1
00:00:00,000 --> 00:00:03,395
[AUDIO LOGO]

2
00:00:03,395 --> 00:00:07,760

3
00:00:07,760 --> 00:00:12,620
So let's begin by talking
about why we want MPLS traffic

4
00:00:12,620 --> 00:00:13,430
engineering.

5
00:00:13,430 --> 00:00:15,230
What does this
bring to the table

6
00:00:15,230 --> 00:00:17,570
that we don't normally have?

7
00:00:17,570 --> 00:00:21,270
Well, we've discussed it a
little bit in previous skills,

8
00:00:21,270 --> 00:00:25,070
but keep in mind
that MPLS normally

9
00:00:25,070 --> 00:00:28,430
follows the IGP when
determining where

10
00:00:28,430 --> 00:00:32,090
to forward the packets to,
in other words, the next hop.

11
00:00:32,090 --> 00:00:37,010
Well, IGPs, such
as OSPF and ISIS,

12
00:00:37,010 --> 00:00:39,440
although technically you
could use other protocols,

13
00:00:39,440 --> 00:00:41,340
they don't support
traffic engineering.

14
00:00:41,340 --> 00:00:43,290
So we won't be discussing those.

15
00:00:43,290 --> 00:00:48,350
But OSPF and ISIS, of course,
have their own path selection

16
00:00:48,350 --> 00:00:49,580
criteria.

17
00:00:49,580 --> 00:00:52,290
OSPF is a little bit
more complicated,

18
00:00:52,290 --> 00:00:55,190
but just in a general
sense, they both sort of

19
00:00:55,190 --> 00:00:57,050
come down to lowest cost.

20
00:00:57,050 --> 00:00:59,550
Of course, it's a little
more complicated than that.

21
00:00:59,550 --> 00:01:01,670
But generally speaking,
that's good enough

22
00:01:01,670 --> 00:01:03,230
for this discussion.

23
00:01:03,230 --> 00:01:07,640
Whatever their criteria,
they choose a best path

24
00:01:07,640 --> 00:01:09,990
to get to a destination.

25
00:01:09,990 --> 00:01:14,570
Now, if there are multiple
paths that are completely equal,

26
00:01:14,570 --> 00:01:18,530
we of course support
ECMP or equal

27
00:01:18,530 --> 00:01:23,330
cost multipathing, in which
case, even MPLS can still

28
00:01:23,330 --> 00:01:28,280
use both of those paths and
send traffic along both paths.

29
00:01:28,280 --> 00:01:31,520
That's not what traffic
engineering is for.

30
00:01:31,520 --> 00:01:34,340
What traffic
engineering is for is

31
00:01:34,340 --> 00:01:38,600
what happens when we
don't have equal paths.

32
00:01:38,600 --> 00:01:40,110
Just as an example.

33
00:01:40,110 --> 00:01:44,900
What if I had a 10-gig
circuit and a 1-gig circuit

34
00:01:44,900 --> 00:01:46,770
that I wanted to use?

35
00:01:46,770 --> 00:01:50,030
Now, obviously,
OSPF and ISIS should

36
00:01:50,030 --> 00:01:54,350
choose the 10-gig circuit to be
our best link, our best path.

37
00:01:54,350 --> 00:01:56,300
And it would be.

38
00:01:56,300 --> 00:02:01,160
But if that 10-gig
circuit is running at 90%,

39
00:02:01,160 --> 00:02:04,250
well, it only has
one gig left, anyway.

40
00:02:04,250 --> 00:02:08,060
So now, you have two
equal paths again.

41
00:02:08,060 --> 00:02:13,310
But your IGP, normally is not
going to keep track of that.

42
00:02:13,310 --> 00:02:16,940
Where traffic engineering comes
into play is where it can say,

43
00:02:16,940 --> 00:02:20,490
yes, the IGP says
this is the best path.

44
00:02:20,490 --> 00:02:23,240
However, in the
right circumstances,

45
00:02:23,240 --> 00:02:27,080
maybe this other path, which
is perfectly fine, perfectly

46
00:02:27,080 --> 00:02:29,810
viable, there's no
loops, it's just

47
00:02:29,810 --> 00:02:33,650
not the best, that doesn't
make it not usable.

48
00:02:33,650 --> 00:02:36,170
Traffic engineering
actually allows

49
00:02:36,170 --> 00:02:39,330
us to use those other paths.

50
00:02:39,330 --> 00:02:42,140
So let's jump in and talk
a little bit more detail

51
00:02:42,140 --> 00:02:45,140
about why we want this and
how it fits into everything.

52
00:02:45,140 --> 00:02:48,030
And well, some of the things
we need to do to support it.

53
00:02:48,030 --> 00:02:50,780
And then in our future
videos here, we're

54
00:02:50,780 --> 00:02:53,640
going to be taking a look at
how it actually functions.

55
00:02:53,640 --> 00:02:57,050
So let's get in and look at some
of the things that make traffic

56
00:02:57,050 --> 00:03:00,370
engineering worth running.

57
00:03:00,370 --> 00:03:03,600
So the main reason that we
want traffic engineering

58
00:03:03,600 --> 00:03:08,190
is to get better usage of
what would be potentially

59
00:03:08,190 --> 00:03:10,330
underutilized resources.

60
00:03:10,330 --> 00:03:12,780
So like we said during
the introduction, maybe

61
00:03:12,780 --> 00:03:16,560
that 1-gig link that's sitting
there not being used at all

62
00:03:16,560 --> 00:03:20,980
because the IGP simply says
it's not the best path.

63
00:03:20,980 --> 00:03:23,700
Now, to ensure that we
can use these other paths

64
00:03:23,700 --> 00:03:26,280
and that they are,
in fact, loop-free,

65
00:03:26,280 --> 00:03:29,040
we need a routing
protocol that actually

66
00:03:29,040 --> 00:03:33,120
has a database with a
view of the whole network

67
00:03:33,120 --> 00:03:34,680
or the topology.

68
00:03:34,680 --> 00:03:37,440
That, of course, is
going to require a link

69
00:03:37,440 --> 00:03:39,210
state routing protocol.

70
00:03:39,210 --> 00:03:41,550
It also requires extensions.

71
00:03:41,550 --> 00:03:45,090
We call them, of course,
the TE extensions.

72
00:03:45,090 --> 00:03:47,070
But these allow it
to also help keep

73
00:03:47,070 --> 00:03:51,000
track of the load on
that link, not just

74
00:03:51,000 --> 00:03:53,040
the available
bandwidth, which is what

75
00:03:53,040 --> 00:03:55,410
these protocols normally track.

76
00:03:55,410 --> 00:03:59,200
Now, given that it has to be
link state, as we mentioned,

77
00:03:59,200 --> 00:04:01,750
there are only two
choices for this.

78
00:04:01,750 --> 00:04:06,120
Of course, we can
either run OSPF or ISIS,

79
00:04:06,120 --> 00:04:09,060
which is why these are, of
course, the two main protocols

80
00:04:09,060 --> 00:04:13,050
that we use under MPLS.

81
00:04:13,050 --> 00:04:17,070
Keep in mind, even if we're not
using MPLS traffic engineering

82
00:04:17,070 --> 00:04:20,680
right now, we may at
some point in the future.

83
00:04:20,680 --> 00:04:22,320
So generally speaking,
if you're going

84
00:04:22,320 --> 00:04:25,440
to build an MPLS environment,
it makes the most sense

85
00:04:25,440 --> 00:04:28,920
to build it on
either OSPF or ISIS.

86
00:04:28,920 --> 00:04:32,340
And of course, most service
providers have done just that.

87
00:04:32,340 --> 00:04:36,630
This is more of a discussion
for enterprise core.

88
00:04:36,630 --> 00:04:38,430
Because again, a
lot of enterprises

89
00:04:38,430 --> 00:04:41,490
also use MPLS in their
core and they may not

90
00:04:41,490 --> 00:04:43,620
be using traffic
engineering right now,

91
00:04:43,620 --> 00:04:45,960
but they may want
to in the future.

92
00:04:45,960 --> 00:04:48,730
Again, from a service
provider perspective,

93
00:04:48,730 --> 00:04:51,600
we are probably already running
one of these two protocols,

94
00:04:51,600 --> 00:04:54,270
and while your environment
is probably already running

95
00:04:54,270 --> 00:04:56,400
traffic engineering as well.

96
00:04:56,400 --> 00:04:59,610
To put this another
way, what we're doing

97
00:04:59,610 --> 00:05:03,750
is we're actually mapping
the traffic to the network.

98
00:05:03,750 --> 00:05:06,420
We're not just routing
down the best path.

99
00:05:06,420 --> 00:05:09,090
Again, that's what
your IGPs normally do,

100
00:05:09,090 --> 00:05:13,020
that's just standard
IP-based forwarding.

101
00:05:13,020 --> 00:05:18,190
TE allows us to use other paths
rather than just the best.

102
00:05:18,190 --> 00:05:21,130
And we are going to be
mapping that traffic.

103
00:05:21,130 --> 00:05:24,570
So as we said, what the IGP,
of course, will normally do,

104
00:05:24,570 --> 00:05:28,780
is it routes traffic
based on the best path.

105
00:05:28,780 --> 00:05:32,430
Whatever that means to that
particular routing protocol,

106
00:05:32,430 --> 00:05:35,010
that's, of course, going
to be based entirely

107
00:05:35,010 --> 00:05:38,850
on the destination address
where the traffic is going.

108
00:05:38,850 --> 00:05:40,830
And of course, the
most important thing

109
00:05:40,830 --> 00:05:42,450
in this whole
discussion really is

110
00:05:42,450 --> 00:05:46,890
that it does not change
based on the traffic load.

111
00:05:46,890 --> 00:05:49,680
These routing protocols
don't track load,

112
00:05:49,680 --> 00:05:52,320
they just look at the
speed of the circuit.

113
00:05:52,320 --> 00:05:54,210
Overall, the cost.

114
00:05:54,210 --> 00:05:56,310
This, of course, is
where traffic engineering

115
00:05:56,310 --> 00:05:59,550
steps in and allows
that traffic to be

116
00:05:59,550 --> 00:06:03,420
mapped based on other criteria.

117
00:06:03,420 --> 00:06:07,650
Not just the criteria for the
IGP, which, again, is usually

118
00:06:07,650 --> 00:06:09,690
going to be lowest cost.

119
00:06:09,690 --> 00:06:13,380
We might want to do this
based on traffic type.

120
00:06:13,380 --> 00:06:15,120
Maybe I want to
send voice traffic

121
00:06:15,120 --> 00:06:17,130
down a particular circuit.

122
00:06:17,130 --> 00:06:20,100
Just as another
example, I did some work

123
00:06:20,100 --> 00:06:22,650
with a company that
did stock trading.

124
00:06:22,650 --> 00:06:24,690
And they had two
remote offices that

125
00:06:24,690 --> 00:06:26,760
were rather far
apart, they needed

126
00:06:26,760 --> 00:06:30,390
to get those stock, quotes, "to
the other location as quickly

127
00:06:30,390 --> 00:06:31,740
as possible."

128
00:06:31,740 --> 00:06:35,430
In doing so, they had put
in this very elaborate, very

129
00:06:35,430 --> 00:06:40,470
direct line of sight, low
speed but incredibly low

130
00:06:40,470 --> 00:06:43,830
latency circuit to get
those stock quotes through.

131
00:06:43,830 --> 00:06:46,350
It was like 3
milliseconds faster

132
00:06:46,350 --> 00:06:47,710
than their normal circuit.

133
00:06:47,710 --> 00:06:49,150
Doesn't sound like much.

134
00:06:49,150 --> 00:06:51,780
But apparently, in the stock
trading market, that's a lot.

135
00:06:51,780 --> 00:06:54,870
Because they spent a lot
of money for this circuit

136
00:06:54,870 --> 00:06:57,240
that was only 3
milliseconds, less

137
00:06:57,240 --> 00:06:59,490
latency than their
other circuit.

138
00:06:59,490 --> 00:07:03,990
But the point is, that circuit
was also low bandwidth and very

139
00:07:03,990 --> 00:07:05,290
high cost.

140
00:07:05,290 --> 00:07:08,430
So we use traffic
engineering to ensure

141
00:07:08,430 --> 00:07:12,030
that the only thing that went
over that special circuit was,

142
00:07:12,030 --> 00:07:16,020
in fact, the stock trading
information, all other traffic

143
00:07:16,020 --> 00:07:18,570
was sent across
the normal circuit.

144
00:07:18,570 --> 00:07:21,690
Now, in truth, that
particular environment,

145
00:07:21,690 --> 00:07:24,420
we were not using MPLS
traffic engineering,

146
00:07:24,420 --> 00:07:26,280
we were doing it a
slightly different way.

147
00:07:26,280 --> 00:07:28,230
But it's still the same concept.

148
00:07:28,230 --> 00:07:31,140
We may want to do this to
send a certain type of traffic

149
00:07:31,140 --> 00:07:34,560
down a certain circuit
for a particular reason,

150
00:07:34,560 --> 00:07:36,490
whatever that may be.

151
00:07:36,490 --> 00:07:38,220
Of course, another
very popular reason,

152
00:07:38,220 --> 00:07:42,360
as we've mentioned numerous
times now, is the traffic load.

153
00:07:42,360 --> 00:07:45,630
Again, this other alternate
circuit may not be the fastest,

154
00:07:45,630 --> 00:07:49,930
it may not be the best, but
it's still perfectly fine.

155
00:07:49,930 --> 00:07:52,950
I always like to use
the analogy of a GPS.

156
00:07:52,950 --> 00:07:55,140
If you're trying to
find your way somewhere,

157
00:07:55,140 --> 00:07:58,870
the GPS will most likely
route you on to the highway.

158
00:07:58,870 --> 00:08:01,470
However, if there's a lot of
traffic on the highway or maybe

159
00:08:01,470 --> 00:08:04,450
an accident, it'll
find an alternate route

160
00:08:04,450 --> 00:08:08,670
and send you around the accident
on what's potentially maybe

161
00:08:08,670 --> 00:08:11,100
side streets or secondary roads.

162
00:08:11,100 --> 00:08:13,583
Normally, those wouldn't
be the best way to go.

163
00:08:13,583 --> 00:08:15,000
But they're certainly
the best way

164
00:08:15,000 --> 00:08:20,230
to go when the best way
to go is all backed up.

165
00:08:20,230 --> 00:08:23,340
So again, we're talking
about using those links that

166
00:08:23,340 --> 00:08:26,440
normally would not be utilized.

167
00:08:26,440 --> 00:08:28,360
So again, maybe
those side streets.

168
00:08:28,360 --> 00:08:30,360
You probably wouldn't
normally take the back way

169
00:08:30,360 --> 00:08:33,000
to get somewhere, but if
the highway is backed up,

170
00:08:33,000 --> 00:08:34,679
it's a great choice.

171
00:08:34,679 --> 00:08:37,120
And ultimately,
what does this do?

172
00:08:37,120 --> 00:08:39,539
It certainly helps the
end user experience,

173
00:08:39,539 --> 00:08:42,390
because we're getting use of
all of our available bandwidth,

174
00:08:42,390 --> 00:08:45,550
maybe a particular circuit
type, that sort of thing.

175
00:08:45,550 --> 00:08:49,800
And ultimately, it's going
to save us money because why?

176
00:08:49,800 --> 00:08:54,870
Well, we're using circuits
that we're paying for.

177
00:08:54,870 --> 00:08:58,030
If we're paying for them,
we should be using that.

178
00:08:58,030 --> 00:09:01,780
Now, one thing to keep in
mind is, in this environment,

179
00:09:01,780 --> 00:09:03,630
if there is a failure.

180
00:09:03,630 --> 00:09:05,130
And of course,
we're really talking

181
00:09:05,130 --> 00:09:09,450
about a failure of these
secondary circuits or possibly

182
00:09:09,450 --> 00:09:10,860
even the primary circuit.

183
00:09:10,860 --> 00:09:13,870
But either way, that
remaining traffic

184
00:09:13,870 --> 00:09:17,140
is still going to have
to traverse the remaining

185
00:09:17,140 --> 00:09:19,540
path, whatever that may be.

186
00:09:19,540 --> 00:09:22,000
And another cool thing
with traffic engineering

187
00:09:22,000 --> 00:09:25,870
is we can actually put
priorities on our tunnels.

188
00:09:25,870 --> 00:09:31,220
And certain tunnels can bump
lesser priority tunnels.

189
00:09:31,220 --> 00:09:35,450
So we can choose, in the event
of lesser bandwidth available,

190
00:09:35,450 --> 00:09:38,170
which tunnels stay up
and which tunnels--

191
00:09:38,170 --> 00:09:39,940
may be something
like a tunnel we

192
00:09:39,940 --> 00:09:42,460
built for TestDev or
something like that

193
00:09:42,460 --> 00:09:44,920
or non-mission
critical information

194
00:09:44,920 --> 00:09:48,610
can go down rather than
our mission critical.

195
00:09:48,610 --> 00:09:54,540
It's essentially QoS for
traffic engineering tunnels.

196
00:09:54,540 --> 00:09:56,690
Now, just to put this out
there-- and hopefully,

197
00:09:56,690 --> 00:09:58,220
it's sort of obvious.

198
00:09:58,220 --> 00:10:01,500
But this is only
beneficial, of course,

199
00:10:01,500 --> 00:10:04,130
if there are resources
available, which

200
00:10:04,130 --> 00:10:06,320
are, in fact, underutilized.

201
00:10:06,320 --> 00:10:08,990
If there's no resources
that we're not using,

202
00:10:08,990 --> 00:10:11,630
well, this isn't really
going to help us much.

203
00:10:11,630 --> 00:10:16,310
Because this is helping us use
those underutilized resources.

204
00:10:16,310 --> 00:10:21,200
This is really designed to
solve long-term issues, not

205
00:10:21,200 --> 00:10:24,530
short term problems
like traffic bursts.

206
00:10:24,530 --> 00:10:29,750
This would be more of an
issue for say, QoS or quality

207
00:10:29,750 --> 00:10:31,010
of service.

208
00:10:31,010 --> 00:10:34,910
QoS and doing things like
traffic prioritization

209
00:10:34,910 --> 00:10:38,060
and having egress
buffers, things like this,

210
00:10:38,060 --> 00:10:42,200
those are more to help short
term traffic burst problems.

211
00:10:42,200 --> 00:10:45,080
This is a long
term, always going

212
00:10:45,080 --> 00:10:48,410
to be using these back
up underutilized paths.

213
00:10:48,410 --> 00:10:51,980
And truth be told, we want this
to be a long term solution,

214
00:10:51,980 --> 00:10:56,700
because we always want to be
using underutilized resources.

215
00:10:56,700 --> 00:10:59,350
We don't want to just
use them temporarily.

216
00:10:59,350 --> 00:11:01,330
The point we're really
trying to make here

217
00:11:01,330 --> 00:11:04,180
is these tunnels have
to be built and set up.

218
00:11:04,180 --> 00:11:07,970
And traffic has to be steered
into these tunnels and so on.

219
00:11:07,970 --> 00:11:09,520
This is not a quick
thing we're going

220
00:11:09,520 --> 00:11:13,150
to do just to resolve
30 milliseconds or even

221
00:11:13,150 --> 00:11:16,210
500 milliseconds worth
of traffic congestion.

222
00:11:16,210 --> 00:11:19,000
This is something we're going
to set up for the long term.

223
00:11:19,000 --> 00:11:21,250
That's really the point
we're trying to drive home

224
00:11:21,250 --> 00:11:23,630
with this particular point.

225
00:11:23,630 --> 00:11:25,720
Now, just for subject
completeness here,

226
00:11:25,720 --> 00:11:29,650
there are, in fact, other
ways that we could accomplish,

227
00:11:29,650 --> 00:11:30,380
the same thing.

228
00:11:30,380 --> 00:11:32,410
We're doing with
traffic engineering.

229
00:11:32,410 --> 00:11:35,810
That company I mentioned
earlier with the stock trading.

230
00:11:35,810 --> 00:11:38,020
What we had actually
done there for them

231
00:11:38,020 --> 00:11:40,720
is policy-based routing.

232
00:11:40,720 --> 00:11:43,240
We did their solution with PBR.

233
00:11:43,240 --> 00:11:46,370
And that worked fine
in their environment.

234
00:11:46,370 --> 00:11:49,780
It's not going to be so good in
a service provider environment,

235
00:11:49,780 --> 00:11:53,560
because policy-based routing
has the same problems

236
00:11:53,560 --> 00:11:57,580
as source-based routing
and static routing.

237
00:11:57,580 --> 00:12:00,400
All of these have
the same issue, which

238
00:12:00,400 --> 00:12:02,482
is they don't scale very well.

239
00:12:02,482 --> 00:12:04,190
And one of the biggest
things, of course,

240
00:12:04,190 --> 00:12:07,090
we need to be concerned with in
a service provider environment

241
00:12:07,090 --> 00:12:09,220
is scalability.

242
00:12:09,220 --> 00:12:12,770
Source-based routing, just
to add insult to injury,

243
00:12:12,770 --> 00:12:15,200
not only does it
not scale very well,

244
00:12:15,200 --> 00:12:19,390
it also introduces
some security concerns.

245
00:12:19,390 --> 00:12:21,910
In this video, we
just took a look

246
00:12:21,910 --> 00:12:24,640
at why we want traffic
engineering, what

247
00:12:24,640 --> 00:12:26,590
it brings to the table
that we don't get

248
00:12:26,590 --> 00:12:29,180
from normal routing solutions.

249
00:12:29,180 --> 00:12:30,940
The next thing to
do is going to be

250
00:12:30,940 --> 00:12:33,910
to take a look at some of
the more detailed information

251
00:12:33,910 --> 00:12:37,270
about how traffic engineering
actually functions

252
00:12:37,270 --> 00:12:39,220
and some of its core components.

253
00:12:39,220 --> 00:12:42,260
And that's exactly
what we get to do next.

254
00:12:42,260 --> 00:12:45,340
Until then, I hope this has
been informative for you,

255
00:12:45,340 --> 00:12:48,360
and I'd like to thank
you for viewing.

256
00:12:48,360 --> 00:12:52,000