,839
and as i've said these name tables can

365
00:14:56,320 --> 00:15:01,920
also be named if it is more convenient

366
00:14:59,839 --> 00:15:04,160
usually these tables just have numbers

367
00:15:01,920 --> 00:15:07,600
and internally of course these

368
00:15:04,160 --> 00:15:09,760
tables have just numbers but there is a

369
00:15:07,600 --> 00:15:11,279
mapping of symbolical names to these

370
00:15:09,760 --> 00:15:14,000
numbers

371
00:15:11,279 --> 00:15:15,120
and we have of course

372
00:15:14,000 --> 00:15:17,519
now

373
00:15:15,120 --> 00:15:19,600
our routing table number one

374
00:15:17,519 --> 00:15:21,760
as i've said with the local scope as

375
00:15:19,600 --> 00:15:23,839
well as the default route and the same

376
00:15:21,760 --> 00:15:26,160
for table number two with local scope

377
00:15:23,839 --> 00:15:28,639
and default route in this case for isp

378
00:15:26,160 --> 00:15:29,839
number two

379
00:15:28,639 --> 00:15:31,920
so

380
00:15:29,839 --> 00:15:34,240
and of course finally we have the main

381
00:15:31,920 --> 00:15:36,959
routing table the default this is the

382
00:15:34,240 --> 00:15:38,480
standard routing table used for all

383
00:15:36,959 --> 00:15:42,560
other packets

384
00:15:38,480 --> 00:15:44,399
not matching to one of our two new

385
00:15:42,560 --> 00:15:46,800
rb holes

386
00:15:44,399 --> 00:15:48,240
so now our working can work

387
00:15:46,800 --> 00:15:49,920
with um

388
00:15:48,240 --> 00:15:51,680
rpv4

389
00:15:49,920 --> 00:15:53,279
questions of course nowadays we have

390
00:15:51,680 --> 00:15:55,600
ipv6

391
00:15:53,279 --> 00:15:57,920
usually in addition to ipv4 does it work

392
00:15:55,600 --> 00:15:59,120
with rpg v6 as well

393
00:15:57,920 --> 00:16:01,519
yes of course

394
00:15:59,120 --> 00:16:02,720
you can do the same also for ipv6 that

395
00:16:01,519 --> 00:16:03,600
is

396
00:16:02,720 --> 00:16:07,199
get

397
00:16:03,600 --> 00:16:08,880
an rgb6 address for ethernet 0

398
00:16:07,199 --> 00:16:11,680
create

399
00:16:08,880 --> 00:16:13,120
also the rounding table number one for

400
00:16:11,680 --> 00:16:15,600
ipv6

401
00:16:13,120 --> 00:16:17,839
important to mention is of course ipv4

402
00:16:15,600 --> 00:16:21,199
and ipv6 are different protocols they

403
00:16:17,839 --> 00:16:24,240
have both their separate routing tables

404
00:16:21,199 --> 00:16:26,000
so in this case of course this is ipp6

405
00:16:24,240 --> 00:16:28,320
table number one

406
00:16:26,000 --> 00:16:30,959
and before we have configured everything

407
00:16:28,320 --> 00:16:33,600
for ipv4 table number one

408
00:16:30,959 --> 00:16:35,759
but otherwise it is the same as before

409
00:16:33,600 --> 00:16:39,440
for ipv4 as we can

410
00:16:35,759 --> 00:16:41,040
define our two routing tables for rb6

411
00:16:39,440 --> 00:16:44,560
and also

412
00:16:41,040 --> 00:16:47,199
set ipv6 routing rules to select cable

413
00:16:44,560 --> 00:16:49,519
number one or table number two according

414
00:16:47,199 --> 00:16:51,120
to the source address of an outgoing

415
00:16:49,519 --> 00:16:54,240
packet

416
00:16:51,120 --> 00:16:55,519
so now depending on the configuration we

417
00:16:54,240 --> 00:16:58,079
have made now

418
00:16:55,519 --> 00:16:59,759
mptcp or if you have an implementation

419
00:16:58,079 --> 00:17:02,079
of cmt stp

420
00:16:59,759 --> 00:17:04,079
can make use

421
00:17:02,079 --> 00:17:07,280
of this new outing and fbtcp

422
00:17:04,079 --> 00:17:10,559
particularly with full mesh pass manager

423
00:17:07,280 --> 00:17:14,880
will then of course use all paths

424
00:17:10,559 --> 00:17:16,160
simultaneously this of course can be

425
00:17:14,880 --> 00:17:17,679
tested

426
00:17:16,160 --> 00:17:20,319
by booting

427
00:17:17,679 --> 00:17:22,559
an mptcp enabled kernel particular

428
00:17:20,319 --> 00:17:24,480
useful mesh pass manager which has to be

429
00:17:22,559 --> 00:17:26,720
configured in the settings of the kernel

430
00:17:24,480 --> 00:17:27,520
to be compiled

431
00:17:26,720 --> 00:17:29,039
and

432
00:17:27,520 --> 00:17:31,440
then of course you can

433
00:17:29,039 --> 00:17:34,240
test this with ping and particularly

434
00:17:31,440 --> 00:17:36,240
tools like for example network meter

435
00:17:34,240 --> 00:17:37,840
i'll come to this tool

436
00:17:36,240 --> 00:17:40,559
later again

437
00:17:37,840 --> 00:17:43,919
and particularly quite recommended if

438
00:17:40,559 --> 00:17:46,720
you learn about mptcp and multi-pass

439
00:17:43,919 --> 00:17:48,720
transport it's quite convenient to also

440
00:17:46,720 --> 00:17:51,280
observe the packet flow

441
00:17:48,720 --> 00:17:54,080
and particularly this is helpful

442
00:17:51,280 --> 00:17:55,760
if something is not working as expected

443
00:17:54,080 --> 00:17:56,480
for debugging

444
00:17:55,760 --> 00:17:59,280
so

445
00:17:56,480 --> 00:18:02,799
you can use wireshark or command to

446
00:17:59,280 --> 00:18:05,520
t-shark to do this

447
00:18:02,799 --> 00:18:08,160
so far the idea of multi-pass transport

448
00:18:05,520 --> 00:18:09,760
particularly with mptcp

449
00:18:08,160 --> 00:18:12,160
and of course

450
00:18:09,760 --> 00:18:14,559
i've suffered

451
00:18:12,160 --> 00:18:15,679
research around multi-path transport is

452
00:18:14,559 --> 00:18:17,919
necessary

453
00:18:15,679 --> 00:18:21,200
so how to do research

454
00:18:17,919 --> 00:18:23,520
well um years ago i worked for a

455
00:18:21,200 --> 00:18:25,280
university in germany at the university

456
00:18:23,520 --> 00:18:28,160
of glensburg essen

457
00:18:25,280 --> 00:18:30,240
and i my colleague wanted to do research

458
00:18:28,160 --> 00:18:32,160
on multi-pass transport particularly

459
00:18:30,240 --> 00:18:34,559
with cmt satp

460
00:18:32,160 --> 00:18:37,520
but also with mpt sigp

461
00:18:34,559 --> 00:18:40,400
so we asked our system to give us

462
00:18:37,520 --> 00:18:43,600
hardware we had some old pcs

463
00:18:40,400 --> 00:18:45,360
some old switches a lot of cables and

464
00:18:43,600 --> 00:18:46,559
some network cards we put everything

465
00:18:45,360 --> 00:18:48,480
together

466
00:18:46,559 --> 00:18:50,160
and obviously

467
00:18:48,480 --> 00:18:53,280
did not work

468
00:18:50,160 --> 00:18:55,440
um the setup was quite cheap was in fact

469
00:18:53,280 --> 00:18:57,919
for free we did not need a special

470
00:18:55,440 --> 00:19:00,559
budget because the old hardware was all

471
00:18:57,919 --> 00:19:00,559
literally there

472
00:19:01,840 --> 00:19:06,640
but it was only treatment of paper

473
00:19:04,480 --> 00:19:09,200
so particularly this hardware had

474
00:19:06,640 --> 00:19:11,200
sometimes issues particularly

475
00:19:09,200 --> 00:19:13,520
we observe that the network cards

476
00:19:11,200 --> 00:19:16,640
sometimes switch from 100 megabits per

477
00:19:13,520 --> 00:19:19,360
second to 10 million per second just for

478
00:19:16,640 --> 00:19:23,440
a few milliseconds and then switch back

479
00:19:19,360 --> 00:19:24,240
but the paths or some flows

480
00:19:23,440 --> 00:19:26,559
were

481
00:19:24,240 --> 00:19:27,760
assumed to be

482
00:19:26,559 --> 00:19:29,440
having

483
00:19:27,760 --> 00:19:31,679
very similar characteristics

484
00:19:29,440 --> 00:19:35,200
particularly same bandwidth of course

485
00:19:31,679 --> 00:19:37,280
became extremely dissimilar when a card

486
00:19:35,200 --> 00:19:39,679
switched to 10 megabits per second it's

487
00:19:37,280 --> 00:19:41,440
just one tenth of the bandwidth on one

488
00:19:39,679 --> 00:19:43,600
pass

489
00:19:41,440 --> 00:19:45,200
so um this trigger of course packs the

490
00:19:43,600 --> 00:19:49,440
implementation so we did a lot of

491
00:19:45,200 --> 00:19:50,240
debugging of sctp in the freebsd kernel

492
00:19:49,440 --> 00:19:51,440
but

493
00:19:50,240 --> 00:19:53,520
finally

494
00:19:51,440 --> 00:19:54,799
we were able to find these bugs fix

495
00:19:53,520 --> 00:19:56,720
these bugs

496
00:19:54,799 --> 00:19:59,120
and prove that

497
00:19:56,720 --> 00:20:01,919
simulations we made in advance with the

498
00:19:59,120 --> 00:20:04,559
simulation model were actually quite

499
00:20:01,919 --> 00:20:09,039
well modeling the reality we were able

500
00:20:04,559 --> 00:20:10,720
to create in this small testbed center

501
00:20:09,039 --> 00:20:12,400
and of course we

502
00:20:10,720 --> 00:20:15,039
were already very committed to open

503
00:20:12,400 --> 00:20:16,960
source community so we provided of

504
00:20:15,039 --> 00:20:18,320
course all these bug fixes back to the

505
00:20:16,960 --> 00:20:20,480
freebsd

506
00:20:18,320 --> 00:20:23,120
satb community

507
00:20:20,480 --> 00:20:24,880
to make the implementation in freebsd

508
00:20:23,120 --> 00:20:27,039
better

509
00:20:24,880 --> 00:20:30,640
and as part of this i wrote they

510
00:20:27,039 --> 00:20:34,080
mentioned two networth meter to do

511
00:20:30,640 --> 00:20:36,000
protocol performance comparisons between

512
00:20:34,080 --> 00:20:39,039
tcp smtp

513
00:20:36,000 --> 00:20:41,120
udp and dccp with the different

514
00:20:39,039 --> 00:20:44,559
parameter settings so quite useful tool

515
00:20:41,120 --> 00:20:47,120
which is also nowadays available

516
00:20:44,559 --> 00:20:49,600
in debian linux ubuntu linux and other

517
00:20:47,120 --> 00:20:51,840
distributions

518
00:20:49,600 --> 00:20:54,240
well this small setup had of course a

519
00:20:51,840 --> 00:20:56,720
lot of learning effects and new ideas

520
00:20:54,240 --> 00:20:57,760
around multi-pass transport and research

521
00:20:56,720 --> 00:21:01,039
on this

522
00:20:57,760 --> 00:21:02,799
so the idea was of course then well we

523
00:21:01,039 --> 00:21:06,240
did research on

524
00:21:02,799 --> 00:21:09,440
internet protocols but obviously such a

525
00:21:06,240 --> 00:21:11,679
testbed is an internet setup

526
00:21:09,440 --> 00:21:13,919
so let's go to the internet

527
00:21:11,679 --> 00:21:16,159
so we were able to convince colleagues

528
00:21:13,919 --> 00:21:18,320
at two other universities

529
00:21:16,159 --> 00:21:21,120
university of berlin science in book

530
00:21:18,320 --> 00:21:22,720
strandford and germany as well as hanan

531
00:21:21,120 --> 00:21:24,799
university in china

532
00:21:22,720 --> 00:21:26,799
to also create

533
00:21:24,799 --> 00:21:29,039
testbed setups

534
00:21:26,799 --> 00:21:30,480
connected at each site to two different

535
00:21:29,039 --> 00:21:33,440
isps

536
00:21:30,480 --> 00:21:36,080
so we had an intercontinental setup with

537
00:21:33,440 --> 00:21:37,200
three sites to germany to try the one in

538
00:21:36,080 --> 00:21:40,400
china

539
00:21:37,200 --> 00:21:42,400
and all these sites had two isps with

540
00:21:40,400 --> 00:21:44,640
rb4 and

541
00:21:42,400 --> 00:21:46,960
mostly also ipv6

542
00:21:44,640 --> 00:21:49,360
so quite interesting we had a small

543
00:21:46,960 --> 00:21:52,080
internet setup and we were able to do

544
00:21:49,360 --> 00:21:53,679
more evaluation work on cmt satp as well

545
00:21:52,080 --> 00:21:56,159
as mptcp

546
00:21:53,679 --> 00:21:58,559
the shredder of course to a lot of many

547
00:21:56,159 --> 00:22:00,960
to a lot of new ideas

548
00:21:58,559 --> 00:22:02,080
but of course one of these ideas was

549
00:22:00,960 --> 00:22:04,080
obviously

550
00:22:02,080 --> 00:22:07,520
we had a three side setup

551
00:22:04,080 --> 00:22:08,799
but still quite small compared to the

552
00:22:07,520 --> 00:22:10,799
big internet

553
00:22:08,799 --> 00:22:13,600
can we make this bigger

554
00:22:10,799 --> 00:22:14,559
and of course yes we can

555
00:22:13,600 --> 00:22:17,520
so

556
00:22:14,559 --> 00:22:19,200
also the similar research about roy in

557
00:22:17,520 --> 00:22:22,799
norway had

558
00:22:19,200 --> 00:22:26,000
ideas around creating a

559
00:22:22,799 --> 00:22:28,000
testbed setup for multi-home systems

560
00:22:26,000 --> 00:22:32,000
so they

561
00:22:28,000 --> 00:22:36,640
published a shop opening i applied

562
00:22:32,000 --> 00:22:39,280
a few weeks later i was moving to norway

563
00:22:36,640 --> 00:22:41,760
to create the nona testbed consisting of

564
00:22:39,280 --> 00:22:42,960
two parts nunet core is the landline

565
00:22:41,760 --> 00:22:45,919
part

566
00:22:42,960 --> 00:22:48,640
this is what i in this talk today will

567
00:22:45,919 --> 00:22:51,360
introduce in a bit more detail

568
00:22:48,640 --> 00:22:53,600
which connects such sort of server

569
00:22:51,360 --> 00:22:56,159
setups as in the figure

570
00:22:53,600 --> 00:22:57,440
at the different sites with up to four

571
00:22:56,159 --> 00:23:00,960
isps

572
00:22:57,440 --> 00:23:03,520
and we had 21 locations with 11 in

573
00:23:00,960 --> 00:23:06,400
norway and 10 aboard

574
00:23:03,520 --> 00:23:09,760
and of course nowadays we all not only

575
00:23:06,400 --> 00:23:10,640
have enzyme connectivity particularly we

576
00:23:09,760 --> 00:23:12,559
have

577
00:23:10,640 --> 00:23:15,039
mobile connectivity

578
00:23:12,559 --> 00:23:16,880
so no net cause it's also of a large

579
00:23:15,039 --> 00:23:19,440
part converted edge

580
00:23:16,880 --> 00:23:20,840
which distributes

581
00:23:19,440 --> 00:23:23,600
these embedded

582
00:23:20,840 --> 00:23:24,799
systems all over norway around 200

583
00:23:23,600 --> 00:23:27,760
locations

584
00:23:24,799 --> 00:23:29,120
to do measurements of the performance of

585
00:23:27,760 --> 00:23:30,720
the norwegian

586
00:23:29,120 --> 00:23:33,360
mobile networks

587
00:23:30,720 --> 00:23:36,080
will not go into too much detail about

588
00:23:33,360 --> 00:23:37,280
known adria because of time constraints

589
00:23:36,080 --> 00:23:39,520
that i will

590
00:23:37,280 --> 00:23:40,960
present you some more details about

591
00:23:39,520 --> 00:23:45,200
monet core

592
00:23:40,960 --> 00:23:49,679
as i've mentioned we have in total

593
00:23:45,200 --> 00:23:50,720
had in 2016 21 sites 11 in norway and 10

594
00:23:49,679 --> 00:23:53,520
aboard

595
00:23:50,720 --> 00:23:56,320
today it's a little bit

596
00:23:53,520 --> 00:23:59,279
fewer sides abroad but still 11 in

597
00:23:56,320 --> 00:24:02,159
norway with up to four isps and wherever

598
00:23:59,279 --> 00:24:05,600
we were able to get ipv6

599
00:24:02,159 --> 00:24:06,480
these isp connections also in the ipv4

600
00:24:05,600 --> 00:24:08,880
have

601
00:24:06,480 --> 00:24:10,559
rbv6 support

602
00:24:08,880 --> 00:24:13,279
and particularly with such an

603
00:24:10,559 --> 00:24:15,039
infrastructure oftentimes on sites you

604
00:24:13,279 --> 00:24:18,559
can get a lot of

605
00:24:15,039 --> 00:24:22,320
details about the internet connectivity

606
00:24:18,559 --> 00:24:25,520
we have developed a tool hyper contraser

607
00:24:22,320 --> 00:24:27,279
to continuously measure the routes by

608
00:24:25,520 --> 00:24:29,360
tracer measurements

609
00:24:27,279 --> 00:24:31,600
inside this infrastructure this is a

610
00:24:29,360 --> 00:24:33,039
view over one month

611
00:24:31,600 --> 00:24:35,360
from

612
00:24:33,039 --> 00:24:37,200
july 2016.

613
00:24:35,360 --> 00:24:39,679
particularly interesting about such

614
00:24:37,200 --> 00:24:41,279
intercontinental setups is of course

615
00:24:39,679 --> 00:24:43,600
that these routes in the internet

616
00:24:41,279 --> 00:24:46,240
frequently change particularly between

617
00:24:43,600 --> 00:24:48,480
for example europe and china so

618
00:24:46,240 --> 00:24:50,880
particularly also depending on the

619
00:24:48,480 --> 00:24:53,360
combinations of isp

620
00:24:50,880 --> 00:24:55,360
locally and remotely you can see that

621
00:24:53,360 --> 00:24:58,480
some of the connections

622
00:24:55,360 --> 00:25:01,600
go westwards over united states while

623
00:24:58,480 --> 00:25:02,559
other connections go eastwards or russia

624
00:25:01,600 --> 00:25:05,360
or

625
00:25:02,559 --> 00:25:06,080
dubai and other locations

626
00:25:05,360 --> 00:25:09,039
to

627
00:25:06,080 --> 00:25:12,320
east area so quite interesting

628
00:25:09,039 --> 00:25:15,279
and particularly of course as mentioned

629
00:25:12,320 --> 00:25:17,919
we have quite

630
00:25:15,279 --> 00:25:19,440
interesting sites and particularly when

631
00:25:17,919 --> 00:25:22,720
it comes to

632
00:25:19,440 --> 00:25:24,880
remote systems for our net setups

633
00:25:22,720 --> 00:25:29,120
well our

634
00:25:24,880 --> 00:25:31,760
servers can be really remote every side

635
00:25:29,120 --> 00:25:34,320
in the city of london on the island of

636
00:25:31,760 --> 00:25:37,600
spitzbergen in norway

637
00:25:34,320 --> 00:25:41,279
is at 78.2 degrees north

638
00:25:37,600 --> 00:25:43,279
so around 1 200 kilometers just from the

639
00:25:41,279 --> 00:25:45,600
north pole probably one of the

640
00:25:43,279 --> 00:25:46,960
northernmost places you can place

641
00:25:45,600 --> 00:25:47,840
servers

642
00:25:46,960 --> 00:25:50,240
so

643
00:25:47,840 --> 00:25:51,600
we deployed the site setup on a nice

644
00:25:50,240 --> 00:25:54,320
spring day

645
00:25:51,600 --> 00:25:55,279
was just minus 30 degree

646
00:25:54,320 --> 00:25:57,760
and

647
00:25:55,279 --> 00:26:00,400
sun was shining so you can imagine how

648
00:25:57,760 --> 00:26:03,679
it looks like in really

649
00:26:00,400 --> 00:26:04,480
wintery conditions with a lot of storm

650
00:26:03,679 --> 00:26:07,120
and

651
00:26:04,480 --> 00:26:09,919
cold temperatures they are also not the

652
00:26:07,120 --> 00:26:13,600
only challenge the other challenges are

653
00:26:09,919 --> 00:26:15,919
wild icebergs homing around this area

654
00:26:13,600 --> 00:26:18,000
so obviously if you think about such a

655
00:26:15,919 --> 00:26:21,760
sight if something goes wrong with the

656
00:26:18,000 --> 00:26:24,799
site and somebody has to go to the site

657
00:26:21,760 --> 00:26:27,360
and this is the way of transport in

658
00:26:24,799 --> 00:26:28,640
longer bean it's a snowmobile so there

659
00:26:27,360 --> 00:26:30,720
are no

660
00:26:28,640 --> 00:26:33,039
regular roads so we have to drive a

661
00:26:30,720 --> 00:26:36,159
snowmobile

662
00:26:33,039 --> 00:26:39,120
so areas where there are wild icebergs

663
00:26:36,159 --> 00:26:40,159
holding around so it's a challenge

664
00:26:39,120 --> 00:26:42,320
so

665
00:26:40,159 --> 00:26:45,440
obviously we have to think about

666
00:26:42,320 --> 00:26:47,520
what can we do when something goes wrong

667
00:26:45,440 --> 00:26:49,760
and things can go wrong because we have

668
00:26:47,520 --> 00:26:52,159
experimental systems and there's no

669
00:26:49,760 --> 00:26:53,279
response anything that can go wrong will

670
00:26:52,159 --> 00:26:56,080
go wrong

671
00:26:53,279 --> 00:26:58,159
so how can we avoid that experimental

672
00:26:56,080 --> 00:27:00,159
software brings down

673
00:26:58,159 --> 00:27:02,320
our infrastructure so that somebody has

674
00:27:00,159 --> 00:27:05,120
to go to the remote side

675
00:27:02,320 --> 00:27:06,880
and do some maintenance work

676
00:27:05,120 --> 00:27:08,960
well the idea was then to use

677
00:27:06,880 --> 00:27:11,840
virtualization

678
00:27:08,960 --> 00:27:15,200
so actually our server setup consists of

679
00:27:11,840 --> 00:27:18,240
a very lightweight setup of ubuntu

680
00:27:15,200 --> 00:27:21,679
server lts lts is the long-term support

681
00:27:18,240 --> 00:27:24,080
version of ubuntu with 5vs support

682
00:27:21,679 --> 00:27:25,760
so therefore is quite stable and quite

683
00:27:24,080 --> 00:27:27,760
well maintained

684
00:27:25,760 --> 00:27:30,399
and everything else

685
00:27:27,760 --> 00:27:31,919
goes into kernel-based virtual machines

686
00:27:30,399 --> 00:27:34,559
kvm

687
00:27:31,919 --> 00:27:37,279
that is particularly the routing

688
00:27:34,559 --> 00:27:40,000
software goes into a

689
00:27:37,279 --> 00:27:42,000
dedicated virtual machine exclusively

690
00:27:40,000 --> 00:27:43,200
running on one of the servers

691
00:27:42,000 --> 00:27:45,039
and all

692
00:27:43,200 --> 00:27:47,440
research

693
00:27:45,039 --> 00:27:49,039
software goes into virtual machines on

694
00:27:47,440 --> 00:27:51,360
the other machines

695
00:27:49,039 --> 00:27:52,880
but it is if something goes wrong

696
00:27:51,360 --> 00:27:55,840
it's possible to log into the host

697
00:27:52,880 --> 00:27:59,200
machine and do some repair work

698
00:27:55,840 --> 00:28:00,320
it is how does such a physical machine

699
00:27:59,200 --> 00:28:03,279
look like

700
00:28:00,320 --> 00:28:06,240
as of that it's you want to serve lds

701
00:28:03,279 --> 00:28:07,919
it's one customization which is

702
00:28:06,240 --> 00:28:11,520
interesting to mention

703
00:28:07,919 --> 00:28:16,480
we decided to use the riser fs version 3

704
00:28:11,520 --> 00:28:20,320
instead of x4 because of its robustness

705
00:28:16,480 --> 00:28:23,440
riser fs actually has a bit bad image

706
00:28:20,320 --> 00:28:26,559
because it's named after the auto

707
00:28:23,440 --> 00:28:28,799
of this file system who killed his x5

708
00:28:26,559 --> 00:28:31,760
but actually his file system is quite

709
00:28:28,799 --> 00:28:34,399
good so it never killed our data and

710
00:28:31,760 --> 00:28:37,919
particularly it never

711
00:28:34,399 --> 00:28:42,080
needed manually fighter step tracks

712
00:28:37,919 --> 00:28:43,760
at boot time which are quite important

713
00:28:42,080 --> 00:28:46,720
this is a quite important property

714
00:28:43,760 --> 00:28:47,840
because if the remote machine wants a

715
00:28:46,720 --> 00:28:49,919
manual

716
00:28:47,840 --> 00:28:52,000
file system check on the console it's

717
00:28:49,919 --> 00:28:53,840
quite annoying we have to the boss of

718
00:28:52,000 --> 00:28:57,440
people at the remote side

719
00:28:53,840 --> 00:29:00,480
which is fast but not with rhiza fs so

720
00:28:57,440 --> 00:29:01,760
therefore we started with arizer fs

721
00:29:00,480 --> 00:29:04,799
nowadays

722
00:29:01,760 --> 00:29:07,200
after unfortunately the rise fs has been

723
00:29:04,799 --> 00:29:09,919
removed from the installer of yuwundo

724
00:29:07,200 --> 00:29:11,039
for new installations we use ptrfs which

725
00:29:09,919 --> 00:29:14,159
became

726
00:29:11,039 --> 00:29:16,159
also much more stable in the last few

727
00:29:14,159 --> 00:29:19,520
years

728
00:29:16,159 --> 00:29:21,520
and as i've said we are using kvm kernel

729
00:29:19,520 --> 00:29:23,600
based virtual machines originally we

730
00:29:21,520 --> 00:29:26,960
used virtualbox

731
00:29:23,600 --> 00:29:29,520
but this was then requiring

732
00:29:26,960 --> 00:29:32,559
the virtualbox is open source itself the

733
00:29:29,520 --> 00:29:35,760
vnc implementation is not so required a

734
00:29:32,559 --> 00:29:37,279
blob form oracle to be installed

735
00:29:35,760 --> 00:29:38,960
and

736
00:29:37,279 --> 00:29:40,480
we were not really satisfied with the

737
00:29:38,960 --> 00:29:43,200
stability of this

738
00:29:40,480 --> 00:29:47,919
so we decided then to move as soon as

739
00:29:43,200 --> 00:29:50,080
possible to kvm which is quite stable

740
00:29:47,919 --> 00:29:52,880
this is the physical machine setup now

741
00:29:50,080 --> 00:29:56,960
coming to the router setup so called

742
00:29:52,880 --> 00:30:00,080
tunnel box come to this name later

743
00:29:56,960 --> 00:30:01,760
well um the tunnel box

744
00:30:00,080 --> 00:30:02,960
connects

745
00:30:01,760 --> 00:30:06,000
the

746
00:30:02,960 --> 00:30:09,840
machine to all the isps of a site

747
00:30:06,000 --> 00:30:12,399
particularly having one ipv4 and if

748
00:30:09,840 --> 00:30:15,840
available one ipv6 address

749
00:30:12,399 --> 00:30:15,840
per site and isp

750
00:30:16,000 --> 00:30:19,360
of course these others are public

751
00:30:17,120 --> 00:30:22,240
addresses but for research experiments

752
00:30:19,360 --> 00:30:24,880
of course for research are nice to have

753
00:30:22,240 --> 00:30:26,880
his own ip addresses

754
00:30:24,880 --> 00:30:29,200
of course if you ask somebody for a

755
00:30:26,880 --> 00:30:31,200
slash 24 public

756
00:30:29,200 --> 00:30:33,760
ipv4 address space

757
00:30:31,200 --> 00:30:37,039
that's of course not possible

758
00:30:33,760 --> 00:30:38,799
those ipv4 address space is exhausted

759
00:30:37,039 --> 00:30:41,360
since several years

760
00:30:38,799 --> 00:30:42,399
so ipv4 addresses are quite scarce

761
00:30:41,360 --> 00:30:43,919
resource

762
00:30:42,399 --> 00:30:46,240
so what we had to do

763
00:30:43,919 --> 00:30:48,720
was then to think about

764
00:30:46,240 --> 00:30:53,840
establishing tunnels between the sites

765
00:30:48,720 --> 00:30:57,039
over gre for ipv4 or ipv6 in rpv6

766
00:30:53,840 --> 00:30:58,720
so that we can apply our own systematic

767
00:30:57,039 --> 00:31:02,799
addressing scheme

768
00:30:58,720 --> 00:31:06,159
with the private rpv4 addresses

769
00:31:02,799 --> 00:31:09,120
and of course then we can give each side

770
00:31:06,159 --> 00:31:11,039
a slash 24 with 10 dot provider id side

771
00:31:09,120 --> 00:31:13,679
id dot node id

772
00:31:11,039 --> 00:31:16,240
and for ipv6 of course we have our own

773
00:31:13,679 --> 00:31:19,279
prefix and can give

774
00:31:16,240 --> 00:31:20,799
for each provider inside uh full slash

775
00:31:19,279 --> 00:31:22,159
64.

776
00:31:20,799 --> 00:31:23,760
it is of course

777
00:31:22,159 --> 00:31:26,480
we can

778
00:31:23,760 --> 00:31:28,960
then allow researchers to use

779
00:31:26,480 --> 00:31:33,440
own addresses for all research

780
00:31:28,960 --> 00:31:35,360
experiments which is quite nice feature

781
00:31:33,440 --> 00:31:37,279
and particularly of course with these

782
00:31:35,360 --> 00:31:39,440
private devices direct communication

783
00:31:37,279 --> 00:31:40,880
between all the sites

784
00:31:39,440 --> 00:31:42,960
is possible

785
00:31:40,880 --> 00:31:46,399
and for

786
00:31:42,960 --> 00:31:49,039
external connectivity we had to decide

787
00:31:46,399 --> 00:31:51,120
well we can of course route

788
00:31:49,039 --> 00:31:54,159
external traffic

789
00:31:51,120 --> 00:31:55,440
at the tunnel box of a remote site

790
00:31:54,159 --> 00:31:58,480
directly

791
00:31:55,440 --> 00:32:01,679
but this causes of course issues for

792
00:31:58,480 --> 00:32:03,360
many hosts because for example we have

793
00:32:01,679 --> 00:32:08,399
sites in germany

794
00:32:03,360 --> 00:32:10,960
and they have very strange regulations

795
00:32:08,399 --> 00:32:12,720
according to copyright and

796
00:32:10,960 --> 00:32:15,360
file download and such

797
00:32:12,720 --> 00:32:18,320
so in fact to convince somebody goes

798
00:32:15,360 --> 00:32:21,039
aside in germany there is no chance

799
00:32:18,320 --> 00:32:23,679
if there is a possibility that people

800
00:32:21,039 --> 00:32:24,880
can do arbitrary downloads from the

801
00:32:23,679 --> 00:32:27,440
internet

802
00:32:24,880 --> 00:32:28,960
so therefore we route all external

803
00:32:27,440 --> 00:32:31,840
traffic

804
00:32:28,960 --> 00:32:34,320
over a similar site so in fact then if

805
00:32:31,840 --> 00:32:36,640
somebody downloads something

806
00:32:34,320 --> 00:32:39,039
from a remote site

807
00:32:36,640 --> 00:32:41,840
even mistakenly for example do a video

808
00:32:39,039 --> 00:32:44,960
download for testing then it goes out

809
00:32:41,840 --> 00:32:47,360
over simula's address in norway and of

810
00:32:44,960 --> 00:32:50,480
course in norway we have much more

811
00:32:47,360 --> 00:32:52,000
relaxed regulations when it comes to

812
00:32:50,480 --> 00:32:54,640
copyright and such

813
00:32:52,000 --> 00:32:56,880
so therefore this was quite important

814
00:32:54,640 --> 00:32:58,240
decision which was necessary

815
00:32:56,880 --> 00:33:00,640
unfortunately

816
00:32:58,240 --> 00:33:04,480
due to such regulations in some

817
00:33:00,640 --> 00:33:05,679
countries particularly germany

818
00:33:04,480 --> 00:33:07,279
well

819
00:33:05,679 --> 00:33:09,600
now we have the router for doing the

820
00:33:07,279 --> 00:33:11,919
routing but of course

821
00:33:09,600 --> 00:33:13,279
although we have quite systematic rp

822
00:33:11,919 --> 00:33:15,519
addressing scheme

823
00:33:13,279 --> 00:33:18,240
it's sometimes difficult to remember all

824
00:33:15,519 --> 00:33:21,919
these rp addresses we have also made our

825
00:33:18,240 --> 00:33:23,120
own dns setup using behind dns server

826
00:33:21,919 --> 00:33:26,000
so we can

827
00:33:23,120 --> 00:33:29,440
name our devices

828
00:33:26,000 --> 00:33:32,320
with the provider and site names so to

829
00:33:29,440 --> 00:33:34,960
have the possibility to do lookup of the

830
00:33:32,320 --> 00:33:37,600
ip addresses and particularly of course

831
00:33:34,960 --> 00:33:39,679
reverse lookups of ip addresses to

832
00:33:37,600 --> 00:33:42,159
symbolic names

833
00:33:39,679 --> 00:33:45,200
and of course since

834
00:33:42,159 --> 00:33:47,200
countries like norway or germany have

835
00:33:45,200 --> 00:33:50,399
extended

836
00:33:47,200 --> 00:33:52,960
a set of characters we have of course

837
00:33:50,399 --> 00:33:56,960
all the support for puni gold works even

838
00:33:52,960 --> 00:33:58,880
for the chinese side with chinese names

839
00:33:56,960 --> 00:34:01,760
so this is one convenience and other

840
00:33:58,880 --> 00:34:04,000
convenience is of course inside the dns

841
00:34:01,760 --> 00:34:05,600
infrastructure we have we can also put

842
00:34:04,000 --> 00:34:08,320
the ssh

843
00:34:05,600 --> 00:34:12,079
keyframe prints as well as

844
00:34:08,320 --> 00:34:14,480
geolocation information as sshfp and loc

845
00:34:12,079 --> 00:34:15,440
resource records

846
00:34:14,480 --> 00:34:17,280
and

847
00:34:15,440 --> 00:34:19,839
since we route

848
00:34:17,280 --> 00:34:23,839
external traffic over simula

849
00:34:19,839 --> 00:34:25,839
makes of course the sense for

850
00:34:23,839 --> 00:34:28,800
the package updates

851
00:34:25,839 --> 00:34:30,240
to be cached by a squid proxy so also on

852
00:34:28,800 --> 00:34:32,879
the tunnel box therefore there is a

853
00:34:30,240 --> 00:34:34,399
screen proxy running to cache

854
00:34:32,879 --> 00:34:37,280
particularly

855
00:34:34,399 --> 00:34:38,800
linux installation packages so that if

856
00:34:37,280 --> 00:34:40,960
multiple

857
00:34:38,800 --> 00:34:42,639
virtual machines at the remote side

858
00:34:40,960 --> 00:34:44,560
have to download the same package that

859
00:34:42,639 --> 00:34:46,839
not all these packages have to be

860
00:34:44,560 --> 00:34:48,560
downloaded via

861
00:34:46,839 --> 00:34:51,679
simulator

862
00:34:48,560 --> 00:34:53,679
this is the router and now let's

863
00:34:51,679 --> 00:34:56,639
introduce the research node

864
00:34:53,679 --> 00:34:58,160
so usually research nodes by default are

865
00:34:56,639 --> 00:35:00,000
managed

866
00:34:58,160 --> 00:35:03,200
as part of

867
00:35:00,000 --> 00:35:05,839
a software called planet central this is

868
00:35:03,200 --> 00:35:08,640
the sense of planet lapios

869
00:35:05,839 --> 00:35:10,240
just with a slight modification

870
00:35:08,640 --> 00:35:14,400
to have

871
00:35:10,240 --> 00:35:15,520
of course an mptcp capable kernel

872
00:35:14,400 --> 00:35:18,880
and

873
00:35:15,520 --> 00:35:20,720
as well as a special configuration for

874
00:35:18,880 --> 00:35:23,119
multi-home routing i will show an

875
00:35:20,720 --> 00:35:26,400
example later this is currently based

876
00:35:23,119 --> 00:35:27,839
still on fedora core 25 which is a bit

877
00:35:26,400 --> 00:35:30,240
out of date now

878
00:35:27,839 --> 00:35:32,160
this soon requires an update so we are

879
00:35:30,240 --> 00:35:35,440
working currently on this

880
00:35:32,160 --> 00:35:35,440
to do some uh

881
00:35:35,599 --> 00:35:39,119
enhancements

882
00:35:36,880 --> 00:35:40,880
and updates

883
00:35:39,119 --> 00:35:43,119
of the software

884
00:35:40,880 --> 00:35:45,280
but what this planet absentee software

885
00:35:43,119 --> 00:35:47,839
does is in fact to provide

886
00:35:45,280 --> 00:35:50,079
containers alexi containers

887
00:35:47,839 --> 00:35:53,119
for research experiments

888
00:35:50,079 --> 00:35:54,160
which have in fact adjust this fedora

889
00:35:53,119 --> 00:35:57,119
core

890
00:35:54,160 --> 00:36:00,000
linux installation but this planet lab

891
00:35:57,119 --> 00:36:01,760
software is actually quite smart

892
00:36:00,000 --> 00:36:04,320
not for each research experiment there

893
00:36:01,760 --> 00:36:06,880
is a requirement to install or complete

894
00:36:04,320 --> 00:36:10,320
fedora core system but in fact it just

895
00:36:06,880 --> 00:36:14,000
works like having a base installation

896
00:36:10,320 --> 00:36:15,359
on our ptifis file system which is then

897
00:36:14,000 --> 00:36:18,640
cloned

898
00:36:15,359 --> 00:36:19,599
so that is just doing copy on white so

899
00:36:18,640 --> 00:36:22,400
it's

900
00:36:19,599 --> 00:36:23,760
quite efficient when it comes to storage

901
00:36:22,400 --> 00:36:26,400
space

902
00:36:23,760 --> 00:36:29,839
and also since it's using containers

903
00:36:26,400 --> 00:36:30,720
also quite efficient for memory so it's

904
00:36:29,839 --> 00:36:34,000
uh

905
00:36:30,720 --> 00:36:36,800
just around 2.5 gigabytes of memory it's

906
00:36:34,000 --> 00:36:39,119
possible to host host multiple research

907
00:36:36,800 --> 00:36:41,839
containers

908
00:36:39,119 --> 00:36:43,280
uh of course besides planet app central

909
00:36:41,839 --> 00:36:45,280
software it's also possible to

910
00:36:43,280 --> 00:36:46,640
instantiate custom virtual machines for

911
00:36:45,280 --> 00:36:49,440
example for

912
00:36:46,640 --> 00:36:50,560
machines running freebastie or more

913
00:36:49,440 --> 00:36:52,720
up-to-date

914
00:36:50,560 --> 00:36:55,440
ubuntu system or systems as owned

915
00:36:52,720 --> 00:36:58,400
kernels if this is necessary for your

916
00:36:55,440 --> 00:37:00,720
search experiment but usually for many

917
00:36:58,400 --> 00:37:04,240
experiments already possible

918
00:37:00,720 --> 00:37:05,280
to run this planet at sample-based

919
00:37:04,240 --> 00:37:08,160
software

920
00:37:05,280 --> 00:37:09,359
what is our physical nodes look like

921
00:37:08,160 --> 00:37:11,359
that

922
00:37:09,359 --> 00:37:13,680
they run

923
00:37:11,359 --> 00:37:13,680
these

924
00:37:13,920 --> 00:37:17,599
simple based

925
00:37:15,920 --> 00:37:20,480
nodes

926
00:37:17,599 --> 00:37:22,720
as kbm virtual machines and inside these

927
00:37:20,480 --> 00:37:23,839
kvm virtual machines are running lxc

928
00:37:22,720 --> 00:37:27,520
containers

929
00:37:23,839 --> 00:37:30,640
instantiating so-called slivers is the

930
00:37:27,520 --> 00:37:32,880
name of a container

931
00:37:30,640 --> 00:37:36,720
in the terminology of the plan lab

932
00:37:32,880 --> 00:37:40,320
centered software so we can run multiple

933
00:37:36,720 --> 00:37:41,680
of these plc nodes as separate virtual

934
00:37:40,320 --> 00:37:45,359
machines

935
00:37:41,680 --> 00:37:46,640
on the same physical host in addition to

936
00:37:45,359 --> 00:37:49,359
possibly

937
00:37:46,640 --> 00:37:51,440
further kvm based virtual machines for

938
00:37:49,359 --> 00:37:55,119
example with special ubuntu

939
00:37:51,440 --> 00:37:57,200
installations or frequency installations

940
00:37:55,119 --> 00:37:58,079
so that is

941
00:37:57,200 --> 00:38:01,440
when

942
00:37:58,079 --> 00:38:04,960
the usual user logs into

943
00:38:01,440 --> 00:38:06,800
one of these slippers

944
00:38:04,960 --> 00:38:09,040
then

945
00:38:06,800 --> 00:38:11,119
what the user sees of course is

946
00:38:09,040 --> 00:38:13,520
that the machine has multiple ip

947
00:38:11,119 --> 00:38:15,200
addresses one for each isp

948
00:38:13,520 --> 00:38:16,880
but this is

949
00:38:15,200 --> 00:38:19,359
a

950
00:38:16,880 --> 00:38:23,280
slipper that is a galaxy container at

951
00:38:19,359 --> 00:38:26,240
hainan university which has two isps

952
00:38:23,280 --> 00:38:28,320
one is centered and one

953
00:38:26,240 --> 00:38:30,560
usb is shiner unicorn

954
00:38:28,320 --> 00:38:35,359
so the machine therefore has twice two

955
00:38:30,560 --> 00:38:35,359
ipv4 addresses and two ipv6 addresses

956
00:38:35,839 --> 00:38:41,839
so that is

957
00:38:38,079 --> 00:38:43,280
in total this machine has four different

958
00:38:41,839 --> 00:38:47,680
ip addresses

959
00:38:43,280 --> 00:38:50,800
so outgoing this four different types

960
00:38:47,680 --> 00:38:52,880
ip addresses can be used then of course

961
00:38:50,800 --> 00:38:56,640
particularly with mpt sapien as i've

962
00:38:52,880 --> 00:38:58,960
said the kernel of the host machine

963
00:38:56,640 --> 00:39:02,079
is running mptcp and since this is a

964
00:38:58,960 --> 00:39:05,440
container of course the user of this

965
00:39:02,079 --> 00:39:07,119
container has the possibility to run

966
00:39:05,440 --> 00:39:08,880
mptcp

967
00:39:07,119 --> 00:39:10,240
as well and particularly with the two

968
00:39:08,880 --> 00:39:13,280
net parameter

969
00:39:10,240 --> 00:39:17,040
it's possible for the researcher to turn

970
00:39:13,280 --> 00:39:20,160
on and off mptcp

971
00:39:17,040 --> 00:39:22,960
or even have concurrent tcp and fbtcp

972
00:39:20,160 --> 00:39:24,480
connections to do some comparisons

973
00:39:22,960 --> 00:39:26,960
for the performance

974
00:39:24,480 --> 00:39:28,800
of mptcp and different settings with

975
00:39:26,960 --> 00:39:32,000
mptcp

976
00:39:28,800 --> 00:39:34,160
and also particularly what is

977
00:39:32,000 --> 00:39:37,680
installed in these containers as well as

978
00:39:34,160 --> 00:39:40,079
a tool system info which displays this

979
00:39:37,680 --> 00:39:44,160
information on login particularly

980
00:39:40,079 --> 00:39:45,200
information about ssh key fingerprints

981
00:39:44,160 --> 00:39:47,359
and

982
00:39:45,200 --> 00:39:49,119
the rp addresses

983
00:39:47,359 --> 00:39:51,599
and of course

984
00:39:49,119 --> 00:39:53,760
these containers have

985
00:39:51,599 --> 00:39:56,480
sudo permissions that is the user of

986
00:39:53,760 --> 00:39:57,839
these containers as possibility to trust

987
00:39:56,480 --> 00:40:00,720
with sudo

988
00:39:57,839 --> 00:40:04,240
install custom software as well for

989
00:40:00,720 --> 00:40:04,240
running research experiments

990
00:40:04,720 --> 00:40:07,839
so

991
00:40:05,680 --> 00:40:09,280
this

992
00:40:07,839 --> 00:40:12,960
concludes

993
00:40:09,280 --> 00:40:14,079
my presentation on new netcore

994
00:40:12,960 --> 00:40:15,280
it is

995
00:40:14,079 --> 00:40:17,599
showing you

996
00:40:15,280 --> 00:40:20,079
the ideas of multi-homing and multi-pass

997
00:40:17,599 --> 00:40:23,119
transfer and of course

998
00:40:20,079 --> 00:40:26,319
introduced to the

999
00:40:23,119 --> 00:40:28,880
new net testbed infrastructure

1000
00:40:26,319 --> 00:40:31,040
and particularly to mention of course no

1001
00:40:28,880 --> 00:40:33,440
net core is an open testbed

1002
00:40:31,040 --> 00:40:35,119
like just like landlab

1003
00:40:33,440 --> 00:40:37,119
so if you are interested in doing

1004
00:40:35,119 --> 00:40:39,760
research in onenet of course it's

1005
00:40:37,119 --> 00:40:41,520
possible to join monet core

1006
00:40:39,760 --> 00:40:44,160
and also use

1007
00:40:41,520 --> 00:40:46,079
this testbed infrastructure

1008
00:40:44,160 --> 00:40:48,240
so quite interesting thing to do

1009
00:40:46,079 --> 00:40:50,880
particularly if you are doing research

1010
00:40:48,240 --> 00:40:53,119
on networks particularly with

1011
00:40:50,880 --> 00:40:55,440
multi-homing but also

1012
00:40:53,119 --> 00:40:58,000
around routing and such

1013
00:40:55,440 --> 00:41:00,400
and currently we are thinking about

1014
00:40:58,000 --> 00:41:01,920
the future of muna particularly we want

1015
00:41:00,400 --> 00:41:05,040
to obtain

1016
00:41:01,920 --> 00:41:08,319
more funding new funding particularly to

1017
00:41:05,040 --> 00:41:11,040
extend the scope of no net core beyond

1018
00:41:08,319 --> 00:41:12,880
this basic idea of multi-pass transport

1019
00:41:11,040 --> 00:41:14,000
but particularly in the areas of network

1020
00:41:12,880 --> 00:41:15,839
function virtualization and

1021
00:41:14,000 --> 00:41:18,400
software-defined networking

1022
00:41:15,839 --> 00:41:21,839
but also around the cloud computing and

1023
00:41:18,400 --> 00:41:24,319
applications particularly this no net

1024
00:41:21,839 --> 00:41:26,240
core infrastructure can also be seen

1025
00:41:24,319 --> 00:41:29,839
as a distributed

1026
00:41:26,240 --> 00:41:31,520
multi-cloud system with several sites

1027
00:41:29,839 --> 00:41:34,079
so it's quite interesting for cloud

1028
00:41:31,520 --> 00:41:36,079
computing applications as well and

1029
00:41:34,079 --> 00:41:38,240
particularly if you are interested in

1030
00:41:36,079 --> 00:41:40,960
more details about on it you can find a

1031
00:41:38,240 --> 00:41:43,839
lot of information about monet

1032
00:41:40,960 --> 00:41:46,839
on the noirnet website

1033
00:41:43,839 --> 00:41:46,839
www.ntp.net

1034
00:41:48,240 --> 00:41:53,280
we're also very committed

1035
00:41:50,079 --> 00:41:56,800
to open source software development

1036
00:41:53,280 --> 00:41:57,680
so a lot of software coming out of no

1037
00:41:56,800 --> 00:41:59,760
net

1038
00:41:57,680 --> 00:42:01,599
is of course available

1039
00:41:59,760 --> 00:42:04,640
in

1040
00:42:01,599 --> 00:42:06,720
different web pages and particularly in

1041
00:42:04,640 --> 00:42:09,599
a lot of github and bitpacket

1042
00:42:06,720 --> 00:42:11,599
repositories so just have a look most of

1043
00:42:09,599 --> 00:42:12,720
the software is licensed

1044
00:42:11,599 --> 00:42:14,319
under

1045
00:42:12,720 --> 00:42:16,800
gpl

1046
00:42:14,319 --> 00:42:19,200
and of course we are research testbed so

1047
00:42:16,800 --> 00:42:20,480
we have of course also a lot of research

1048
00:42:19,200 --> 00:42:23,119
literature

1049
00:42:20,480 --> 00:42:25,680
with the results out of

1050
00:42:23,119 --> 00:42:29,440
experiments in onenet core available you

1051
00:42:25,680 --> 00:42:31,839
can find all these papers also

1052
00:42:29,440 --> 00:42:34,560
on the new netweb server on the

1053
00:42:31,839 --> 00:42:35,760
publication section

1054
00:42:34,560 --> 00:42:39,839
so

1055
00:42:35,760 --> 00:42:41,400
now we are almost over time

1056
00:42:39,839 --> 00:42:45,359
but we still have

1057
00:42:41,400 --> 00:42:47,200
a few moments for questions so thank you

1058
00:42:45,359 --> 00:42:48,160
for attention

1059
00:42:47,200 --> 00:42:50,720
and

1060
00:42:48,160 --> 00:42:54,079
now we have probably around two minutes

1061
00:42:50,720 --> 00:42:54,079
for questions left

1062
00:42:54,560 --> 00:42:58,400
okay thank you thomas

1063
00:42:56,400 --> 00:42:59,599
i'm here now to

1064
00:42:58,400 --> 00:43:03,280
give you the questions there is

1065
00:42:59,599 --> 00:43:06,480
currently one question available

1066
00:43:03,280 --> 00:43:09,760
and that is where is the planet lab

1067
00:43:06,480 --> 00:43:13,520
uh central distro available

1068
00:43:09,760 --> 00:43:16,720
uh it's on our no net web server there

1069
00:43:13,520 --> 00:43:18,319
you can find the links to all the github

1070
00:43:16,720 --> 00:43:20,800
repositories

1071
00:43:18,319 --> 00:43:22,960
so the original upstream software is

1072
00:43:20,800 --> 00:43:24,400
available of course

1073
00:43:22,960 --> 00:43:27,440
from planet lab

1074
00:43:24,400 --> 00:43:30,640
and our changes are in separate

1075
00:43:27,440 --> 00:43:34,240
repositories but they are linked on our

1076
00:43:30,640 --> 00:43:34,240
new net web server

1077
00:43:34,960 --> 00:43:38,720
and the questions or forecasts if you do

1078
00:43:36,800 --> 00:43:44,000
not find it you can just write me an

1079
00:43:38,720 --> 00:43:46,480
email then i can send you the links

1080
00:43:44,000 --> 00:43:48,560
thank you that's all we have in terms of

1081
00:43:46,480 --> 00:43:50,319
questions

1082
00:43:48,560 --> 00:43:53,200
would you like to wrap up that point or

1083
00:43:50,319 --> 00:43:55,040
do you have something else to add

1084
00:43:53,200 --> 00:43:57,040
uh particularly if you later have

1085
00:43:55,040 --> 00:43:59,280
questions just write me an email and

1086
00:43:57,040 --> 00:44:01,599
also will be available

1087
00:43:59,280 --> 00:44:03,520
at the conference in

1088
00:44:01,599 --> 00:44:04,640
chat rooms

1089
00:44:03,520 --> 00:44:07,760
excellent

1090
00:44:04,640 --> 00:44:08,560
okay thank you thomas

1091
00:44:07,760 --> 00:44:10,560
you

1092
00:44:08,560 --> 00:44:14,599
have a nice day

1093
00:44:10,560 --> 00:44:14,599
and enjoy the rest of the conference