1

00:00:12,620  -->  00:00:18,060
Welcome back to Backspace Academy. In
this lecture on Elastic Compute cloud or

2

00:00:18,060  -->  00:00:22,350
ec2 for short, we're going to start off
by looking at the different ways that

3

00:00:22,350  -->  00:00:28,080
you can purchase an ec2 server and also
look at the different instance types are

4

00:00:28,080  -->  00:00:32,219
available and for the different
applications you may use those. Then

5

00:00:32,219  -->  00:00:36,780
we'll look at the instance lifecycle
what happens to an instance from when it

6

00:00:36,780  -->  00:00:42,030
is launched through that process through
to when it is actually terminated. We'll

7

00:00:42,030  -->  00:00:48,020
look at cluster networking and how we
can group clusters of ec2 servers and

8

00:00:48,020  -->  00:00:52,649
interconnect them with high-speed
networking. We'll look at the storage

9

00:00:52,649  -->  00:00:58,320
options that are available to attach to
our EC2 instances and finally we'll look

10

00:00:58,320  -->  00:01:03,539
at the options that are available for us
to connect remotely in to our ec2 server

11

00:01:03,539  -->  00:01:11,070
operating system. There are a number of
different options that are available for

12

00:01:11,070  -->  00:01:15,990
us when we're purchasing ec2 servers. Up
till now we've been using on-demand

13

00:01:15,990  -->  00:01:19,020
instances and they're the most easy to
understand.

14

00:01:19,020  -->  00:01:23,759
We simply launch an instance and from
that point onwards we're paying by the

15

00:01:23,759  -->  00:01:29,159
second at a flat rate for that instance.
There's no upfront cost for launching it

16

00:01:29,159  -->  00:01:33,990
and there's no cost when we terminate it.
We just pay by the second for what we use.

17

00:01:33,990  -->  00:01:41,460
Now for AWS to have capacity
available for these on-demand instances

18

00:01:41,460  -->  00:01:48,090
AWS needs to have at a certain
level of reserve capacity available to

19

00:01:48,090  -->  00:01:51,780
fulfill that. So if all of a sudden
everyone just comes along and wants to

20

00:01:51,780  -->  00:01:56,280
start launching ec2 instances, there
needs to be a little bit of reserve

21

00:01:56,280  -->  00:02:02,670
capacity available. Now instead of having
that capacity just sitting there doing

22

00:02:02,670  -->  00:02:06,030
nothing
what AWS does is that they have

23

00:02:06,030  -->  00:02:11,970
that capacity available for people to
bid on and so whoever bids with the

24

00:02:11,970  -->  00:02:16,800
highest price
get that capacity and it's a little bit

25

00:02:16,800  -->  00:02:22,950
like eBay for reserve capacity of ec2
and that helps you to significantly

26

00:02:22,950  -->  00:02:28,130
reduce the cost of running instances and
it's generally the cheapest option

27

00:02:28,130  -->  00:02:35,640
although not always. The reason is, if all
of that capacity has been taken and the

28

00:02:35,640  -->  00:02:40,770
spot price goes through the roof and
your maximum spot price is above an

29

00:02:40,770  -->  00:02:48,120
on-demand instance then you're going to
be paying a premium for that instance.

30

00:02:48,120  -->  00:02:52,380
So you're really putting down there a
maximum price that you are willing to

31

00:02:52,380  -->  00:02:59,490
pay. The downside is that AWS they can
take that capacity back whenever they

32

00:02:59,490  -->  00:03:04,710
want and that's called a spot instance
interruption and they can also take it

33

00:03:04,710  -->  00:03:11,670
back when the spot price for that next
hour exceeds your maximum price. So that

34

00:03:11,670  -->  00:03:17,490
is something that you need to take into
consideration. It's really best for batch

35

00:03:17,490  -->  00:03:22,590
type compute processes, for example if
you had big data stuff that need to be

36

00:03:22,590  -->  00:03:27,180
processed and once it's processed you
don't need that capacity anymore. A spot

37

00:03:27,180  -->  00:03:32,790
instance is great for that. Now if your
spot instance is terminated or stopped

38

00:03:32,790  -->  00:03:39,570
by AWS within that first instance hour
you won't be charged anything for that usage.

39

00:03:39,570  -->  00:03:45,090
If you terminate that instance
within the first error or any other hour

40

00:03:45,090  -->  00:03:51,450
after that you're going to be charged
per second or the per nearest second for

41

00:03:51,450  -->  00:03:59,340
the use of that instance. So as we
mentioned before there is a significant

42

00:03:59,340  -->  00:04:04,980
amount of reserve capacity available for
these on-demand instances and for the

43

00:04:04,980  -->  00:04:12,450
variation in usage of those instances.
Now there is also a very unlikely

44

00:04:12,450  -->  00:04:17,310
possibility that that reserved capacity
is actually used up as well by all the

45

00:04:17,310  -->  00:04:21,900
on-demand instances. So there are a
number of other options that are

46

00:04:21,900  -->  00:04:27,500
available for you to reserve capacity
outside of on-demand.

47

00:04:27,500  -->  00:04:34,500
The first option there is a reserved
instance. Now what that means is that if

48

00:04:34,500  -->  00:04:40,470
you find that you are going to be using
a certain level of capacity of EC2 over

49

00:04:40,470  -->  00:04:45,840
the next year or two years or up to
three years, you can go to AWS and say

50

00:04:45,840  -->  00:04:50,460
right I would like to have this capacity
available for me for that period of time

51

00:04:50,460  -->  00:04:58,380
and you can use that capacity and, by
making that commitment to AWS over that

52

00:04:58,380  -->  00:05:05,300
period of time you are going to get a
discount on the cost of your instances.

53

00:05:05,300  -->  00:05:10,410
A scheduled instance is where you can
purchase instances but instead of being

54

00:05:10,410  -->  00:05:16,370
over a one year period continuously you
can have it over a one year period

55

00:05:16,370  -->  00:05:22,020
intermittently on a schedule, for example
you might use it for one day in a month

56

00:05:22,020  -->  00:05:27,180
for, I don't know, accounts processing or
something like that and so that will

57

00:05:27,180  -->  00:05:32,790
give you that reserved capacity that you
need for that specific time of every

58

00:05:32,790  -->  00:05:39,440
month or week or whatever. On demand
capacity reservations are different to

59

00:05:39,440  -->  00:05:45,270
reserved and scheduled instances in that
you're not entering into a long term

60

00:05:45,270  -->  00:05:50,610
contract. These reservations are on
demand you use them when you need them

61

00:05:50,610  -->  00:05:55,830
or you create a capacity reservation
when you need it, when you don't need it

62

00:05:55,830  -->  00:06:00,120
you just stop it
and there's no long term contract.

63

00:06:00,120  -->  00:06:05,640
Now the downside is that, it is at the
on-demand rate. So you might have a

64

00:06:05,640  -->  00:06:11,970
capacity reservation that you never use
but you will still be paying for the

65

00:06:11,970  -->  00:06:17,370
cost of an on-demand instance. So that
needs to be taken into consideration

66

00:06:17,370  -->  00:06:23,640
because there is no cost benefit with a
on demand capacity reservation but it

67

00:06:23,640  -->  00:06:30,180
does reserve capacity for you in the
event that something goes wrong with AWS

68

00:06:30,180  -->  00:06:35,070
and they run out of on-demand instances
that are available. You will always have

69

00:06:35,070  -->  00:06:40,470
an on-demand instance of available for you,
and finally we have

70

00:06:40,470  -->  00:06:46,320
dedicated instances and dedicated hosts.
With a dedicated instance, you're going

71

00:06:46,320  -->  00:06:51,540
to be paying by the hour for an instance
that will be running on single tenant

72

00:06:51,540  -->  00:06:56,310
hardware. They will be hardware that only
you will be using and we also have

73

00:06:56,310  -->  00:07:01,290
dedicated hosts where you're going to be
paying for a physical host that is

74

00:07:01,290  -->  00:07:11,070
fully dedicated to running your
dedicated instances. Another way that you

75

00:07:11,070  -->  00:07:18,120
can reduce your ec2 costs is through a
savings plan. That allows for reduced

76

00:07:18,120  -->  00:07:24,270
pricing in exchange for consistent usage
in dollars per hour over a one or three

77

00:07:24,270  -->  00:07:30,360
year commitment. The difference between
this and a reserved instance is that

78

00:07:30,360  -->  00:07:36,150
this is based on your historical usage
in the AWS Cost Explorer and those

79

00:07:36,150  -->  00:07:41,370
recommendations within the AWS Cost
Explorer will be used as a basis for

80

00:07:41,370  -->  00:07:47,490
that saving plan. There are two types of
savings plan you can have. A compute

81

00:07:47,490  -->  00:07:52,110
savings plan, that will be the most
flexible and that will apply to all the

82

00:07:52,110  -->  00:07:57,030
different instance types, the size of the
instance, the operating system or any

83

00:07:57,030  -->  00:08:03,390
tenancy type, and that is going to
provide up to 66% off your normal

84

00:08:03,390  -->  00:08:10,530
on-demand rates. The other option there
is an ec2 instance savings plan and that

85

00:08:10,530  -->  00:08:16,320
is applied to an individual instance
family, for example if you select an m2

86

00:08:16,320  -->  00:08:21,810
type, and that is applied to a specific
region as well and that can provide up

87

00:08:21,810  -->  00:08:28,950
to 72 percent off of the on demand rates.
You can also reduce cost depending on

88

00:08:28,950  -->  00:08:33,750
what payment option you come up with. So
if you decide to have no upfront payment

89

00:08:33,750  -->  00:08:38,430
and just simply pay on a monthly basis
that is one option. If you want to reduce

90

00:08:38,430  -->  00:08:44,130
it from that you can go to a partial
upfront cost and that is going to reduce

91

00:08:44,130  -->  00:08:49,620
a price over a no upfront, or you can do
everything upfront in one payment and

92

00:08:49,620  -->  00:08:54,270
that will provide you with the lowest
price of your savings plan.

93

00:08:54,270  -->  00:09:00,300
Another thing to take into consideration
is that a savings plan does not provide

94

00:09:00,300  -->  00:09:05,490
capacity reservations, but that said
there is nothing stopping you from

95

00:09:05,490  -->  00:09:11,610
including your capacity reserved
instances into a savings plan, but the

96

00:09:11,610  -->  00:09:19,110
savings plan itself does not provide
that capacity reservation. To prepare

97

00:09:19,110  -->  00:09:23,280
for the cloud practitioner exam you will
need to understand these different

98

00:09:23,280  -->  00:09:27,320
reserving options that are available, but
it can be a little bit confusing.

99

00:09:27,320  -->  00:09:32,400
So having a look here, I've got the capacity
reservations and the two different types

100

00:09:32,400  -->  00:09:37,170
of reserved instances. So we have
reserved instances that are reserved

101

00:09:37,170  -->  00:09:41,310
based upon an availability zones they're
the zonal reserved instances we have

102

00:09:41,310  -->  00:09:47,160
regional reserved instances that are
reserved within a specific region and

103

00:09:47,160  -->  00:09:53,010
then finally we've had our Savings Plan.
The first point there is the term with a

104

00:09:53,010  -->  00:09:57,870
capacity reservation there is no
long-term commitment required, whereas

105

00:09:57,870  -->  00:10:03,060
with the others there is a required
either a fixed one year or a three-year

106

00:10:03,060  -->  00:10:10,110
commitment, with capacity benefits the
capacity reservations and the zonal

107

00:10:10,110  -->  00:10:17,820
reserved instances they reserve capacity
within a specific availability zone but

108

00:10:17,820  -->  00:10:23,520
if you're going for a regional reserved
instance not a zonal reserved instance

109

00:10:23,520  -->  00:10:28,140
that will not provide capacity in that
availability zone. So although you're

110

00:10:28,140  -->  00:10:33,270
going to get the cost benefits that
capacity will not be reserved. The same

111

00:10:33,270  -->  00:10:38,670
again with a savings plan. That does not
provide a capacity reservation but with

112

00:10:38,670  -->  00:10:43,890
a savings plan you can always include as
part of that savings plan a capacity

113

00:10:43,890  -->  00:10:49,530
reservation as well. Capacity
reservations do not provide for a

114

00:10:49,530  -->  00:10:55,170
billing and discount but that said you
can use capacity reservations with a

115

00:10:55,170  -->  00:11:01,530
savings plan or as part of a regional
reserved instance plan as well. So both

116

00:11:01,530  -->  00:11:07,020
the zonal reserved instance and the
regional reserved instances and the

117

00:11:07,020  -->  00:11:12,110
savings plan all the
those provide for a billing discount, and

118

00:11:12,110  -->  00:11:16,290
finally there are different instance
limits that are applied to these

119

00:11:16,290  -->  00:11:20,940
different plans as well. So if you're
going for a capacity reservation that is

120

00:11:20,940  -->  00:11:25,740
simply going to be limited by your ec2
on-demand instance limits which are

121

00:11:25,740  -->  00:11:30,300
applied per region. If you're going for a
zonal reserved instance that's going to

122

00:11:30,300  -->  00:11:35,790
be limited to 20 per availability zone.
If it's a regional reserved instance

123

00:11:35,790  -->  00:11:40,740
that will be limited to 20 per region
but both of those if you find that you

124

00:11:40,740  -->  00:11:45,540
need to exceed that, you need more than
20, you can put in an application to AWS

125

00:11:45,540  -->  00:11:51,750
support to have that limit increase
effected for you, and with a savings plan

126

00:11:51,750  -->  00:11:56,310
there is no limit on the amount of
instances that can be applied to that

127

00:11:56,310  -->  00:12:02,790
savings plan. There are a number of
different instance types that are

128

00:12:02,790  -->  00:12:07,290
available for us to select with ec2. The
first one and the one that we've been

129

00:12:07,290  -->  00:12:13,080
using so far is the general-purpose
types they're great for as a name

130

00:12:13,080  -->  00:12:18,720
suggests for general-purpose databases
data processing tasks, back-end servers

131

00:12:18,720  -->  00:12:23,310
and the like. The next one there is
compute optimized. It's going to have

132

00:12:23,310  -->  00:12:28,230
a lot of compute capacity available and
that's going to be good for web servers,

133

00:12:28,230  -->  00:12:33,690
for batch processing and the like. Memory
optimize is great for high performance

134

00:12:33,690  -->  00:12:40,230
databases and for in-memory cases where
we need to respond very quickly to

135

00:12:40,230  -->  00:12:45,330
requests instead of having to go back to
a hard disk to retrieve that that

136

00:12:45,330  -->  00:12:50,190
information, and we've got GPU or
graphics processing units and

137

00:12:50,190  -->  00:12:55,710
accelerator computing instant types and
they're great for 3d applications, for

138

00:12:55,710  -->  00:13:01,550
video encoding, they're great for machine
learning and AI applications as well, and

139

00:13:01,550  -->  00:13:06,840
finally we've got storaged optimized
which is going to be great for your Big

140

00:13:06,840  -->  00:13:12,420
Data stuff. It's going to be great for no
SQL databases, for MongoDB, it's going

141

00:13:12,420  -->  00:13:17,250
to be great for elastic MapReduce, for
data warehousing anything that requires

142

00:13:17,250  -->  00:13:21,990
a large amount of storage. Now with all
these ec2 instance types

143

00:13:21,990  -->  00:13:27,330
we have a choice of Linux and a
whole heap of varieties of Linux or

144

00:13:27,330  -->  00:13:31,680
Windows Server and if you want more
information on those types there is a

145

00:13:31,680  -->  00:13:39,450
link there to the ec2 instance types
page. The instance types that we've

146

00:13:39,450  -->  00:13:45,959
looked at previously are running under a
virtualized environment. So what that

147

00:13:45,959  -->  00:13:52,470
means is that the ec2 service has
physical hardware and on top of that

148

00:13:52,470  -->  00:13:59,640
hardware is running a hypervisor and on
top of that hypervisor are virtualized

149

00:13:59,640  -->  00:14:06,360
servers running on that, and so a Linux
operating system is not actually running

150

00:14:06,360  -->  00:14:13,020
on physical Hardware it's running on a
hypervisor a virtual environment and

151

00:14:13,020  -->  00:14:19,649
that goes the same whether it's Windows
or Linux under EC2. EC2 bare metal

152

00:14:19,649  -->  00:14:24,899
instances are running on a non
virtualized environment and the

153

00:14:24,899  -->  00:14:33,270
operating system is running directly on
that ec2 physical Hardware. It's

154

00:14:33,270  -->  00:14:40,050
particularly suitable for applications
that benefit from deep analysis of the

155

00:14:40,050  -->  00:14:46,170
performance of the underlying hardware
or specialized workloads that require

156

00:14:46,170  -->  00:14:53,430
direct access to that physical hardware
as well. There could be legacy workloads

157

00:14:53,430  -->  00:14:57,690
that are just not running on a
virtualized environment and you need to

158

00:14:57,690  -->  00:15:03,320
have the actual access to a physical
environment to run it. There may also be

159

00:15:03,320  -->  00:15:10,230
licensing restricted applications that
rely upon that physical hardware for

160

00:15:10,230  -->  00:15:18,120
license validation. Examples of bare
metal instances are the i3, m5 metal, r5

161

00:15:18,120  -->  00:15:23,660
metal and z1 D metal instance types.

162

00:15:24,260  -->  00:15:30,960
The t2 burstable performance instances
now they don't blow up on you they don't

163

00:15:30,960  -->  00:15:35,940
explode or burst but they're a very
important instance that you need to

164

00:15:35,940  -->  00:15:39,810
understand.
So what burstable means is that this

165

00:15:39,810  -->  00:15:46,050
instance will operate at what we define
as a baseline performance but it will

166

00:15:46,050  -->  00:15:52,260
have an ability to burst above that for
a short period of time if need be, and

167

00:15:52,260  -->  00:15:58,410
that will be governed by what we call
CPU credits. So the more CPU credits that

168

00:15:58,410  -->  00:16:02,730
we have available the longer that we can
burst and for the most amount that we

169

00:16:02,730  -->  00:16:07,380
can burst above our baseline performance.
So that's great

170

00:16:07,380  -->  00:16:12,120
if we have demand that is not
predictable, it's a bit all over the place,

171

00:16:12,120  -->  00:16:19,040
but we kind of know where our
average for that 24-hour period is.

172

00:16:19,040  -->  00:16:24,570
T2 instances are great when used with an
auto scaling group because when an

173

00:16:24,570  -->  00:16:31,260
instance is launched it may take 15, 20,
25 minutes or whatever to go through

174

00:16:31,260  -->  00:16:37,680
that whole launch process that may be
required. So that may be too long and by

175

00:16:37,680  -->  00:16:45,120
that period of time that burst of
demand has disappeared. So the t2 burstable

176

00:16:45,120  -->  00:16:49,950
performance instance will fill
that void and provide instantaneous

177

00:16:49,950  -->  00:16:56,460
response to that increased demand.
Credits they're built up over a 24-hour

178

00:16:56,460  -->  00:17:03,030
period. So if you stay below that
baseline you're going to build up

179

00:17:03,030  -->  00:17:08,010
credits for every hour that you are
under that baseline and then you can use

180

00:17:08,010  -->  00:17:13,770
those credits where needed to burst
above that baseline capacity, and after

181

00:17:13,770  -->  00:17:20,490
the 24-hour period that will be reset.
Now if you're finding that you're never

182

00:17:20,490  -->  00:17:26,520
having enough CPU credit balance to
burst, that means that the average demand

183

00:17:26,520  -->  00:17:30,440
that you're getting over that period of
time, that 24 hour period of time,

184

00:17:30,440  -->  00:17:35,730
is above what your baseline capacity that
you're paying for is. So you need to

185

00:17:35,730  -->  00:17:39,720
consider upgrading
to a larger instance to keep

186

00:17:39,720  -->  00:17:45,390
below that baseline capacity. Now there's
another option there called the t2

187

00:17:45,390  -->  00:17:52,140
unlimited and that will provide that
burstable high CPU availability and a

188

00:17:52,140  -->  00:18:01,290
flat additional rate of five cents per Vcpu hour, and that is for as long as you

189

00:18:01,290  -->  00:18:06,330
need it. So instead of building up CPU
credits, you just get billed for what you use.

190

00:18:06,330  -->  00:18:14,430
Graviton instance types they are
based upon a 64-bit ARM Neoverse core.

191

00:18:14,430  -->  00:18:19,500
If you've used a Raspberry Pi before you
have obviously taken advantage of an ARM

192

00:18:19,500  -->  00:18:26,130
Cortex core that provides very low cost
high performance but this is a ARM

193

00:18:26,130  -->  00:18:31,950
Neoverse cores and are optimized for
large infrastructures such as cloud

194

00:18:31,950  -->  00:18:37,230
infrastructure, and they will provide a
very low cost and high performance

195

00:18:37,230  -->  00:18:42,000
compute capacity. So if you have
something that is really compute

196

00:18:42,000  -->  00:18:47,850
intensive and it will run on an ARM core
then by all means use that and it will

197

00:18:47,850  -->  00:18:53,340
provide a fantastic performance for you
at a much more reasonable cost.

198

00:18:53,340  -->  00:18:57,450
The first-generation are called the a1
instance type and the second generation

199

00:18:57,450  -->  00:19:04,470
are the graviton2 processors and they
can provide up to 40% better price

200

00:19:04,470  -->  00:19:12,120
performance over the comparable current
generation x86 based instances.

201

00:19:12,120  -->  00:19:17,520
Graviton instances are perfect for scale out
applications that can run on multiple

202

00:19:17,520  -->  00:19:23,310
small cores, but you need to take into
consideration that not all software will

203

00:19:23,310  -->  00:19:28,920
run on an ARM core, but if it does, this
is a great option for compute intensive

204

00:19:28,920  -->  00:19:35,640
applications.
If you find that you want to have a mix

205

00:19:35,640  -->  00:19:40,470
of different instance types and you
might have a mix of on-demand and spot

206

00:19:40,470  -->  00:19:47,820
instance options available you can
create an ec2 fleet of instances, and you

207

00:19:47,820  -->  00:19:54,150
can create an unlimited number of
instance types per ec2 fleet. So you can

208

00:19:54,150  -->  00:20:00,120
mix and match. You can put a t2 in there
with a m1 whatever and you have the

209

00:20:00,120  -->  00:20:04,980
choice of mixing both on-demand
instances and spot instances. So you can

210

00:20:04,980  -->  00:20:10,350
have a baseline capacity there but you
can also have also spot instances

211

00:20:10,350  -->  00:20:16,800
available in that ec2 fleet as well. So
it's only available through using the

212

00:20:16,800  -->  00:20:22,020
AWS API or one of the software
development kits or using the

213

00:20:22,020  -->  00:20:26,580
command-line interface. You cannot do
this through the console and it's

214

00:20:26,580  -->  00:20:32,520
available across multiple availability
zones but not across multiple regions. So

215

00:20:32,520  -->  00:20:38,040
you need to create an ec2 fleet in each
region if you want to have that

216

00:20:38,040  -->  00:20:44,630
happening in each region, and there's no
additional charge for using EC2 fleet.

217

00:20:47,399  -->  00:20:53,769
We already know what an Amazon machine
image is or an AMI we have used an AMI

218

00:20:53,769  -->  00:20:59,109
to launch a wordpress server and we
know that it provides the information

219

00:20:59,109  -->  00:21:05,229
that is required to launch that instance,
and that includes a template for the

220

00:21:05,229  -->  00:21:09,399
root volume of that instance, for
example the operating system will have

221

00:21:09,399  -->  00:21:15,129
the Linux operating system, it will have
the WordPress application there that

222

00:21:15,129  -->  00:21:19,629
will be required, that goes on that root
volume. It will also have launch

223

00:21:19,629  -->  00:21:25,629
permissions for the AMI. So if it's a
public AMI it will have availability to

224

00:21:25,629  -->  00:21:31,570
all AWS accounts. If it's a private one
it will be private to that AWS account

225

00:21:31,570  -->  00:21:37,419
or certain AWS accounts. It will also
have a block device mapping and that

226

00:21:37,419  -->  00:21:42,849
will specify any volumes that are
attached to that instance when it's

227

00:21:42,849  -->  00:21:49,570
launched. We have the AWS marketplace
where people can put AMIs on there and

228

00:21:49,570  -->  00:21:54,190
sell them or give them away for free. So
they can be paid, they can be free, they

229

00:21:54,190  -->  00:22:00,789
can be trial versions of AMI or of
software or they could be BYO license

230

00:22:00,789  -->  00:22:04,239
software, for example it could be an
Oracle database saying you bring your

231

00:22:04,239  -->  00:22:09,249
own license to use it, and so you can
normally find a great deal of

232

00:22:09,249  -->  00:22:13,599
applications on that AWS marketplace and
we've already used that before because

233

00:22:13,599  -->  00:22:20,679
we've used it to launch a WordPress
application. There are a number of

234

00:22:20,679  -->  00:22:28,479
different states of an ec2 instance
depending on whether the root volume of

235

00:22:28,479  -->  00:22:35,639
the instance is an EBS or elastic block
storage type and we'll talk more about

236

00:22:35,639  -->  00:22:41,559
EBS and instance store volumes but if
you have an EBS type you can actually

237

00:22:41,559  -->  00:22:47,679
stop and start those instances but you
can't do that if you have an instance

238

00:22:47,679  -->  00:22:53,200
store type of root volume attached to
that instance. So with an EBS backed

239

00:22:53,200  -->  00:22:57,700
instance you can stop it and their
instance will be shut down and you're

240

00:22:57,700  -->  00:23:02,680
not going to be get getting any more
instance charges but you will still be

241

00:23:02,680  -->  00:23:11,410
charged for any EBS volumes or any drives,
volumes that are attached to that ec2

242

00:23:11,410  -->  00:23:17,530
instance, and if you stop an instance in
less than one minute then you're still

243

00:23:17,530  -->  00:23:21,850
going to be charged for that one minute
when you restart it. There is another

244

00:23:21,850  -->  00:23:25,930
option and that is stop hibernate and
again it's only for EBS backed

245

00:23:25,930  -->  00:23:31,900
volumes and that is very good because it
will suspend all of your RAM to disk. So

246

00:23:31,900  -->  00:23:38,560
it will save your RAM to the EBS volume.
So it will allow for your instance to

247

00:23:38,560  -->  00:23:44,170
come back online very quickly and it
will keep all of that session data that

248

00:23:44,170  -->  00:23:50,710
was stored in that RAM. The other option
there is that we can reboot and an

249

00:23:50,710  -->  00:23:54,910
instance and we can do that for either
an EBS or an instance store backed

250

00:23:54,910  -->  00:24:00,220
volume and that is simply an operating
system reboot, and finally we can

251

00:24:00,220  -->  00:24:08,080
terminate those instances when we don't
need them anymore. Ok so first off we can

252

00:24:08,080  -->  00:24:14,500
use our ami to launch an ec2 instance,
while that instance is launching it will

253

00:24:14,500  -->  00:24:20,890
have the state of pending, after a period
of time it will then transition to the

254

00:24:20,890  -->  00:24:25,270
running state after it's done all of its
checks, while it's running we have an

255

00:24:25,270  -->  00:24:29,560
option there to reboot the operating
system, we can do that no matter what

256

00:24:29,560  -->  00:24:35,980
type of instance at it is it's simply an
operating system reboot, if we have an

257

00:24:35,980  -->  00:24:41,770
EBS backed instance and not an instance
source instance we have the opportunity

258

00:24:41,770  -->  00:24:46,900
there to stop that instance when we're
not using it and then restart it later

259

00:24:46,900  -->  00:24:51,730
on, it'll go back to that pending stage
and then obviously on to running again,

260

00:24:51,730  -->  00:24:57,520
and finally we can issue a terminate
command and while that terminate is in

261

00:24:57,520  -->  00:25:02,230
process it will be at the status of
shutting down and then it will

262

00:25:02,230  -->  00:25:07,050
transition to terminated after that.

263

00:25:10,010  -->  00:25:16,770
If you require a cluster of ec2
instances that need to be connected

264

00:25:16,770  -->  00:25:22,980
between each other with high-speed
networking, you can place those inside of

265

00:25:22,980  -->  00:25:32,610
an ec2 placement group. An ec2 placement
group will provide a low network latency

266

00:25:32,610  -->  00:25:38,550
it will be using a 10 gig interconnect
between those instances which will allow

267

00:25:38,550  -->  00:25:44,730
for a high network throughput. It's
available for all instances that support

268

00:25:44,730  -->  00:25:51,510
enhanced networking. It can't span
multiple availability zones. The cluster

269

00:25:51,510  -->  00:25:57,090
or the placement group must be in a
single availability zone but it can span

270

00:25:57,090  -->  00:26:04,860
a peared VPC but you're not going to get
that full bandwidth between that cluster

271

00:26:04,860  -->  00:26:12,510
and you cannot merge placement groups.
The enhanced networking required within

272

00:26:12,510  -->  00:26:19,370
a placement group uses single route IO
virtualization or SI-IOV for short and

273

00:26:19,370  -->  00:26:26,670
it provides a higher IO performance and
lower CPU utilization at no additional

274

00:26:26,670  -->  00:26:33,300
charge. It's supported in a number of
instances but not all of them and it

275

00:26:33,300  -->  00:26:38,880
requires an EBS optimized instance and
they are designed to deliver the

276

00:26:38,880  -->  00:26:48,320
provisioned IOPS that you specified for
99.9% of the time that they're operating.

277

00:26:52,910  -->  00:26:59,690
There are two options available for
storage directly to an ec2 instance and

278

00:26:59,690  -->  00:27:05,520
that's not including the external
storage for example using an Amazon s3

279

00:27:05,520  -->  00:27:11,280
bucket or using elastic file service. So
the two options there are elastic block

280

00:27:11,280  -->  00:27:18,870
store or instant store. Now the EBS is
the most common type and and maybe as

281

00:27:18,870  -->  00:27:24,780
volume will be replicated within an
availability zone which is great so if

282

00:27:24,780  -->  00:27:32,070
your drive fails there will be another
copy in another availability zone. Now

283

00:27:32,070  -->  00:27:38,750
the EBS volumes if they're attached at
instance launch they will be by default

284

00:27:38,750  -->  00:27:44,490
deleted when that instance is terminated.
Now that is a default behavior and you

285

00:27:44,490  -->  00:27:50,190
can modify that by changing the delete
on terminate checkbox in the console or

286

00:27:50,190  -->  00:27:53,669
the delete on
terminate flag if you're using the

287

00:27:53,669  -->  00:28:00,150
command line interface or an SDK. The 
EBS volumes that are attached to a

288

00:28:00,150  -->  00:28:06,030
running instance are not deleted by
default when the instance is terminated

289

00:28:06,030  -->  00:28:11,280
but they will be detached with that data
intact and you will still be getting

290

00:28:11,280  -->  00:28:17,250
billed for those abs volumes even though
you won't be getting billed for that ec2

291

00:28:17,250  -->  00:28:24,590
instance, but again that behavior can be
modified on the delete on terminate flag.

292

00:28:24,590  -->  00:28:30,809
Instance store volumes they are
physically attached to that host ec2

293

00:28:30,809  -->  00:28:37,500
instance and data is not lost when the
operating system is rebooted and that's

294

00:28:37,500  -->  00:28:42,780
the same for an elastic block store
backed ec2 instance as well but the data

295

00:28:42,780  -->  00:28:48,240
will definitely be lost if there's an
underlying drive failure because we

296

00:28:48,240  -->  00:28:54,330
don't have any replication there and
also if the instance is terminated that

297

00:28:54,330  -->  00:29:00,630
will be lost as well. So you don't rely
on an instance store for any valuable

298

00:29:00,630  -->  00:29:05,200
long-term data
but they are good because they do

299

00:29:05,200  -->  00:29:12,280
provide a little bit of a performance
advantage over EBS and you cannot detach

300

00:29:12,280  -->  00:29:22,090
an instance store volume and
attach it over to another instance. There

301

00:29:22,090  -->  00:29:28,419
are a number of storage options
available for EBS, the first one there is

302

00:29:28,419  -->  00:29:35,140
the default choice and is a
general-purpose SSD or it's called a GP2.

303

00:29:35,140  -->  00:29:41,200
The next one there is if you've got
some IO intensive applications such as a

304

00:29:41,200  -->  00:29:47,289
large relational database or a no sequel
database you can use a provisioned IOps

305

00:29:47,289  -->  00:29:54,850
Drive SSD as well and that's called an
IO1drive and that will give you high

306

00:29:54,850  -->  00:30:00,220
performance consistent and low latency.
The next one there are or the next two

307

00:30:00,220  -->  00:30:05,980
magnetic drives so they're going to be
lower cost lower performance options . So

308

00:30:05,980  -->  00:30:10,179
the first one there is a cold hard disk
drive and that will provide you the

309

00:30:10,179  -->  00:30:15,789
lowest cost per GB it's a 
standard magnetic drive. The next one

310

00:30:15,789  -->  00:30:20,380
there is a throughput optimized hard
disk drive and that's going to provide a

311

00:30:20,380  -->  00:30:27,750
low cost per bit per gigabyte but not as
low as the cold but it is very good for

312

00:30:27,750  -->  00:30:32,289
frequently accessed workloads for
example if you had a large amount of

313

00:30:32,289  -->  00:30:36,010
data and you wanted to low costs but you
were going to be accessing that data

314

00:30:36,010  -->  00:30:43,450
quite a bit for example for a big data
data warehouse. If we want to back up our

315

00:30:43,450  -->  00:30:50,049
EBS volumes we can take EBS snapshots
and they are a point-in-time backup of

316

00:30:50,049  -->  00:30:56,260
that EBS volume and they will be backed
up to the Amazon s3 service. They're an

317

00:30:56,260  -->  00:30:59,620
incremental backup so if you take
another snapshot after that it will be

318

00:30:59,620  -->  00:31:06,100
incremental and it can be copied to
other regions or other accounts to

319

00:31:06,100  -->  00:31:12,909
create new EBS volumes from that
snapshot. We also have encryption

320

00:31:12,909  -->  00:31:19,910
available on EBS. We can use the AWS KMS
master keys or you can use your own

321

00:31:19,910  -->  00:31:25,760
customer master key for that encryption
and the data stored at rest will be

322

00:31:25,760  -->  00:31:32,180
encrypted including any snapshots that
are created from that encrypted drive

323

00:31:32,180  -->  00:31:38,990
and also as well as data that is in
transit between the EBS volume and that

324

00:31:38,990  -->  00:31:42,370
ec2 instance.

325

00:31:46,610  -->  00:31:54,330
Another feature of ec2 is that we can
connect remotely into our ec2 instances

326

00:31:54,330  -->  00:32:00,590
so for example if we're sitting here at
our desktop with our Windows or Mac PC

327

00:32:00,590  -->  00:32:08,460
we can have a AWS access key which will
have an access key ID and a secret

328

00:32:08,460  -->  00:32:13,680
access key provided we've got that we
can use that to connect over the

329

00:32:13,680  -->  00:32:22,050
Internet to our ec2 instance and access
and run commands on the operating system

330

00:32:22,050  -->  00:32:28,110
of that instance. Not only do you require
those AWS access keys but you also need

331

00:32:28,110  -->  00:32:34,020
to have permission to access that
network so you need to have a security

332

00:32:34,020  -->  00:32:40,680
group inbound rule that will allow
access from your desktop you can connect

333

00:32:40,680  -->  00:32:48,750
to a Linux system using the secure shell
or SSH and mac or linux desktops will

334

00:32:48,750  -->  00:32:53,460
use a console straight out of the box to
do that. Windows computers will require

335

00:32:53,460  -->  00:33:00,870
an ssh client such as putty or a bash
clients such as cygwin to be installed

336

00:33:00,870  -->  00:33:06,320
on their computer before they can access
their Linux instance over SSH and

337

00:33:06,320  -->  00:33:12,030
connecting to a Windows Server instance
can be done using the remote desktop

338

00:33:12,030  -->  00:33:18,240
protocol or RDP. So that brings us to the
end of this lecture coming up we're

339

00:33:18,240  -->  00:33:24,000
going to be having quite a number of
pretty complicated labs on ec2 I look

340

00:33:24,000  -->  00:33:27,320
forward to seeing you in those.
