1
00:00:00,000 --> 00:00:07,150

2
00:00:07,150 --> 00:00:10,060
So in this video,
let's start off

3
00:00:10,060 --> 00:00:14,810
by looking at the BGP path
selection criteria as a whole.

4
00:00:14,810 --> 00:00:17,330
So we're going to go through
all of the different steps.

5
00:00:17,330 --> 00:00:19,790
Now some of the
steps are, of course,

6
00:00:19,790 --> 00:00:22,760
going to be broken
down in future videos,

7
00:00:22,760 --> 00:00:26,320
the ones that require a little
bit more discussion, and/or you

8
00:00:26,320 --> 00:00:28,840
could say the ones that
we use more commonly

9
00:00:28,840 --> 00:00:30,940
for path selection.

10
00:00:30,940 --> 00:00:33,940
Some of the other ones,
we will cover in full

11
00:00:33,940 --> 00:00:35,780
here in this video.

12
00:00:35,780 --> 00:00:37,390
But there are some
of the rules that

13
00:00:37,390 --> 00:00:40,030
sound relatively
straightforward that

14
00:00:40,030 --> 00:00:42,200
do have some caveats with them.

15
00:00:42,200 --> 00:00:45,160
So we're going to make sure
we cover those either here

16
00:00:45,160 --> 00:00:48,190
or in the future
videos, when we cover

17
00:00:48,190 --> 00:00:50,330
those topics individually.

18
00:00:50,330 --> 00:00:53,380
And we will make sure you
understand which ones are which

19
00:00:53,380 --> 00:00:54,850
as we go through.

20
00:00:54,850 --> 00:00:58,360
And I'll mention whether this
is going to be in a future video

21
00:00:58,360 --> 00:01:01,550
or if this is the most that
we're going to be covering it.

22
00:01:01,550 --> 00:01:04,840
So let's jump in and look at
the path selection criteria,

23
00:01:04,840 --> 00:01:07,750
looking at, well,
some of the details,

24
00:01:07,750 --> 00:01:09,970
and saving some
for future videos.

25
00:01:09,970 --> 00:01:12,950

26
00:01:12,950 --> 00:01:14,960
Our first step is
that we're going

27
00:01:14,960 --> 00:01:19,580
to prefer the path with
the highest weight.

28
00:01:19,580 --> 00:01:23,990
Now weight is actually a
Cisco proprietary attribute.

29
00:01:23,990 --> 00:01:29,060
And as such, it does not
get sent in BGP updates.

30
00:01:29,060 --> 00:01:33,470
Note, as we go down through the
path selection criteria, some

31
00:01:33,470 --> 00:01:38,580
of these rules, such as rule one
here, will be listed in blue.

32
00:01:38,580 --> 00:01:41,360
Some of them will be in yellow.

33
00:01:41,360 --> 00:01:45,470
The blue indicates that
there's a more detailed video

34
00:01:45,470 --> 00:01:49,580
on how this works coming
up later in this skill.

35
00:01:49,580 --> 00:01:52,190
If the rule is
yellow, it means we're

36
00:01:52,190 --> 00:01:54,140
going to have the
full discussion on it

37
00:01:54,140 --> 00:01:58,460
here in this video, because
it's a fairly simple rule,

38
00:01:58,460 --> 00:02:00,740
and maybe it just has
a few little caveats.

39
00:02:00,740 --> 00:02:03,050
So keep that in mind
as we go down here.

40
00:02:03,050 --> 00:02:06,080
We're not going to be talking
a lot about weight right now.

41
00:02:06,080 --> 00:02:10,770
Because again, there's a
separate video coming up next.

42
00:02:10,770 --> 00:02:13,920
Sort of along the same
lines is rule number two.

43
00:02:13,920 --> 00:02:18,150
This is the highest
local preference.

44
00:02:18,150 --> 00:02:20,550
This one is not
Cisco proprietary.

45
00:02:20,550 --> 00:02:24,000
It is part of the BGP
standard, and therefore

46
00:02:24,000 --> 00:02:26,850
is carried in BGP attributes.

47
00:02:26,850 --> 00:02:29,790
Once again, the highest wins.

48
00:02:29,790 --> 00:02:34,350
Number three is our first rule,
that's rather straightforward.

49
00:02:34,350 --> 00:02:37,230
And it simply says
that we are going

50
00:02:37,230 --> 00:02:42,150
to prefer the path that
was locally originated

51
00:02:42,150 --> 00:02:47,940
via either the network or the
aggregate BGP subcommands,

52
00:02:47,940 --> 00:02:53,460
or through redistribution
from an IGP.

53
00:02:53,460 --> 00:02:55,690
Now this is fairly
straightforward.

54
00:02:55,690 --> 00:02:59,200
It's just saying that locally
originated is preferred.

55
00:02:59,200 --> 00:03:03,270
However, note the caveat, adding
a little more detail here,

56
00:03:03,270 --> 00:03:07,650
that local paths coming
from either network

57
00:03:07,650 --> 00:03:12,930
or redistribution are
preferred over something

58
00:03:12,930 --> 00:03:16,030
coming from the aggregate
address command.

59
00:03:16,030 --> 00:03:19,080
So even within the
locally originated,

60
00:03:19,080 --> 00:03:22,170
there's still a
hierarchy where, again,

61
00:03:22,170 --> 00:03:27,210
network or redistribution
beats out an aggregate address.

62
00:03:27,210 --> 00:03:30,540
Our next rule is simply
to prefer the path

63
00:03:30,540 --> 00:03:34,020
with the shortest AS path.

64
00:03:34,020 --> 00:03:36,630
When we say shortest,
we're talking

65
00:03:36,630 --> 00:03:42,730
about the number of AS's,
not the actual length.

66
00:03:42,730 --> 00:03:55,680
So for example, if I had a
path 65001 versus the path 123,

67
00:03:55,680 --> 00:03:59,820
the winner here is
going to be the 65001.

68
00:03:59,820 --> 00:04:02,250
Because even though the bottom
path might only be three

69
00:04:02,250 --> 00:04:06,060
numbers, it's three
separate AS's.

70
00:04:06,060 --> 00:04:10,110
So again, it's the number
of AS's in the path.

71
00:04:10,110 --> 00:04:13,350
Next is another fairly
straightforward rule.

72
00:04:13,350 --> 00:04:18,550
We prefer the path with
the lowest origin type.

73
00:04:18,550 --> 00:04:24,060
So the lowest here is
an I, which means IGP.

74
00:04:24,060 --> 00:04:27,600
In the case of a Cisco
device, this generally

75
00:04:27,600 --> 00:04:32,040
means that the route was brought
in with the network statement.

76
00:04:32,040 --> 00:04:35,250
Next would be E, meaning EGP.

77
00:04:35,250 --> 00:04:38,010
This is generally not
going to be set anymore

78
00:04:38,010 --> 00:04:41,850
unless an administrator
actually set the origin code

79
00:04:41,850 --> 00:04:48,210
purposefully to E. And then
finally is the question mark.

80
00:04:48,210 --> 00:04:51,420
This simply means
unknown origin.

81
00:04:51,420 --> 00:04:55,650
On Cisco devices,
this generally means

82
00:04:55,650 --> 00:04:59,970
that this path was originated
through redistribution.

83
00:04:59,970 --> 00:05:01,920
So all we really
have to remember

84
00:05:01,920 --> 00:05:04,470
is that I is going
to beat E, and E

85
00:05:04,470 --> 00:05:07,480
is going to be question mark.

86
00:05:07,480 --> 00:05:09,940
Now again another
quick note here--

87
00:05:09,940 --> 00:05:13,930
origin code is not
very commonly used

88
00:05:13,930 --> 00:05:17,110
to manipulate path
selection, which

89
00:05:17,110 --> 00:05:20,570
is why we're not discussing
it in great detail here.

90
00:05:20,570 --> 00:05:25,480
However, this is still a part
of the path selection criteria,

91
00:05:25,480 --> 00:05:29,920
and we can set this through
route maps and route policy

92
00:05:29,920 --> 00:05:30,700
language.

93
00:05:30,700 --> 00:05:33,910
So keep in mind,
although not common,

94
00:05:33,910 --> 00:05:38,920
we can certainly use the origin
code to manipulate BGP path

95
00:05:38,920 --> 00:05:41,650
selection if we wanted to.

96
00:05:41,650 --> 00:05:47,620
Next up is to prefer the path
with the lowest multi exit

97
00:05:47,620 --> 00:05:50,770
discriminator, or MED.

98
00:05:50,770 --> 00:05:53,890
Cisco generally refers
to this as the metric

99
00:05:53,890 --> 00:05:55,330
when you're configuring it.

100
00:05:55,330 --> 00:05:58,540
And in show and
debug commands, there

101
00:05:58,540 --> 00:06:01,390
are a few exceptions where
they refer to it as the MED,

102
00:06:01,390 --> 00:06:06,190
such as when you're setting
it in RPL on iOS XR.

103
00:06:06,190 --> 00:06:09,130
But generally, speaking
when you see metric,

104
00:06:09,130 --> 00:06:10,900
they're referring to the MED.

105
00:06:10,900 --> 00:06:13,120
And again, this is
something that we have

106
00:06:13,120 --> 00:06:14,810
another video on coming up.

107
00:06:14,810 --> 00:06:18,460
So we're not going to spend a
lot of detail on it right now.

108
00:06:18,460 --> 00:06:24,700
Next up is fairly simply to
prefer external paths, or EBGP

109
00:06:24,700 --> 00:06:29,810
paths, over internal
paths, or IBGP paths.

110
00:06:29,810 --> 00:06:33,170
Now once again, a
few little caveats.

111
00:06:33,170 --> 00:06:36,340
First, if best path
is selected, we're

112
00:06:36,340 --> 00:06:38,650
going to continue
on to step nine,

113
00:06:38,650 --> 00:06:40,750
or multipathing, which
we're going to be

114
00:06:40,750 --> 00:06:43,880
talking about in just a moment.

115
00:06:43,880 --> 00:06:50,200
Also note that if these paths
contain confed attributes,

116
00:06:50,200 --> 00:06:54,040
they are considered local
to the confederation.

117
00:06:54,040 --> 00:06:58,000
And for the sake of this
path selection criteria,

118
00:06:58,000 --> 00:07:01,820
they will be treated
as internal paths.

119
00:07:01,820 --> 00:07:04,660
There's no distinction
between confederation

120
00:07:04,660 --> 00:07:07,810
external and
confederation internal.

121
00:07:07,810 --> 00:07:14,110
Next is to prefer the path
with the lowest IGP metric

122
00:07:14,110 --> 00:07:18,100
to the next hop, which is
essentially just saying

123
00:07:18,100 --> 00:07:23,830
if this traffic is leaving
the network anyway,

124
00:07:23,830 --> 00:07:28,630
then let's get it to the next
hop through the closest IGP

125
00:07:28,630 --> 00:07:29,920
exit.

126
00:07:29,920 --> 00:07:33,580
And here we will continue,
even if best path is already

127
00:07:33,580 --> 00:07:34,960
selected.

128
00:07:34,960 --> 00:07:39,100
Then we have step nine,
which is our multiple paths,

129
00:07:39,100 --> 00:07:42,240
or our multi path rule.

130
00:07:42,240 --> 00:07:48,000
Now keep in mind, BGP
will select one best path.

131
00:07:48,000 --> 00:07:50,550
That's what it's going to
advertise to its neighbors

132
00:07:50,550 --> 00:07:51,590
and things like that.

133
00:07:51,590 --> 00:07:54,390
That doesn't mean
that it can only put

134
00:07:54,390 --> 00:07:57,000
one route in the routing table.

135
00:07:57,000 --> 00:08:00,300
We can enable BGP
for multipathing,

136
00:08:00,300 --> 00:08:03,390
and that's what rule
9 here is all about.

137
00:08:03,390 --> 00:08:06,330
The default on most
Cisco platforms

138
00:08:06,330 --> 00:08:10,330
is to install one route from
BGP into the routing table.

139
00:08:10,330 --> 00:08:13,365
But if you change the max paths
to something like four, six,

140
00:08:13,365 --> 00:08:16,350
eight, something
like that, then it's

141
00:08:16,350 --> 00:08:20,310
going to hit rule nine here
and choose, essentially,

142
00:08:20,310 --> 00:08:24,990
which routes are a tie
down to this point,

143
00:08:24,990 --> 00:08:28,800
and put all of those
up to the max value

144
00:08:28,800 --> 00:08:30,970
into the routing table.

145
00:08:30,970 --> 00:08:34,440
And this is precisely
why rules seven and eight

146
00:08:34,440 --> 00:08:38,700
said to continue, even if
best path was selected.

147
00:08:38,700 --> 00:08:41,429
Because we had to still
see if they were ties

148
00:08:41,429 --> 00:08:44,440
that could be put into
the routing table.

149
00:08:44,440 --> 00:08:49,110
Now at this point, we're
going to continue only if best

150
00:08:49,110 --> 00:08:52,360
path is not yet selected.

151
00:08:52,360 --> 00:08:56,640
So if it gets down to step nine
here and we still have a tie,

152
00:08:56,640 --> 00:08:58,740
then we are going
to still keep going.

153
00:08:58,740 --> 00:09:00,720
Because again, even
though we might

154
00:09:00,720 --> 00:09:03,930
be putting all of those
prefixes into the routing table,

155
00:09:03,930 --> 00:09:07,110
they can't be chosen
as the best path.

156
00:09:07,110 --> 00:09:11,340
Because we only advertise
one path to our neighbors.

157
00:09:11,340 --> 00:09:16,890
Next is rule 10, which simply
says if both of the paths

158
00:09:16,890 --> 00:09:19,650
are external, then
it's simply going

159
00:09:19,650 --> 00:09:23,460
to go with the one that
was received first.

160
00:09:23,460 --> 00:09:26,220
The idea, of course,
being that it's proven

161
00:09:26,220 --> 00:09:28,560
itself to be the most stable.

162
00:09:28,560 --> 00:09:32,700
The idea here is, it minimizes
route flaps so that a newer

163
00:09:32,700 --> 00:09:36,330
path doesn't displace the
older one all the time, even

164
00:09:36,330 --> 00:09:40,020
if the newer path would be
the preferred route based

165
00:09:40,020 --> 00:09:45,090
on the next lesson criteria,
such as 11, 12, and 13.

166
00:09:45,090 --> 00:09:47,830
Now this step won't
always come into play.

167
00:09:47,830 --> 00:09:51,700
So we're going to skip this
step in certain criteria.

168
00:09:51,700 --> 00:09:56,370
The first is if you have
actually enabled BGP

169
00:09:56,370 --> 00:09:59,400
best path compare router ID.

170
00:09:59,400 --> 00:10:01,140
The reason for this
is, of course, we

171
00:10:01,140 --> 00:10:03,420
haven't seen router ID yet.

172
00:10:03,420 --> 00:10:06,810
That's one of these
steps, 11, 12 or 13.

173
00:10:06,810 --> 00:10:07,450
It's a mystery.

174
00:10:07,450 --> 00:10:08,280
We'll find out.

175
00:10:08,280 --> 00:10:11,200
But we haven't gotten
to router ID yet.

176
00:10:11,200 --> 00:10:13,800
So if we're telling it it
has to compare router ID,

177
00:10:13,800 --> 00:10:17,970
then clearly, we're not going
to stop here at step 10.

178
00:10:17,970 --> 00:10:19,860
It will also skip
it if the paths are

179
00:10:19,860 --> 00:10:22,770
from the same
neighboring router ID,

180
00:10:22,770 --> 00:10:26,650
or if there's no
current best path.

181
00:10:26,650 --> 00:10:28,770
It's not going to use
this as the criteria

182
00:10:28,770 --> 00:10:30,630
to choose the best path.

183
00:10:30,630 --> 00:10:33,540
And it's also not going to look
at this if the neighbor that

184
00:10:33,540 --> 00:10:36,370
was offering this goes down.

185
00:10:36,370 --> 00:10:39,090
It'll continue on
with other steps.

186
00:10:39,090 --> 00:10:41,610
The next step is step 11.

187
00:10:41,610 --> 00:10:44,070
And as you can see, the
mystery is solved now,

188
00:10:44,070 --> 00:10:48,660
because this is coming
from the lowest router ID.

189
00:10:48,660 --> 00:10:50,710
The router ID, by
default, by the way,

190
00:10:50,710 --> 00:10:54,600
is going to be the highest
IP address on the router,

191
00:10:54,600 --> 00:10:58,500
with preference given
to loopback interfaces.

192
00:10:58,500 --> 00:11:02,700
Or more preferably, you
as the administrator

193
00:11:02,700 --> 00:11:05,490
can use the BGP
router ID command

194
00:11:05,490 --> 00:11:09,020
and manually set the router ID.

195
00:11:09,020 --> 00:11:11,890
Now a quick note on
the router ID here--

196
00:11:11,890 --> 00:11:15,100
if you are doing
BGP synchronization,

197
00:11:15,100 --> 00:11:17,830
which is not commonly
done anymore,

198
00:11:17,830 --> 00:11:24,490
but if you are, and your
underlying IGP is OSPF,

199
00:11:24,490 --> 00:11:30,790
there is an RFC that states that
the BGP router ID and the OSPF

200
00:11:30,790 --> 00:11:35,660
router ID must match in
order to do synchronization.

201
00:11:35,660 --> 00:11:38,530
So keep that in
mind in your design

202
00:11:38,530 --> 00:11:43,030
if you are doing BGP and OSPF
with synchronization, which,

203
00:11:43,030 --> 00:11:45,140
again, not commonly done.

204
00:11:45,140 --> 00:11:48,520
But just keep that in mind
with your router ID design.

205
00:11:48,520 --> 00:11:50,590
The fact of the matter
is, you probably

206
00:11:50,590 --> 00:11:52,610
want to make them
the same anyway,

207
00:11:52,610 --> 00:11:55,360
just because why have
more than one address

208
00:11:55,360 --> 00:11:59,000
to remember a router by in
two different protocols?

209
00:11:59,000 --> 00:12:02,050
So unless you had some
outstanding reason to do so,

210
00:12:02,050 --> 00:12:04,810
we would probably make these
numbers the same anyway.

211
00:12:04,810 --> 00:12:08,380
But just keep it in mind,
that in that synchronization

212
00:12:08,380 --> 00:12:12,380
scenario, they actually
do have to match.

213
00:12:12,380 --> 00:12:17,090
And one caveat on this rule,
that if there's a route

214
00:12:17,090 --> 00:12:23,430
reflector, the originator ID is
used instead of the router ID.

215
00:12:23,430 --> 00:12:26,000
So keep that caveat
in mind when we're

216
00:12:26,000 --> 00:12:28,530
dealing with route reflectors.

217
00:12:28,530 --> 00:12:31,790
So we just had a caveat
related to route reflectors.

218
00:12:31,790 --> 00:12:35,870
Now we have a whole rule
for route reflectors.

219
00:12:35,870 --> 00:12:38,480
You can see right
here, our first note,

220
00:12:38,480 --> 00:12:43,520
that this is only present in BGP
route reflector environments.

221
00:12:43,520 --> 00:12:45,180
Well, what are we talking about?

222
00:12:45,180 --> 00:12:47,070
Well, it's quite simple.

223
00:12:47,070 --> 00:12:51,250
If that originator ID is the
same for multiple paths--

224
00:12:51,250 --> 00:12:53,750
which, again, that's what this
whole path selection criteria

225
00:12:53,750 --> 00:12:55,220
is for--

226
00:12:55,220 --> 00:13:00,860
we're going to go with the
minimum cluster list length.

227
00:13:00,860 --> 00:13:06,290
Just like the AS path length,
it's the number of clusters

228
00:13:06,290 --> 00:13:09,870
that this particular
path goes through.

229
00:13:09,870 --> 00:13:12,890
So we're going to prefer
to go through the lower

230
00:13:12,890 --> 00:13:15,150
number of route reflectors.

231
00:13:15,150 --> 00:13:17,240
So because of this
rule, we can, in fact,

232
00:13:17,240 --> 00:13:20,810
peer with route reflectors
or clients in other clusters

233
00:13:20,810 --> 00:13:22,740
without creating an issue.

234
00:13:22,740 --> 00:13:26,270
However, be aware that in
order for this to work,

235
00:13:26,270 --> 00:13:31,490
the receiving router has to be
aware of those route reflector

236
00:13:31,490 --> 00:13:34,160
specific BGP attributes.

237
00:13:34,160 --> 00:13:37,640
I can't imagine that any
devices today would not

238
00:13:37,640 --> 00:13:38,970
be aware of that.

239
00:13:38,970 --> 00:13:41,930
However, maybe there's
something out there that's not.

240
00:13:41,930 --> 00:13:43,790
Just be aware that
this rule would not

241
00:13:43,790 --> 00:13:45,890
take effect in that case.

242
00:13:45,890 --> 00:13:52,160
Finally, we prefer the path that
comes from the lowest neighbor

243
00:13:52,160 --> 00:13:53,450
address.

244
00:13:53,450 --> 00:13:56,870
This is going to be using
the IP address that's

245
00:13:56,870 --> 00:14:00,120
used for the BGP neighborship.

246
00:14:00,120 --> 00:14:03,140
So this might be the directly
connected physical interface.

247
00:14:03,140 --> 00:14:05,270
It may be a loopback interface.

248
00:14:05,270 --> 00:14:09,200
It's whatever address is
used in the neighborship.

249
00:14:09,200 --> 00:14:12,560
Again, it's going to
correspond with the remote peer

250
00:14:12,560 --> 00:14:15,720
in that TCP connection.

251
00:14:15,720 --> 00:14:19,910
So as you can see, this is
flat out down to a tiebreaker.

252
00:14:19,910 --> 00:14:22,880
There's no way we're going
to come out of Rule 13

253
00:14:22,880 --> 00:14:25,430
with two best paths.

254
00:14:25,430 --> 00:14:28,990
There will only be one.

255
00:14:28,990 --> 00:14:33,340
So in this video, we went
through the BGP best path

256
00:14:33,340 --> 00:14:35,300
selection criteria.

257
00:14:35,300 --> 00:14:40,100
So when BGP has multiple paths
to get to a single destination,

258
00:14:40,100 --> 00:14:43,240
it will go down this
list, one by one,

259
00:14:43,240 --> 00:14:48,680
all 13 steps, if needed, till
it comes down to one best path.

260
00:14:48,680 --> 00:14:52,480
Now we also saw that
at rule nine, if we're

261
00:14:52,480 --> 00:14:55,630
putting multiple routes
into the routing table,

262
00:14:55,630 --> 00:14:58,390
it can, of course,
support multipathing.

263
00:14:58,390 --> 00:15:00,730
But it will still
continue until it

264
00:15:00,730 --> 00:15:03,550
has one best path,
because that's what it

265
00:15:03,550 --> 00:15:05,980
has to send to its neighbors.

266
00:15:05,980 --> 00:15:08,530
So we'll go down the
list until there's

267
00:15:08,530 --> 00:15:13,700
a single path, eliminating
all the other paths as we go.

268
00:15:13,700 --> 00:15:15,790
I hope this has been
informative for you,

269
00:15:15,790 --> 00:15:18,900
and I'd like to thank
you for viewing.

270
00:15:18,900 --> 00:15:22,000