1
00:00:06,730 --> 00:00:10,470
- Now let's review Amazon
EC2 Billing Options.

2
00:00:10,470 --> 00:00:13,240
Within Amazon EC2 we have a number

3
00:00:13,240 --> 00:00:16,250
of different billing
options to choose from.

4
00:00:16,250 --> 00:00:19,090
In this particular table as we can see

5
00:00:19,090 --> 00:00:21,683
we have the On-Demand billing model.

6
00:00:22,640 --> 00:00:25,050
And On-Demand essentially says that

7
00:00:25,050 --> 00:00:28,290
it follows what we might call the MSRP.

8
00:00:28,290 --> 00:00:32,720
If AWS advertises a
particular type of machine

9
00:00:32,720 --> 00:00:35,590
to cost say 20 cents an hour

10
00:00:35,590 --> 00:00:38,780
then launching it with the
On-Demand billing model

11
00:00:38,780 --> 00:00:41,550
we will pay the 20 cents an hour.

12
00:00:41,550 --> 00:00:45,363
Or per second if we're
using per second billing.

13
00:00:46,200 --> 00:00:50,160
Now if we need a discount,
if we wanted to be able

14
00:00:50,160 --> 00:00:52,550
to run applications and lower our cost

15
00:00:52,550 --> 00:00:54,550
and gain access to discounts.

16
00:00:54,550 --> 00:00:59,400
Then we have the reserved
instance billing model

17
00:01:00,400 --> 00:01:04,350
and we also have Spot
Instance billing model.

18
00:01:04,350 --> 00:01:06,220
And we're gonna talk about those in detail

19
00:01:06,220 --> 00:01:08,250
here in just a moment.

20
00:01:08,250 --> 00:01:11,040
Now the biggest thing that I
wanna draw your attention to

21
00:01:11,040 --> 00:01:13,670
on this table, you can
always pause the video

22
00:01:13,670 --> 00:01:17,170
and review this, but the biggest thing

23
00:01:17,170 --> 00:01:20,600
that I want to draw your
attention to is the Availability.

24
00:01:20,600 --> 00:01:24,590
And I want to highlight the
fact that with On-Demand

25
00:01:26,060 --> 00:01:30,770
and with the Spot Market, Availability

26
00:01:30,770 --> 00:01:32,623
is not guaranteed.

27
00:01:33,620 --> 00:01:38,010
And so AWS, we have to
keep in mind that AWS

28
00:01:38,010 --> 00:01:43,010
does have an unimaginably large capacity

29
00:01:43,110 --> 00:01:44,600
for compute power.

30
00:01:44,600 --> 00:01:48,200
In fact at one time
Gartner reported that AWS

31
00:01:48,200 --> 00:01:52,020
had several times the
amount of compute power

32
00:01:52,020 --> 00:01:55,130
as the other cloud providers combined.

33
00:01:55,130 --> 00:01:58,010
And so I think we would have
a hard time even imagining

34
00:01:58,010 --> 00:02:01,710
how much compute capacity AWS has.

35
00:02:01,710 --> 00:02:05,300
But and it's constantly
growing, AWS is constantly

36
00:02:05,300 --> 00:02:08,850
adding machines to that
and growing that capacity

37
00:02:08,850 --> 00:02:10,633
in any particular region.

38
00:02:11,480 --> 00:02:15,710
But at any one time, at the
time we launched an instance

39
00:02:15,710 --> 00:02:18,420
in that moment there is a finite capacity.

40
00:02:18,420 --> 00:02:22,850
And so it is possible
that using the On-Demand

41
00:02:22,850 --> 00:02:27,170
or the Spot billing models, it is possible

42
00:02:27,170 --> 00:02:31,620
that we could ask for a
particular type of EC2 instance

43
00:02:31,620 --> 00:02:35,370
and then the API come back
and say sorry we don't

44
00:02:35,370 --> 00:02:38,920
have capacity for that type machine

45
00:02:38,920 --> 00:02:42,010
in that availability zone
please try again later

46
00:02:42,010 --> 00:02:44,020
or try a different size machine

47
00:02:44,020 --> 00:02:47,330
or try a different
availability zone right?

48
00:02:47,330 --> 00:02:49,250
So keep that in mind that and of course

49
00:02:49,250 --> 00:02:52,860
I've been using AWS for almost 10 years

50
00:02:52,860 --> 00:02:57,350
and I have only ever seen
that happen maybe four times.

51
00:02:57,350 --> 00:03:00,200
So it is rare but it is possible.

52
00:03:00,200 --> 00:03:04,900
Now with Reserved Instances,
there may or may not be

53
00:03:04,900 --> 00:03:06,470
an upfront payment, it all depends

54
00:03:06,470 --> 00:03:08,670
on how we want to approach it.

55
00:03:08,670 --> 00:03:13,670
Generally the more we pay
upfront the more discount we get.

56
00:03:13,840 --> 00:03:18,160
And so we can choose to
purchase a Reserved Instance

57
00:03:18,160 --> 00:03:20,470
by making no upfront payment.

58
00:03:20,470 --> 00:03:23,750
That means that there's
no capital expenditure

59
00:03:23,750 --> 00:03:25,730
we get to keep more money in our pocket

60
00:03:25,730 --> 00:03:29,160
but we are getting the least
amount of discount right?

61
00:03:29,160 --> 00:03:32,800
So the more we pay upfront
the more discount we get.

62
00:03:32,800 --> 00:03:35,920
We could choose to pay a little
bit, we could choose to pay

63
00:03:35,920 --> 00:03:38,130
make a full upfront payment right?

64
00:03:38,130 --> 00:03:40,310
So by making a full upfront payment

65
00:03:40,310 --> 00:03:43,210
what we are essentially
doing is saying that

66
00:03:43,210 --> 00:03:45,980
for the term they're
available in either a one

67
00:03:45,980 --> 00:03:47,670
or a three-year term.

68
00:03:47,670 --> 00:03:50,470
For that term if I make
a full upfront payment,

69
00:03:50,470 --> 00:03:54,610
I am paying for every hour of every month

70
00:03:54,610 --> 00:03:57,330
for that one or three year term.

71
00:03:57,330 --> 00:04:00,810
And in doing so and making
a full payment upfront

72
00:04:00,810 --> 00:04:05,060
I get the most discount on
that particular instance.

73
00:04:05,060 --> 00:04:08,040
And so the thing to keep in
mind with a Reserved Instance

74
00:04:08,040 --> 00:04:12,030
is that, and that this may
be a little bit challenging

75
00:04:12,030 --> 00:04:14,800
to understand at first
but it will become clearer

76
00:04:14,800 --> 00:04:17,180
as you actually get in
there and start using it.

77
00:04:17,180 --> 00:04:19,870
With Reserved Instances, it's one thing

78
00:04:19,870 --> 00:04:23,230
to purchase a Reserved
Instance and it's another thing

79
00:04:23,230 --> 00:04:25,860
to actually launch an instance right?

80
00:04:25,860 --> 00:04:28,210
Purchasing a Reserved Instance
and launching an instance

81
00:04:28,210 --> 00:04:31,170
are two completely different things.

82
00:04:31,170 --> 00:04:34,010
Just because you purchased
a Reserved Instance

83
00:04:34,010 --> 00:04:36,000
doesn't mean you launched one.

84
00:04:36,000 --> 00:04:38,170
And just because you
launched one doesn't mean

85
00:04:38,170 --> 00:04:39,970
it qualifies for the discount

86
00:04:39,970 --> 00:04:42,160
of the Reserved Instance right?

87
00:04:42,160 --> 00:04:45,510
And so if you were to
purchase a Reserved Instance

88
00:04:45,510 --> 00:04:49,810
with no upfront payment or
a partial upfront payment

89
00:04:49,810 --> 00:04:54,810
then you are essentially
committing to paying every hour

90
00:04:55,110 --> 00:04:58,910
of every month even if
you don't actually launch

91
00:04:58,910 --> 00:05:00,130
an instance right?

92
00:05:00,130 --> 00:05:02,770
You're gonna pay that fee regardless

93
00:05:02,770 --> 00:05:05,200
of how many instances you're running.

94
00:05:05,200 --> 00:05:08,800
And so when you purchase
a Reserved Instance

95
00:05:08,800 --> 00:05:11,400
and then what that
essentially what that does

96
00:05:11,400 --> 00:05:15,110
is create a like a billing
reconciliation model.

97
00:05:15,110 --> 00:05:17,810
It creates a line in the billing that says

98
00:05:17,810 --> 00:05:21,100
you qualify for a
discount on any instances

99
00:05:21,100 --> 00:05:23,000
that you run right?

100
00:05:23,000 --> 00:05:25,770
And then when you launch an instance

101
00:05:25,770 --> 00:05:27,380
they have to match right?

102
00:05:27,380 --> 00:05:30,660
So if you purchase a Reserved
Instance for a particular

103
00:05:32,110 --> 00:05:35,570
type of machine in a
particular availability zone,

104
00:05:35,570 --> 00:05:39,620
if you launch that same
thing that same size machine

105
00:05:39,620 --> 00:05:43,690
in that same AZ the billing
reconciliation will apply

106
00:05:43,690 --> 00:05:45,463
to the instance that you launch.

107
00:05:47,020 --> 00:05:51,693
And so it's possible to gain
a guaranteed availability.

108
00:05:52,580 --> 00:05:54,780
This is one of the
options you don't have to

109
00:05:55,970 --> 00:05:58,440
when you purchase the Reserved Instance

110
00:05:58,440 --> 00:06:02,570
if we choose the
availability zone, then AWS

111
00:06:02,570 --> 00:06:05,280
will reserve availability for us

112
00:06:05,280 --> 00:06:08,140
and that's partly why we
call it a Reserved Instance

113
00:06:08,140 --> 00:06:10,470
because capacity is being reserved.

114
00:06:10,470 --> 00:06:13,450
It means that when we
actually launch the instance

115
00:06:13,450 --> 00:06:15,583
we have a guarantee that we will get one.

116
00:06:17,210 --> 00:06:20,810
And of course the benefit
of Reserved Instances

117
00:06:20,810 --> 00:06:23,770
are that we gain a significant discount

118
00:06:23,770 --> 00:06:26,010
versus the On-Demand price.

119
00:06:26,010 --> 00:06:28,810
So Reserved Instances are generally best

120
00:06:28,810 --> 00:06:33,800
for Constant applications or a Databases.

121
00:06:33,800 --> 00:06:36,280
When you look at the
application that's running

122
00:06:36,280 --> 00:06:38,760
and if you determine
that, that application,

123
00:06:38,760 --> 00:06:43,760
it will be running for more than 55

124
00:06:43,920 --> 00:06:46,000
to 60% of the month.

125
00:06:46,000 --> 00:06:49,800
Then that's generally
where the threshold is.

126
00:06:49,800 --> 00:06:54,130
Beyond 55 to 60% of the
month is when you would see

127
00:06:54,130 --> 00:06:58,260
a savings by using Reserved
Instances as opposed

128
00:06:58,260 --> 00:07:00,230
to using On-Demand.

129
00:07:00,230 --> 00:07:01,840
The other billing model that we have

130
00:07:01,840 --> 00:07:03,593
would be Spot Instances.

131
00:07:04,480 --> 00:07:09,390
Now with the Spot, what we
call the Spot Market, the fee

132
00:07:09,390 --> 00:07:11,600
is based on supply and demand.

133
00:07:11,600 --> 00:07:14,370
I mentioned earlier that AWS has

134
00:07:14,370 --> 00:07:18,600
an enormous unimaginable
capacity for generally

135
00:07:18,600 --> 00:07:22,570
any type of machine in any
given availability zone.

136
00:07:22,570 --> 00:07:27,080
And so what AWS does is
they look at the supply

137
00:07:27,080 --> 00:07:29,620
that they have, the
capacity that they have

138
00:07:29,620 --> 00:07:31,810
for a particular size machine

139
00:07:31,810 --> 00:07:34,990
and the actual demand for that machine.

140
00:07:34,990 --> 00:07:37,880
And the Spot Market is essentially a way

141
00:07:37,880 --> 00:07:41,550
of selling unused capacity

142
00:07:41,550 --> 00:07:43,630
or giving us the ability to gain

143
00:07:45,100 --> 00:07:47,750
those resources at a lower price.

144
00:07:47,750 --> 00:07:52,750
And so we can see with the Spot Market

145
00:07:53,040 --> 00:07:54,500
our hourly fees

146
00:07:54,500 --> 00:07:58,210
be an absolute fraction
of the On-Demand price.

147
00:07:58,210 --> 00:08:01,690
We can save upwards of 80 maybe even 90%

148
00:08:01,690 --> 00:08:03,343
as opposed to On-Demand.

149
00:08:06,390 --> 00:08:09,740
And so the way that we
approach the Spot Market

150
00:08:09,740 --> 00:08:12,230
is that when we launch an
EC2 instance and we say

151
00:08:12,230 --> 00:08:15,480
I want to use a Spot
Instance, keep in mind that

152
00:08:15,480 --> 00:08:16,890
there's no functional difference

153
00:08:16,890 --> 00:08:19,963
between On-Demand, Reserved and Spot.

154
00:08:20,900 --> 00:08:24,170
These are all EC2 instances
there's no functional difference

155
00:08:24,170 --> 00:08:27,480
between them, the only
difference is how we actually pay

156
00:08:27,480 --> 00:08:29,640
for them, how we are charged right?

157
00:08:29,640 --> 00:08:32,370
And so when we launch an EC2 instance

158
00:08:32,370 --> 00:08:36,630
and if we want to gain access
to the Spot Market discounts,

159
00:08:36,630 --> 00:08:40,970
we choose that we want
it as a Spot Instance

160
00:08:40,970 --> 00:08:43,250
and we specify our maximum spend.

161
00:08:43,250 --> 00:08:44,940
How much are we willing to spend

162
00:08:44,940 --> 00:08:46,513
on that particular instance?

163
00:08:47,440 --> 00:08:52,440
Now if the current market rate

164
00:08:52,760 --> 00:08:54,573
according to supply and demand.

165
00:08:55,480 --> 00:08:59,370
If the current market rate rises above

166
00:08:59,370 --> 00:09:03,410
our maximum spend then the
instance can be terminated

167
00:09:03,410 --> 00:09:06,170
by AWS and essentially given to someone

168
00:09:06,170 --> 00:09:08,020
who's willing to spend more money.

169
00:09:08,020 --> 00:09:10,340
Now of course we do get
a two-minute warning

170
00:09:10,340 --> 00:09:12,730
and that's it made
available to the instance

171
00:09:12,730 --> 00:09:14,340
our application can poll

172
00:09:14,340 --> 00:09:17,710
for that inquiry the
instance metadata service

173
00:09:17,710 --> 00:09:20,800
looking for that warning and
when it sees that warning

174
00:09:20,800 --> 00:09:24,780
it can record some kind of
a checkpoint save its work

175
00:09:24,780 --> 00:09:26,060
that way it can come back

176
00:09:26,060 --> 00:09:31,060
and resume that work later
when the market price lowers.

177
00:09:33,100 --> 00:09:38,100
And so the Spot Market is
generally best for Temporary work.

178
00:09:38,810 --> 00:09:42,330
Work that can be interrupted
like Batch Processing,

179
00:09:42,330 --> 00:09:45,640
Testing and Experimentation, Billing

180
00:09:45,640 --> 00:09:47,060
and Sales reports right?

181
00:09:47,060 --> 00:09:51,140
Any kind of work that can
be done quickly and work

182
00:09:51,140 --> 00:09:55,880
that is easy to essentially
checkpoint and save

183
00:09:55,880 --> 00:09:58,010
and come back to you later.

184
00:09:58,010 --> 00:10:00,990
The really important thing
about the Spot Instances

185
00:10:00,990 --> 00:10:02,390
and I think I've already mentioned this,

186
00:10:02,390 --> 00:10:06,520
is that apps need to
tolerate interruption.

187
00:10:06,520 --> 00:10:11,520
If your application cannot
tolerate interruption whatsoever

188
00:10:11,600 --> 00:10:13,980
then the Spot Market is the wrong choice,

189
00:10:13,980 --> 00:10:17,220
we would definitely want
to look to either On-Demand

190
00:10:17,220 --> 00:10:18,440
or Reserved right?

191
00:10:18,440 --> 00:10:21,060
So let's keep those billing models in mind

192
00:10:21,060 --> 00:10:25,410
as we move forward using
EC2 and if we want to create

193
00:10:26,990 --> 00:10:30,020
an environment that is cost-effective.

194
00:10:30,020 --> 00:10:32,330
It's very important that we be aware

195
00:10:32,330 --> 00:10:35,000
of the billing models that are available;

196
00:10:35,000 --> 00:10:38,820
On-Demand, Reserved and
Spot and determine which one

197
00:10:38,820 --> 00:10:41,620
is most appropriate for
that particular type

198
00:10:41,620 --> 00:10:42,870
of work that we're doing.