1
00:00:00,000 --> 00:00:07,880

2
00:00:07,880 --> 00:00:12,350
So let's wrap up by looking
at topology independent

3
00:00:12,350 --> 00:00:14,210
loop free alternative.

4
00:00:14,210 --> 00:00:16,940
Our main focus here
is going to be looking

5
00:00:16,940 --> 00:00:20,300
at the differences
between traditional loop

6
00:00:20,300 --> 00:00:23,750
free alternative and
topology independent.

7
00:00:23,750 --> 00:00:25,200
What's the difference?

8
00:00:25,200 --> 00:00:26,700
What does that mean?

9
00:00:26,700 --> 00:00:28,970
Keep in mind that
the overall idea

10
00:00:28,970 --> 00:00:33,440
behind loop free alternative is
that our routing protocol will

11
00:00:33,440 --> 00:00:38,000
have pre-chosen a repair
path or a backup path

12
00:00:38,000 --> 00:00:41,610
that it's going to use in
the event of a failure.

13
00:00:41,610 --> 00:00:45,770
Now, the problem with loop free
alternative, the legacy method,

14
00:00:45,770 --> 00:00:49,980
is it doesn't always
find the loop free path.

15
00:00:49,980 --> 00:00:54,020
However, with topology
independent, it will.

16
00:00:54,020 --> 00:00:56,010
Now, this is
assuming, of course,

17
00:00:56,010 --> 00:00:58,770
that a backup path
actually exists.

18
00:00:58,770 --> 00:01:02,690
The other thing we want to do
is avoid suboptimal routing.

19
00:01:02,690 --> 00:01:06,140
Traditional LFA in the
wrong circumstances

20
00:01:06,140 --> 00:01:08,940
could lead to
suboptimal routing.

21
00:01:08,940 --> 00:01:12,330
And one more thing to keep
in mind as we jump in here,

22
00:01:12,330 --> 00:01:15,480
we're talking about
loop free alternative,

23
00:01:15,480 --> 00:01:21,690
meaning having that backup path
ready if there's a failure.

24
00:01:21,690 --> 00:01:26,730
We are not talking about regular
routing protocol convergence.

25
00:01:26,730 --> 00:01:29,430
So throughout this
video when we say things

26
00:01:29,430 --> 00:01:32,580
like it can't use this path
or it won't use that path,

27
00:01:32,580 --> 00:01:36,510
we mean for LFA,
in this context.

28
00:01:36,510 --> 00:01:40,740
Of course, if there's a failure,
and there's another path,

29
00:01:40,740 --> 00:01:44,490
the routing protocol will
find it, it will converge.

30
00:01:44,490 --> 00:01:48,330
We're talking about what
paths it will find as a backup

31
00:01:48,330 --> 00:01:51,600
before it has to
completely reconverge.

32
00:01:51,600 --> 00:01:54,810
So all of that said, let's
go ahead and jump in and take

33
00:01:54,810 --> 00:01:57,540
a more detailed look
at topology independent

34
00:01:57,540 --> 00:02:00,420
loop free alternative
and how that compares

35
00:02:00,420 --> 00:02:05,570
to traditional or legacy LFA.

36
00:02:05,570 --> 00:02:10,310
So TI-LFA is a
resiliency feature

37
00:02:10,310 --> 00:02:14,330
that allows for sub 50
millisecond convergence

38
00:02:14,330 --> 00:02:17,030
because what it does
is it calculates

39
00:02:17,030 --> 00:02:21,770
that repair path before a
failure actually occurs.

40
00:02:21,770 --> 00:02:26,180
Now, one thing this has over
classic or traditional LFA

41
00:02:26,180 --> 00:02:30,410
is this can provide
100% coverage

42
00:02:30,410 --> 00:02:33,950
because it's
topology independent.

43
00:02:33,950 --> 00:02:39,410
Classic LFA is, of
course topology dependent

44
00:02:39,410 --> 00:02:44,250
and therefore may not always
provide a backup path.

45
00:02:44,250 --> 00:02:46,650
Let's take a look
at what that means.

46
00:02:46,650 --> 00:02:49,520
So let's bring up a
diagram here, and let's

47
00:02:49,520 --> 00:02:52,040
put some costs on these links.

48
00:02:52,040 --> 00:02:58,130
We'll make this one a 10,
10, make this one a 20.

49
00:02:58,130 --> 00:03:01,010
This one will be
a 10, and this is

50
00:03:01,010 --> 00:03:05,310
going to be the suboptimal
path up here through router 5,

51
00:03:05,310 --> 00:03:08,370
so we'll make that
one a cost of 100.

52
00:03:08,370 --> 00:03:10,500
During this whole
discussion, we'll

53
00:03:10,500 --> 00:03:13,590
be talking about traffic
going from this source

54
00:03:13,590 --> 00:03:15,338
to this destination.

55
00:03:15,338 --> 00:03:17,130
So we're not going to
focus on those a lot.

56
00:03:17,130 --> 00:03:19,200
Just keep in mind that
that's the traffic

57
00:03:19,200 --> 00:03:20,970
path we're talking about.

58
00:03:20,970 --> 00:03:24,300
We would, of course, normally
send this to router 1.

59
00:03:24,300 --> 00:03:27,540
Router 1 would typically
send this to router 2

60
00:03:27,540 --> 00:03:29,190
because it's a cost of 10.

61
00:03:29,190 --> 00:03:32,700
And of course, it would then get
delivered to the destination.

62
00:03:32,700 --> 00:03:34,320
But the problem,
of course, comes

63
00:03:34,320 --> 00:03:38,830
in is what happens when
this link goes down.

64
00:03:38,830 --> 00:03:41,160
This is where the
question comes in.

65
00:03:41,160 --> 00:03:44,110
Do we have a loop
free alternative?

66
00:03:44,110 --> 00:03:46,950
So in this case,
let's talk about what

67
00:03:46,950 --> 00:03:50,745
would be happening
down here on router 3.

68
00:03:50,745 --> 00:03:53,730
Router 3 would have been
going through these two

69
00:03:53,730 --> 00:03:59,700
links with a cost of 10
each for a total cost of 20.

70
00:03:59,700 --> 00:04:02,850
If it had gone across the bottom
that, of course, would have

71
00:04:02,850 --> 00:04:05,370
been a cost of 30, and
that's not the path

72
00:04:05,370 --> 00:04:06,840
we would have chosen.

73
00:04:06,840 --> 00:04:09,360
The problem is
router 3's path would

74
00:04:09,360 --> 00:04:13,570
have been pointing up to router
1 to get to the destination.

75
00:04:13,570 --> 00:04:17,399
So if we tried to use that
as an alternative path,

76
00:04:17,399 --> 00:04:20,670
the source would send
the traffic to router 1.

77
00:04:20,670 --> 00:04:23,130
Router 1 would send
it to router 3.

78
00:04:23,130 --> 00:04:25,980
Router 3 would send it
right back up to router 1,

79
00:04:25,980 --> 00:04:28,690
and right there's
your routing loop.

80
00:04:28,690 --> 00:04:33,960
So in this example, there's
not a loop free alternative

81
00:04:33,960 --> 00:04:36,060
using the classic method.

82
00:04:36,060 --> 00:04:37,980
And that's what
we want to address

83
00:04:37,980 --> 00:04:41,130
with topology independent
loop free alternative

84
00:04:41,130 --> 00:04:45,120
because we, as humans,
can look at this

85
00:04:45,120 --> 00:04:49,320
and see that there's clearly
an alternative path where

86
00:04:49,320 --> 00:04:53,830
we could go to 4 and then 2
and then to the destination.

87
00:04:53,830 --> 00:04:56,280
And it's not even that
much higher of a cost.

88
00:04:56,280 --> 00:04:58,950
It just would have
been a cost of 30.

89
00:04:58,950 --> 00:05:02,750
However, traditional LFA
will not address that issue.

90
00:05:02,750 --> 00:05:06,400
That's where TI-LFA can help.

91
00:05:06,400 --> 00:05:09,280
Now TI-LFA can
provide protection

92
00:05:09,280 --> 00:05:14,210
for the link, a node, or
a shared risk link group,

93
00:05:14,210 --> 00:05:17,350
which is something you as the
administrator create to tell

94
00:05:17,350 --> 00:05:19,540
it not to use certain links.

95
00:05:19,540 --> 00:05:22,660
And it does prevent
suboptimal routing

96
00:05:22,660 --> 00:05:25,120
because what could
have happened here,

97
00:05:25,120 --> 00:05:27,820
we could have possibly chosen
to go through this path

98
00:05:27,820 --> 00:05:30,520
through the top
through router 5.

99
00:05:30,520 --> 00:05:34,520
That could have possibly
been a loop free alternative.

100
00:05:34,520 --> 00:05:38,200
However, that's a
really suboptimal path.

101
00:05:38,200 --> 00:05:41,950
So if router 3 did
choose to go to router 1,

102
00:05:41,950 --> 00:05:45,580
and router 1 decided
to go to router 5,

103
00:05:45,580 --> 00:05:47,950
would the traffic get
to the destination?

104
00:05:47,950 --> 00:05:51,910
Yes, most likely, but definitely
through a suboptimal path.

105
00:05:51,910 --> 00:05:56,140
This would end up
being a cost of 200,

106
00:05:56,140 --> 00:05:59,170
where with TI-LFA
using the bottom path,

107
00:05:59,170 --> 00:06:01,090
it only would have been 30.

108
00:06:01,090 --> 00:06:05,530
So even if LFA could have
provided this loop free path,

109
00:06:05,530 --> 00:06:09,080
it would have been
suboptimal in this topology.

110
00:06:09,080 --> 00:06:12,440
So this is all the things
we're trying to avoid.

111
00:06:12,440 --> 00:06:14,380
And the cool thing
about this whole thing

112
00:06:14,380 --> 00:06:17,110
is this is a very,
very simple operation.

113
00:06:17,110 --> 00:06:22,400
This is all computed by the
IGP, so either OSPF or IS-IS.

114
00:06:22,400 --> 00:06:25,300

115
00:06:25,300 --> 00:06:28,450
TI-LFA works by
calculating what's

116
00:06:28,450 --> 00:06:32,410
called the post convergence
path, the path we're

117
00:06:32,410 --> 00:06:36,610
going to use after the
convergence of, well, just

118
00:06:36,610 --> 00:06:38,750
realizing the link is down.

119
00:06:38,750 --> 00:06:42,160
So this is, again, going to be
calculated before any failure.

120
00:06:42,160 --> 00:06:43,720
That's the whole point here.

121
00:06:43,720 --> 00:06:49,970
And this path, again, may
not actually be an LFA.

122
00:06:49,970 --> 00:06:53,600
Remember that router would have
this looping situation going on

123
00:06:53,600 --> 00:06:58,220
over here if this link went
down and therefore not really

124
00:06:58,220 --> 00:06:59,420
have an LFA.

125
00:06:59,420 --> 00:07:02,690
But with TI-LFA,
the idea is we want

126
00:07:02,690 --> 00:07:05,990
to be able to use this
path through the bottom

127
00:07:05,990 --> 00:07:08,400
to get to the destination.

128
00:07:08,400 --> 00:07:11,780
And as we said, with
traditional LFA,

129
00:07:11,780 --> 00:07:15,410
this would not be
a loop free path.

130
00:07:15,410 --> 00:07:20,540
The way we get around this is
we enforce a loop free path

131
00:07:20,540 --> 00:07:23,600
by using the segment list.

132
00:07:23,600 --> 00:07:25,700
Essentially what's
happening here

133
00:07:25,700 --> 00:07:30,620
is router 1 is
making the decision

134
00:07:30,620 --> 00:07:33,650
to go through routers
3, 4, 2, and then

135
00:07:33,650 --> 00:07:36,980
to the destination, this
path across the bottom.

136
00:07:36,980 --> 00:07:41,720
The way we stop router 3
from engaging in this loop

137
00:07:41,720 --> 00:07:44,840
is by using our segment list.

138
00:07:44,840 --> 00:07:48,710
Remember, we tell router
3 to go to router 4.

139
00:07:48,710 --> 00:07:51,930
This is where that
source-based routing kicks in

140
00:07:51,930 --> 00:07:55,030
to force this loop free path.

141
00:07:55,030 --> 00:07:57,220
So the real question
then becomes,

142
00:07:57,220 --> 00:08:01,320
how does TI-LFA determine
this path to begin with?

143
00:08:01,320 --> 00:08:07,320
If legacy LFA can't do it,
then how does TI-LFA actually

144
00:08:07,320 --> 00:08:09,870
pull this off?

145
00:08:09,870 --> 00:08:13,710
One of the key factors
here is that TI-LFA

146
00:08:13,710 --> 00:08:18,150
uses the concept of two
different spaces, the P

147
00:08:18,150 --> 00:08:20,280
space and the Q space.

148
00:08:20,280 --> 00:08:25,260
Both spaces are calculating
using pre-convergence paths.

149
00:08:25,260 --> 00:08:29,230
Remember this is all calculated
before there is a failure.

150
00:08:29,230 --> 00:08:33,360
The P space is a
set of nodes that

151
00:08:33,360 --> 00:08:38,640
are reachable from the point
of local repair or the PLR.

152
00:08:38,640 --> 00:08:44,400
In our example, router
1 would be the PLR.

153
00:08:44,400 --> 00:08:46,800
This is the point
of local repair.

154
00:08:46,800 --> 00:08:49,330
So let's think about
this for a moment.

155
00:08:49,330 --> 00:08:55,980
Which set of nodes are reachable
from the point of local repair

156
00:08:55,980 --> 00:08:59,410
without using the
protected link?

157
00:08:59,410 --> 00:09:03,750
So let's remember that we were
saying this was a cost of 100.

158
00:09:03,750 --> 00:09:05,646
This was 100.

159
00:09:05,646 --> 00:09:12,310
We had a 10, a 10,
a 10, and a 20.

160
00:09:12,310 --> 00:09:17,920
So where can the PLR get without
using the protected link?

161
00:09:17,920 --> 00:09:20,500
Remember we're focusing
on this link right

162
00:09:20,500 --> 00:09:24,170
here as being the link that
we're trying to protect.

163
00:09:24,170 --> 00:09:28,540
So where can router 1 get
without using that link?

164
00:09:28,540 --> 00:09:30,580
Well, let's think about this.

165
00:09:30,580 --> 00:09:33,220
It can definitely
get to router 3.

166
00:09:33,220 --> 00:09:36,550
So router 3 is definitely
going to be in the P space.

167
00:09:36,550 --> 00:09:39,505
It's a cost of 10
to get to router 3,

168
00:09:39,505 --> 00:09:41,590
and there's certainly
no better path.

169
00:09:41,590 --> 00:09:45,400
If it tried to go from
router 1 to router 2

170
00:09:45,400 --> 00:09:49,840
to router 4 to router 3,
that would be a cost of 40

171
00:09:49,840 --> 00:09:51,370
to go around that way.

172
00:09:51,370 --> 00:09:53,500
And of course, going
up through the top

173
00:09:53,500 --> 00:09:56,150
would definitely not
be a better option.

174
00:09:56,150 --> 00:09:59,830
So router 1 can get to
router 3, no problem.

175
00:09:59,830 --> 00:10:01,780
That's in the P space.

176
00:10:01,780 --> 00:10:05,560
Router 4 would also
be in the P space

177
00:10:05,560 --> 00:10:10,780
because, again, it's 10 and 10
to go through the bottom, where

178
00:10:10,780 --> 00:10:15,310
it would be 10 and 20 to go
through the protected link.

179
00:10:15,310 --> 00:10:18,460
So it wouldn't choose to go
through the protected link.

180
00:10:18,460 --> 00:10:21,370
And we could say the same
thing about router 5.

181
00:10:21,370 --> 00:10:25,090
Yes, it's a cost of
100 to get to router 5,

182
00:10:25,090 --> 00:10:28,150
but going through router 2
through the protected link

183
00:10:28,150 --> 00:10:29,920
would be 110.

184
00:10:29,920 --> 00:10:31,690
That would be even more.

185
00:10:31,690 --> 00:10:33,910
So these are the
routers that are

186
00:10:33,910 --> 00:10:36,010
reachable through the P space.

187
00:10:36,010 --> 00:10:39,130
Of course, router 2 is
not because router 2

188
00:10:39,130 --> 00:10:42,020
would literally be trying
to use the protected link.

189
00:10:42,020 --> 00:10:44,980
So these are the
P space routers.

190
00:10:44,980 --> 00:10:47,740
Again, the routers
that can be reached

191
00:10:47,740 --> 00:10:52,480
by the point of local repair
without using the protected

192
00:10:52,480 --> 00:10:55,950
link using standard metrics.

193
00:10:55,950 --> 00:10:58,490
Next is the Q space.

194
00:10:58,490 --> 00:11:03,320
These are the nodes that
can reach the destination

195
00:11:03,320 --> 00:11:06,060
without using the
protected link.

196
00:11:06,060 --> 00:11:09,080
So let's go back down to
look at our diagram again.

197
00:11:09,080 --> 00:11:11,870
Of course, router 1
can't because it's

198
00:11:11,870 --> 00:11:14,210
going straight across
the protected link

199
00:11:14,210 --> 00:11:17,420
to get to router 2 and
therefore the destination.

200
00:11:17,420 --> 00:11:22,820
Router 3, well, router
3 would prefer 10 and 10

201
00:11:22,820 --> 00:11:27,290
over 10 and 20,
so router 3 would

202
00:11:27,290 --> 00:11:29,400
be using the protected link.

203
00:11:29,400 --> 00:11:33,410
Therefore router 3 is
not in the Q space.

204
00:11:33,410 --> 00:11:35,000
Router 4.

205
00:11:35,000 --> 00:11:39,740
Router 4 would, of
course, go the path of 20

206
00:11:39,740 --> 00:11:44,520
rather than 10, 20, 30 to
go around the other way.

207
00:11:44,520 --> 00:11:48,740
So router 4 is in the Q space.

208
00:11:48,740 --> 00:11:52,430
And then router 5, of course, is
going to take the path directly

209
00:11:52,430 --> 00:11:55,950
to router 2 so that would
make router 5 in the Q space

210
00:11:55,950 --> 00:11:56,830
as well.

211
00:11:56,830 --> 00:12:01,120
And of course, router 2 goes
directly to the destination.

212
00:12:01,120 --> 00:12:04,470
So it's definitely
in the Q space.

213
00:12:04,470 --> 00:12:08,190
So once the Q and the P
spaces have been defined,

214
00:12:08,190 --> 00:12:10,360
couple things to keep in mind.

215
00:12:10,360 --> 00:12:14,220
First, we can certainly
have PQ routers.

216
00:12:14,220 --> 00:12:17,790
This is a router that's in both
the P space and the Q space.

217
00:12:17,790 --> 00:12:19,280
And of course, in
our example, we

218
00:12:19,280 --> 00:12:21,420
can see that router
5 and router 4

219
00:12:21,420 --> 00:12:24,810
are both in the P and Q spaces.

220
00:12:24,810 --> 00:12:27,990
And then ultimately,
the final decision here

221
00:12:27,990 --> 00:12:31,890
is realizing that the
packet can then be released

222
00:12:31,890 --> 00:12:35,500
on any router in the Q space.

223
00:12:35,500 --> 00:12:37,500
So essentially what
it's going to do

224
00:12:37,500 --> 00:12:40,320
is router 1 is going to
look at this and say, OK

225
00:12:40,320 --> 00:12:45,190
the TI-LFA path I could
release this on router 4.

226
00:12:45,190 --> 00:12:48,610
I could release it on router
2 as well or router 5.

227
00:12:48,610 --> 00:12:51,670
But router 4 is going to be
the best path because it only

228
00:12:51,670 --> 00:12:55,480
costs me 20 to get to router 4.

229
00:12:55,480 --> 00:12:58,910
So it's going to try to
go across the bottom.

230
00:12:58,910 --> 00:13:02,440
So ultimately, what
this means is the source

231
00:13:02,440 --> 00:13:04,540
sends the update to router 1.

232
00:13:04,540 --> 00:13:08,770
Router 1 sends it to router 3
with the appropriate segment

233
00:13:08,770 --> 00:13:11,350
list on it so that
it doesn't loop.

234
00:13:11,350 --> 00:13:14,140
Router 3 sends it to router 4.

235
00:13:14,140 --> 00:13:17,200
Router 4 sends it
up to router 2.

236
00:13:17,200 --> 00:13:19,690
Router 2 sends it
to the destination,

237
00:13:19,690 --> 00:13:24,130
and we've provided a topology
independent loop free path

238
00:13:24,130 --> 00:13:28,150
to the destination without
using the protected link.

239
00:13:28,150 --> 00:13:31,600
And again, remember all
of this is calculated

240
00:13:31,600 --> 00:13:35,450
before a failure occurs.

241
00:13:35,450 --> 00:13:40,640
In this video, we took a look
at topology independent LFA.

242
00:13:40,640 --> 00:13:45,230
And unlike traditional
or classic LFA,

243
00:13:45,230 --> 00:13:49,130
TI-LFA will always
provide a backup path,

244
00:13:49,130 --> 00:13:51,740
assuming one actually exists.

245
00:13:51,740 --> 00:13:55,590
Of course, if there's a single
path, we can't back that up.

246
00:13:55,590 --> 00:13:59,210
But if there is one,
TI-LFA will find it

247
00:13:59,210 --> 00:14:03,630
by using these two separate
spaces, the P space and the Q

248
00:14:03,630 --> 00:14:08,720
space and then using our
source-based segment routing

249
00:14:08,720 --> 00:14:11,330
to get that traffic
through the environment

250
00:14:11,330 --> 00:14:15,020
to the necessary
router without creating

251
00:14:15,020 --> 00:14:17,780
a loop, which as
we demonstrated,

252
00:14:17,780 --> 00:14:21,410
could clearly happen
with classic LFA,

253
00:14:21,410 --> 00:14:23,150
and that's why
there's not always

254
00:14:23,150 --> 00:14:26,630
a backup path in
traditional LFA.

255
00:14:26,630 --> 00:14:28,730
So that wraps up
everything that we

256
00:14:28,730 --> 00:14:31,760
need to discuss in
this skill, and that

257
00:14:31,760 --> 00:14:35,880
leaves us with just the
review in the form of a quiz.

258
00:14:35,880 --> 00:14:38,490
And that's exactly what
we're going to do next.

259
00:14:38,490 --> 00:14:41,940
Till then, I hope this has
been informative for you,

260
00:14:41,940 --> 00:14:45,050
and I'd like to thank
you for viewing.

261
00:14:45,050 --> 00:14:48,000