1
00:00:00,000 --> 00:00:03,880
[AUDIO LOGO]

2
00:00:03,880 --> 00:00:06,790

3
00:00:06,790 --> 00:00:11,350
So now it's time to get a better
look at deterministic MED.

4
00:00:11,350 --> 00:00:14,680
We're going to take a look at
the difference between what

5
00:00:14,680 --> 00:00:20,590
IOS XE does by default versus
what deterministic MED does

6
00:00:20,590 --> 00:00:23,180
to the path selection process.

7
00:00:23,180 --> 00:00:25,780
Now, we'll see, as we go
through this discussion,

8
00:00:25,780 --> 00:00:28,420
that deterministic
MED is actually

9
00:00:28,420 --> 00:00:31,720
enabled by default on IOS XR.

10
00:00:31,720 --> 00:00:34,060
So this is mostly
going to be about what

11
00:00:34,060 --> 00:00:38,710
we have to do to effect
this change on IOS XE.

12
00:00:38,710 --> 00:00:41,680
And it is generally
considered best practice

13
00:00:41,680 --> 00:00:44,170
to enable deterministic MED.

14
00:00:44,170 --> 00:00:47,200
So let's jump in and take a
look at the details of how this

15
00:00:47,200 --> 00:00:50,810
affects the overall process.

16
00:00:50,810 --> 00:00:53,800
So our first example
here is where

17
00:00:53,800 --> 00:00:58,070
we're ordered from the newest
to the oldest, in other words,

18
00:00:58,070 --> 00:01:00,730
just the order in which
the routes were learned.

19
00:01:00,730 --> 00:01:04,870
And this is the
default on IOS XE.

20
00:01:04,870 --> 00:01:09,530
Now, what's the problem with
this, or why do we care?

21
00:01:09,530 --> 00:01:13,570
Well, the way the IOS
picks the best path is it

22
00:01:13,570 --> 00:01:18,460
takes the first two entries
and compares them together.

23
00:01:18,460 --> 00:01:20,230
Here's the problem.

24
00:01:20,230 --> 00:01:23,290
This one is from AS 500.

25
00:01:23,290 --> 00:01:26,830
This one is from AS 100.

26
00:01:26,830 --> 00:01:29,510
And since they're from
different AS numbers,

27
00:01:29,510 --> 00:01:33,880
that means that the MED
doesn't get considered.

28
00:01:33,880 --> 00:01:37,870
Now, moving on,
they're both external.

29
00:01:37,870 --> 00:01:42,730
This decision is going to come
down to the lowest router ID.

30
00:01:42,730 --> 00:01:46,180
Well, the lowest
router ID is 1.1.1.1.

31
00:01:46,180 --> 00:01:53,390
And this becomes the
best of those two paths.

32
00:01:53,390 --> 00:01:55,810
So now, entry2 is our best.

33
00:01:55,810 --> 00:02:00,400
Now we compare entry2 to entry3.

34
00:02:00,400 --> 00:02:06,841
Once again, AS 100, AS
500, we don't look at MED.

35
00:02:06,841 --> 00:02:08,960
MED is not considered.

36
00:02:08,960 --> 00:02:11,830
So now, as we go down
the BGP path selection,

37
00:02:11,830 --> 00:02:15,130
we're going to prefer
external over internal.

38
00:02:15,130 --> 00:02:18,440
And this one is going
to become the best.

39
00:02:18,440 --> 00:02:22,810
And as we can see here,
entry2 is our best entry

40
00:02:22,810 --> 00:02:25,300
in this scenario.

41
00:02:25,300 --> 00:02:28,060
I think the biggest
thing to realize here,

42
00:02:28,060 --> 00:02:30,640
whether deterministic
MED changes

43
00:02:30,640 --> 00:02:33,550
the actual ultimate
decision or not,

44
00:02:33,550 --> 00:02:37,820
MED wasn't even considered
in this discussion.

45
00:02:37,820 --> 00:02:41,590
We just chose the best
path, and we crossed out

46
00:02:41,590 --> 00:02:43,990
MED the whole way through.

47
00:02:43,990 --> 00:02:47,890
MED didn't even get a chance
to affect the decision.

48
00:02:47,890 --> 00:02:53,560
That's not very deterministic if
MED doesn't even get a chance.

49
00:02:53,560 --> 00:02:56,710
Again, will it ultimately
change the final decision?

50
00:02:56,710 --> 00:02:58,480
Maybe, maybe not.

51
00:02:58,480 --> 00:03:02,320
But it doesn't even get
considered in, we'll say,

52
00:03:02,320 --> 00:03:05,500
just the wrong
scenario like this one.

53
00:03:05,500 --> 00:03:07,480
MED wasn't even looked at.

54
00:03:07,480 --> 00:03:10,870
Let's look at that same
scenario now but now

55
00:03:10,870 --> 00:03:14,010
with deterministic MED.

56
00:03:14,010 --> 00:03:18,120
So now, we're taking
those same prefixes,

57
00:03:18,120 --> 00:03:22,080
but this time, they're
ordered by the AS number.

58
00:03:22,080 --> 00:03:25,540
And this is using the
deterministic MED feature,

59
00:03:25,540 --> 00:03:30,570
which is actually the
default on IOS XR.

60
00:03:30,570 --> 00:03:32,280
So what's changed?

61
00:03:32,280 --> 00:03:35,820
Well, we can actually see
here that they are now

62
00:03:35,820 --> 00:03:40,470
grouped by the AS number,
not the order in which they

63
00:03:40,470 --> 00:03:41,830
were learned.

64
00:03:41,830 --> 00:03:47,410
Remember that before, router 1
was entry2 up here at the top.

65
00:03:47,410 --> 00:03:50,940
Now, router 1 is entry1.

66
00:03:50,940 --> 00:03:59,080
So now what happens is we pick
the best by the AS number.

67
00:03:59,080 --> 00:04:00,750
In other words, we're
going to start off

68
00:04:00,750 --> 00:04:03,720
by looking at these two
entries because they're

69
00:04:03,720 --> 00:04:04,923
from the same AS.

70
00:04:04,923 --> 00:04:07,340
And what we're going to do is
we're going to go down here,

71
00:04:07,340 --> 00:04:08,715
and we're going
to go OK, they're

72
00:04:08,715 --> 00:04:11,760
both from a single AS, AS 500.

73
00:04:11,760 --> 00:04:14,610
This one has a MED of 100.

74
00:04:14,610 --> 00:04:17,579
This one has a MED of 150.

75
00:04:17,579 --> 00:04:26,040
Lowest wins, so entry2 becomes
the best, the best from AS 500.

76
00:04:26,040 --> 00:04:29,880
But then we clear that
out, and we say, OK,

77
00:04:29,880 --> 00:04:36,640
now take the best from AS
500 and compare it to AS 100.

78
00:04:36,640 --> 00:04:40,260
OK, well, again, they're
both one AS number.

79
00:04:40,260 --> 00:04:44,010
In this case, they're from
different AS's, so once again,

80
00:04:44,010 --> 00:04:46,120
MED gets thrown out.

81
00:04:46,120 --> 00:04:50,880
We prefer external over
internal, so guess what?

82
00:04:50,880 --> 00:04:54,710
Entry1 becomes the best.

83
00:04:54,710 --> 00:04:58,010
Now, you might note that
it's still router 1.

84
00:04:58,010 --> 00:05:00,890
As I mentioned,
deterministic MED

85
00:05:00,890 --> 00:05:03,780
may not affect the
ultimate end decision.

86
00:05:03,780 --> 00:05:05,780
And in this case, it didn't.

87
00:05:05,780 --> 00:05:07,700
But you see the
important factor here

88
00:05:07,700 --> 00:05:10,670
is that MED at least
got considered.

89
00:05:10,670 --> 00:05:13,190
It may not have been the
final decision-maker,

90
00:05:13,190 --> 00:05:15,500
but it was at least
considered when

91
00:05:15,500 --> 00:05:17,970
making our best path selection.

92
00:05:17,970 --> 00:05:20,300
So let's clear this
off, and let's take

93
00:05:20,300 --> 00:05:22,820
a look at this same scenario.

94
00:05:22,820 --> 00:05:25,380
But what if we turn on both?

95
00:05:25,380 --> 00:05:29,930
So what if we do deterministic
MED and always compare MED?

96
00:05:29,930 --> 00:05:32,160
Let's see what that changes.

97
00:05:32,160 --> 00:05:34,460
So once again, we're
going to start off

98
00:05:34,460 --> 00:05:37,760
by looking at entry2 and
entry3 together because they're

99
00:05:37,760 --> 00:05:39,440
from the same AS.

100
00:05:39,440 --> 00:05:42,950
And once again,
they're both from 500.

101
00:05:42,950 --> 00:05:44,540
This one has a MED of 100.

102
00:05:44,540 --> 00:05:46,160
This has a MED of 150.

103
00:05:46,160 --> 00:05:50,270
Therefore, once
again, entry2 wins.

104
00:05:50,270 --> 00:05:53,840
So so far, entry2 is our
winner out of this pairing.

105
00:05:53,840 --> 00:05:56,120
So let's get rid of these.

106
00:05:56,120 --> 00:06:00,870
And now we compare
entry2 and entry1.

107
00:06:00,870 --> 00:06:02,510
This one's from a single AS.

108
00:06:02,510 --> 00:06:03,800
This is from a single AS.

109
00:06:03,800 --> 00:06:06,920
So the AS path
length is the same.

110
00:06:06,920 --> 00:06:11,670
But now, remember, we have BGP
always compare MED enabled.

111
00:06:11,670 --> 00:06:15,800
So now it is going to look
at the MED of 200 and 100.

112
00:06:15,800 --> 00:06:22,740
And again, 100 is lower, so this
now becomes the overall best.

113
00:06:22,740 --> 00:06:25,500
So using these two
features together,

114
00:06:25,500 --> 00:06:27,710
we end up with a
different answer.

115
00:06:27,710 --> 00:06:30,980
Now, entry2 is the best.

116
00:06:30,980 --> 00:06:33,180
Now, using these features
together, of course,

117
00:06:33,180 --> 00:06:36,500
we still have to make sure
that those neighboring AS's,

118
00:06:36,500 --> 00:06:41,630
100 and 500 in this case,
are using compatible MEDs.

119
00:06:41,630 --> 00:06:44,100
But you can see that
at bare minimum,

120
00:06:44,100 --> 00:06:47,190
we should turn on
deterministic MED.

121
00:06:47,190 --> 00:06:49,980
Deterministic MED
is, at this point,

122
00:06:49,980 --> 00:06:53,340
considered the best choice
for new installations.

123
00:06:53,340 --> 00:06:57,570
This is, again, part of why
it's on by default on IOS XR.

124
00:06:57,570 --> 00:07:01,900
You should probably enable
it on IOS XE as well.

125
00:07:01,900 --> 00:07:04,110
Now, do you always compare MED?

126
00:07:04,110 --> 00:07:08,950
That one, of course,
depends on your scenario.

127
00:07:08,950 --> 00:07:13,270
In this video, we saw how
using deterministic MED

128
00:07:13,270 --> 00:07:18,400
actually ensures that MED is
always at least considered

129
00:07:18,400 --> 00:07:20,930
if we don't use this feature.

130
00:07:20,930 --> 00:07:23,560
Then we saw, in
our first example,

131
00:07:23,560 --> 00:07:26,770
that depending on the exact
order that the routes are

132
00:07:26,770 --> 00:07:30,800
learned in, MED might
not be considered at all.

133
00:07:30,800 --> 00:07:34,420
Now, again, MED might not
be making the final decision

134
00:07:34,420 --> 00:07:37,210
as far as the best
path, but at least it

135
00:07:37,210 --> 00:07:40,390
gets to contribute or
at least be involved

136
00:07:40,390 --> 00:07:42,700
if we use deterministic MED.

137
00:07:42,700 --> 00:07:45,160
And again, that is
generally considered

138
00:07:45,160 --> 00:07:47,380
to be the best
practice, and that's

139
00:07:47,380 --> 00:07:51,850
part of why it's enabled
by default on IOS XR.

140
00:07:51,850 --> 00:07:54,010
I hope this has been
informative for you,

141
00:07:54,010 --> 00:07:57,150
and I'd like to thank
you for viewing.

142
00:07:57,150 --> 00:08:00,000