WEBVTT

1
00:00.000 --> 00:05.040
The hierarchical network model is a great model. It's one that we absolutely

2
00:05.040 --> 00:05.680
want to use

3
00:05.680 --> 00:11.240
for the foundations for our network designs. But notice that our focus for why

4
00:11.240 --> 00:12.080
the hierarchical

5
00:12.080 --> 00:16.860
network model is so important is because of scale. The larger our network gets,

6
00:16.860 --> 00:18.080
if it's built on

7
00:18.080 --> 00:22.590
the foundational principles of the model, then we're going to have no problem

8
00:22.590 --> 00:23.520
scaling out.

9
00:23.520 --> 00:27.840
But things get a little bit interesting when we scale back in. Specifically,

10
00:27.840 --> 00:28.400
when we're looking

11
00:28.400 --> 00:33.430
at smaller networks, and we realize maybe the hierarchical network model doesn
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

11
00:28.400 --> 00:33.430
at smaller networks, and we realize maybe the hierarchical network model doesn

12
00:33.430 --> 00:34.960
't fit. Or is that

13
00:34.960 --> 00:38.880
even maybe not the best way to say it? Let's take a look at an example here.

14
00:38.880 --> 00:40.800
Let's say that we have

15
00:40.800 --> 00:45.400
a small network. So we've got a small, we'll call it a headquarters site, but

16
00:45.400 --> 00:46.880
maybe it's just a small

17
00:46.880 --> 00:52.800
company. Just getting going here. And we have a couple of access closets. And

18
00:52.800 --> 00:54.240
so we'll call these are

19
00:54.240 --> 00:59.370
access layer switches in here. So we've got a couple of switches in each access

20
00:59.370 --> 01:00.160
closet.
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

20
00:59.370 --> 01:00.160
closet.

21
01:00.160 --> 01:05.840
And then we're going to connect those back, according to the model, to a pair

22
01:05.840 --> 01:06.800
of distribution

23
01:06.800 --> 01:11.760
switches. So here's our distribution model here. And then we know that we've

24
01:11.760 --> 01:13.040
also got an internet

25
01:13.040 --> 01:17.470
circuit coming in. So here's the question, how would the model suggest that I

26
01:17.470 --> 01:18.560
build this out?

27
01:19.200 --> 01:23.520
Well, the model would say, hey, we've got maybe a pair of firewalls up here.

28
01:23.520 --> 01:25.520
And we're going to bring

29
01:25.520 --> 01:30.430
those into a set of core switches. So we bring those into the core switches. So
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

29
01:25.520 --> 01:30.430
those into a set of core switches. So we bring those into the core switches. So

30
01:30.430 --> 01:31.520
this is the core.

31
01:31.520 --> 01:37.510
Now we have distribution down here. And we connect all of these together like

32
01:37.510 --> 01:38.880
we said that we would.

33
01:38.880 --> 01:44.070
And now we have our network built. But here's the big question. What on earth

34
01:44.070 --> 01:45.200
is the value of the

35
01:45.200 --> 01:50.790
core switches here? All we're doing is taking our internet traffic and forward

36
01:50.790 --> 01:51.840
ing it down to the

37
01:51.840 --> 01:56.000
distribution layer, whereas we could have really just brought that internet

38
01:56.000 --> 01:57.440
circuit directly into

39
01:57.440 --> 02:02.810
the distribution layer. So is the core doing anything? Is it worth having in
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

39
01:57.440 --> 02:02.810
the distribution layer. So is the core doing anything? Is it worth having in

40
02:02.810 --> 02:04.320
the answer is no,

41
02:04.320 --> 02:10.110
no, we'd probably be better off without having the core layer. So what we might

42
02:10.110 --> 02:11.680
do is revise our

43
02:11.680 --> 02:16.820
network design here. And we'll say, all right, we still have those access clos

44
02:16.820 --> 02:18.320
ets. We still need those

45
02:18.320 --> 02:23.670
access switches. And then we'll put a pair of probably pretty small

46
02:23.670 --> 02:26.320
distribution switches here

47
02:26.320 --> 02:31.640
that are going to connect down. So this is still our dist layer. And then our
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

47
02:26.320 --> 02:31.640
that are going to connect down. So this is still our dist layer. And then our

48
02:31.640 --> 02:33.600
internet circuit is

49
02:33.600 --> 02:40.190
coming into a pair of firewalls. Draw these a little bit bigger now. And so we

50
02:40.190 --> 02:41.440
've got an internet

51
02:41.440 --> 02:45.640
circuit coming in probably redundantly in some way into these firewalls. Maybe

52
02:45.640 --> 02:46.160
it's a high

53
02:46.160 --> 02:49.840
availability pair. But either way, we now can connect these into the

54
02:49.840 --> 02:51.600
distribution layer. And now this

55
02:51.600 --> 02:57.370
is a much cleaner network model. Or maybe I should say it's it's right sized.

56
02:57.370 --> 02:58.800
Again, doing network

57
02:58.800 --> 03:05.270
designs, there's always this sense of trying to find the balance between what
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

57
02:58.800 --> 03:05.270
designs, there's always this sense of trying to find the balance between what

58
03:05.270 --> 03:06.880
the right type of

59
03:06.880 --> 03:11.150
design is for a particular organization. And also maybe what something like the

60
03:11.150 --> 03:12.320
hierarchical model

61
03:12.320 --> 03:16.580
is trying to tell us because the hierarchical model says, Hey, we need a a core

62
03:16.580 --> 03:17.760
layer or a core block.

63
03:17.760 --> 03:23.790
And I'm looking at this saying, Hey, I'm maybe consulting or maybe this is my

64
03:23.790 --> 03:25.360
organization. And I'm

65
03:25.360 --> 03:30.340
trying to go and justify switches that cost multiple thousands of dollars to
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

65
03:25.360 --> 03:30.340
trying to go and justify switches that cost multiple thousands of dollars to

66
03:30.340 --> 03:31.520
just sit there as a

67
03:31.520 --> 03:36.300
core layer. It doesn't make sense. So we have to adapt our model and we have to

68
03:36.300 --> 03:38.240
figure out a couple

69
03:38.240 --> 03:41.830
of things. First of all, what is the right size for this network? We want to

70
03:41.830 --> 03:43.280
make sure that we're not

71
03:43.280 --> 03:47.300
you building something that is way too big and way too expensive that this

72
03:47.300 --> 03:48.480
company doesn't need.

73
03:48.480 --> 03:53.280
But at the same time, we need to ask ourselves, Okay, but what about that whole

74
03:53.280 --> 03:54.720
scaling concept? Because

75
03:54.720 --> 04:00.530
I our company is small, but we might well grow. So am I setting myself up for
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

75
03:54.720 --> 04:00.530
I our company is small, but we might well grow. So am I setting myself up for

76
04:00.530 --> 04:03.200
failure by in some way

77
04:03.200 --> 04:08.000
violating the model? And what we're going to find is that we're not actually in

78
04:08.000 --> 04:08.720
violation of

79
04:08.720 --> 04:13.270
anything here. We are adhering to the model, even though it doesn't look like

80
04:13.270 --> 04:14.400
it quite at first.

81
04:14.400 --> 04:18.480
So what did we do here? Why can we say that we're adhering to the model? Well,

82
04:18.480 --> 04:19.360
first of all, we see

83
04:19.360 --> 04:23.920
that there are a dedicated distribution block. We have our access block down

84
04:23.920 --> 04:25.600
here. We have our

85
04:25.600 --> 04:30.940
internet block up here at the top, but we don't have a core layer or do we? See
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

85
04:25.600 --> 04:30.940
internet block up here at the top, but we don't have a core layer or do we? See

86
04:30.940 --> 04:33.440
, when we build out the

87
04:33.440 --> 04:41.140
hierarchical network model, the model appears to be a physical model. We use

88
04:41.140 --> 04:42.400
that building block

89
04:42.400 --> 04:45.640
concept. I put the core in the middle and I connect it out to all of these

90
04:45.640 --> 04:47.520
different blocks. And we

91
04:48.160 --> 04:52.160
want to make the assumption that each one of these blocks is a physical set of

92
04:52.160 --> 04:53.440
switches. So I've got

93
04:53.440 --> 04:58.800
switches in here and pair of switches and all of this is a physical model. But

94
04:58.800 --> 05:00.400
this isn't necessarily
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

94
04:58.800 --> 05:00.400
this isn't necessarily

95
05:00.400 --> 05:07.680
true. Instead, the model isn't really focused as much on the physical, but on

96
05:07.680 --> 05:10.640
the logical. And specifically,

97
05:12.160 --> 05:18.800
the word I like to use for this is function or functions. We'd spent a lot of

98
05:18.800 --> 05:19.840
time breaking down the

99
05:19.840 --> 05:24.110
functions of the different layers, especially as it relates to the access layer

100
05:24.110 --> 05:25.120
, the distribution

101
05:25.120 --> 05:30.780
layer, the core layer, wherever that is up here. Core distribution and access
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

101
05:25.120 --> 05:30.780
layer, the core layer, wherever that is up here. Core distribution and access

102
05:30.780 --> 05:32.160
all have very

103
05:32.160 --> 05:38.010
different functions. But notice that in our case here, we don't have a physical

104
05:38.010 --> 05:39.920
core layer. And so

105
05:39.920 --> 05:45.050
is that a problem? Well, not really, because we actually do have core

106
05:45.050 --> 05:46.480
functionality here.

107
05:46.480 --> 05:50.640
What's the core supposed to do? Fast layer three switching. What are these

108
05:50.640 --> 05:52.080
distribution switches do?

109
05:52.080 --> 05:56.120
They're performing fast layer three switching between the internet block up

110
05:56.120 --> 05:58.080
here and the access

111
05:58.080 --> 06:06.460
blocks down below. So we do have core and distribution functions happening,
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

111
05:58.080 --> 06:06.460
blocks down below. So we do have core and distribution functions happening,

112
06:06.460 --> 06:08.480
even though it's a,

113
06:08.480 --> 06:13.900
they're sharing a single pair of switches. So again, not focused on the

114
06:13.900 --> 06:16.000
physical, we are focused on

115
06:16.000 --> 06:21.340
the functions, the logical side of the model. And in that sense, we absolutely

116
06:21.340 --> 06:22.480
are adhering to

117
06:22.480 --> 06:27.860
the hierarchical network model, even for this smaller network design. Now,

118
06:27.860 --> 06:29.280
certainly in a perfect

119
06:29.280 --> 06:33.780
world, each of these logical functions would map to a physical set of switches
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

119
06:29.280 --> 06:33.780
world, each of these logical functions would map to a physical set of switches

120
06:33.780 --> 06:34.960
that are dedicated

121
06:34.960 --> 06:39.150
for that purpose. And for large at scale networks, we absolutely should be

122
06:39.150 --> 06:41.840
doing that. But this concept

123
06:41.840 --> 06:47.270
of placing multiple functions into the same physical set of hardware has a word

124
06:47.270 --> 06:48.080
that we use often

125
06:48.080 --> 06:54.060
in this industry and it's known as collapsing. So in this case, we would

126
06:54.060 --> 06:56.320
especially refer to this as

127
06:56.320 --> 07:03.680
a collapsed core design, where we don't have a dedicated core. Instead, we have
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

127
06:56.320 --> 07:03.680
a collapsed core design, where we don't have a dedicated core. Instead, we have

128
07:03.680 --> 07:04.480
collapsed

129
07:04.960 --> 07:09.980
this word collapsing here. We have collapsed the core functions into the

130
07:09.980 --> 07:11.920
distribution switches.

131
07:11.920 --> 07:15.840
Or I guess you could argue that we collapse the distribution functions up into

132
07:15.840 --> 07:16.640
the core, but

133
07:16.640 --> 07:20.880
usually we go sort of a top down approach when we're talking about this.

134
07:20.880 --> 07:27.050
The way I like to refer to this is not just saying collapsed core, but we might

135
07:27.050 --> 07:28.800
say it's a collapsed

136
07:28.800 --> 07:38.400
core and distribution layer. So however we want to word it is fine, but when we
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

136
07:28.800 --> 07:38.400
core and distribution layer. So however we want to word it is fine, but when we

137
07:38.400 --> 07:39.440
are referring to this

138
07:39.440 --> 07:43.570
no matter what, we are going to refer to it as being collapsed. And so we can

139
07:43.570 --> 07:45.120
take one physical set of

140
07:45.120 --> 07:50.060
devices and give it multiple functions using this collapsing method. Now,

141
07:50.060 --> 07:51.520
collapsing doesn't have to

142
07:51.520 --> 07:57.280
stop there. We don't necessarily just stop with the idea of collapsing our core

143
07:57.280 --> 07:58.080
and distribution

144
07:58.080 --> 08:07.060
blocks together. Let's imagine a site that has 10 employees. What kind of
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

144
07:58.080 --> 08:07.060
blocks together. Let's imagine a site that has 10 employees. What kind of

145
08:07.060 --> 08:09.600
network does this site need?

146
08:09.600 --> 08:15.290
Well, if I truly try to adhere to the model, we're going to have way more

147
08:15.290 --> 08:16.800
hardware than we ever

148
08:16.800 --> 08:21.400
needed if we're mapping each function to physical sets of hardware. But even if

149
08:21.400 --> 08:22.240
we build out a

150
08:22.240 --> 08:27.210
network like this over here, this is overkill because we only have 10 employees

151
08:27.210 --> 08:28.640
. And by assumption here,

152
08:28.640 --> 08:33.000
we're saying that we have roughly 10 devices. Maybe we have some wireless
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

152
08:28.640 --> 08:33.000
we're saying that we have roughly 10 devices. Maybe we have some wireless

153
08:33.000 --> 08:33.920
devices in there.

154
08:33.920 --> 08:40.460
But ultimately, we don't need a ton of connectivity. So how would we build out

155
08:40.460 --> 08:41.200
this network?

156
08:41.200 --> 08:46.160
Well, probably we have some kind of internet circuit coming in to a firewall.

157
08:46.160 --> 08:46.720
It's probably not

158
08:46.720 --> 08:53.360
even redundant for this little site. And then we could have a single switch and

159
08:53.360 --> 08:55.120
all of our users and

160
08:55.120 --> 09:01.450
all of our phones and all of our access points could all connect into a 24 or
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

160
08:55.120 --> 09:01.450
all of our phones and all of our access points could all connect into a 24 or

161
09:01.450 --> 09:03.680
maybe 48 port switch

162
09:03.680 --> 09:07.700
if we have enough extra devices in there. And then we just connect the firewall

163
09:07.700 --> 09:08.080
up.

164
09:08.080 --> 09:14.080
So does this add here to the model? And technically the answer once again is

165
09:14.080 --> 09:16.240
yes. This does add here

166
09:16.240 --> 09:20.620
to this model, but not from a physical perspective, yet instead from a

167
09:20.620 --> 09:22.240
functional perspective,

168
09:22.240 --> 09:26.170
we would say that it does. Because at this point, what we've done is we've

169
09:26.170 --> 09:27.360
created a collapsed

170
09:27.360 --> 09:34.710
environment. And we have created a single switch here that is a collapsed core,
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

170
09:27.360 --> 09:34.710
environment. And we have created a single switch here that is a collapsed core,

171
09:34.710 --> 09:35.680
distribution,

172
09:35.680 --> 09:41.530
and access switch. It's performing all of the functions. We have connectivity

173
09:41.530 --> 09:42.160
from an access

174
09:42.160 --> 09:46.160
perspective here. We're creating the connectivity. We're assigning VLANs,

175
09:46.160 --> 09:48.160
potentially doing QOS tags.

176
09:48.160 --> 09:53.780
We have SBI's on this switch. And we're not running first-top redundancy

177
09:53.780 --> 09:55.120
protocols because

178
09:55.120 --> 10:01.720
we don't even have a second switch. But we do have the SBI's. We are creating
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

178
09:55.120 --> 10:01.720
we don't even have a second switch. But we do have the SBI's. We are creating

179
10:01.720 --> 10:02.640
the functionality

180
10:02.640 --> 10:06.410
that the distribution performs. And then the core functionality is simply

181
10:06.410 --> 10:07.760
switching back and forth

182
10:07.760 --> 10:14.000
between the firewall and the users. And so yeah, it's a single switch. And yet

183
10:14.000 --> 10:15.520
it adheres to the

184
10:15.520 --> 10:22.480
higher, higher article network model. And I suppose we might even call this a

185
10:22.480 --> 10:23.920
one-tier model.

186
10:23.920 --> 10:30.290
Whereas before over here with our collapsed core and distribution plus our
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

186
10:23.920 --> 10:30.290
Whereas before over here with our collapsed core and distribution plus our

187
10:30.290 --> 10:30.640
access,

188
10:30.640 --> 10:38.510
this would be more of a traditional two-tier architecture. Now ultimately, even

189
10:38.510 --> 10:39.280
though probably

190
10:39.280 --> 10:42.170
stretches a little beyond the scope of this course, if we're thinking about

191
10:42.170 --> 10:43.040
being network

192
10:43.040 --> 10:49.450
architects or studying network architecture, a big part of designing networks

193
10:49.450 --> 10:50.880
is staying flexible.

194
10:50.880 --> 10:57.750
We want to stay flexible because we need to find a balance between the spend of

195
10:57.750 --> 10:58.560
a network. In other

196
10:58.560 --> 11:05.600
words, how much does this network cost? And compare that to really just having
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

196
10:58.560 --> 11:05.600
words, how much does this network cost? And compare that to really just having

197
11:05.600 --> 11:07.680
structure. So adhering

198
11:07.680 --> 11:13.880
to this model. There was a time when my brother actually gave me a call. He was

199
11:13.880 --> 11:14.640
looking at starting

200
11:14.640 --> 11:18.540
up a company and he asked me for a quote and he decided, "Hey, not necessarily

201
11:18.540 --> 11:19.680
a quote." But he asked

202
11:19.680 --> 11:23.760
me for a budget. Like, how much should I expect to spend on a network for this

203
11:23.760 --> 11:25.520
site? And because it's

204
11:25.520 --> 11:29.770
my brother and I just like messing with him, like he likes messing with me, I

205
11:29.770 --> 11:30.640
sent him a quote for
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

205
11:29.770 --> 11:30.640
sent him a quote for

206
11:30.640 --> 11:37.680
$150,000 worth of equipment. And after laughing a little bit, we had to start

207
11:37.680 --> 11:39.360
having more real conversations.

208
11:39.360 --> 11:46.090
But a small startup company probably just needed a small network switch in a

209
11:46.090 --> 11:47.040
firewall,

210
11:47.040 --> 11:52.800
not $150,000 worth of gear. And so it is always a matter again from a design

211
11:52.800 --> 11:54.080
perspective of saying,

212
11:54.080 --> 11:58.730
"I know what I'm trying to do. I'm trying to do this model. I'm trying to make

213
11:58.730 --> 12:00.320
it so we can scale up."
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

213
11:58.730 --> 12:00.320
it so we can scale up."

214
12:00.320 --> 12:07.060
But as long as we are adhering to the principles of the model and making it so

215
12:07.060 --> 12:08.080
that we can scale up

216
12:08.080 --> 12:13.240
one day, then we're going to be in good shape as long as we're considering all

217
12:13.240 --> 12:14.960
of these principles.

218
12:14.960 --> 12:19.600
So that's certainly one of the takeaways from this conversation. But when we're

219
12:19.600 --> 12:20.880
saying flexible,

220
12:20.880 --> 12:24.690
we need to think about all of the things. We need to think about the fact that

221
12:24.690 --> 12:25.520
we might deploy this

222
12:25.520 --> 12:29.200
one-tier model. We might deploy a two-tier model where core and distribution

223
12:29.200 --> 12:30.160
are collapsed.
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

223
12:29.200 --> 12:30.160
are collapsed.

224
12:30.160 --> 12:34.620
We might actually have a situation where we collapse our distribution and

225
12:34.620 --> 12:35.760
access together. So

226
12:35.760 --> 12:41.600
what happens if I have a distribution switch and I have access switches out in

227
12:41.600 --> 12:42.640
access closets?

228
12:42.640 --> 12:48.520
But inside of this distribution closet, I also have runs that are going out to

229
12:48.520 --> 12:49.520
the floor where

230
12:49.520 --> 12:54.300
users are going to connect. Does it make sense for me to deploy dedicated

231
12:54.300 --> 12:56.160
access switches for those

232
12:56.160 --> 12:59.860
users? Or can they just come into the distribution switch? At that point, this

233
12:59.860 --> 13:00.720
would be a collapsed
WEBVTT
X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:900000

233
12:59.860 --> 13:00.720
would be a collapsed

234
13:00.720 --> 13:06.130
distribution and access block. And so these are the kinds of ways that we can

235
13:06.130 --> 13:08.400
leverage what we

236
13:08.400 --> 13:12.410
simply call a two-tier architecture. But ultimately, it comes back to saying we

237
13:12.410 --> 13:13.680
're going to leverage

238
13:13.680 --> 13:18.500
the collapsing of these different functions together, not necessarily lock

239
13:18.500 --> 13:19.360
every single

240
13:19.360 --> 13:24.960
function to a physical set of hardware. And again, try to strike that balance

241
13:24.960 --> 13:25.840
in order to make

242
13:25.840 --> 13:29.200
sure that we're not over-sizing networks for the need.