WEBVTT

00:00:00.940 --> 00:00:02.280
Earlier in this course,

00:00:02.280 --> 00:00:06.120
we took a look at a quick overview of the essential

00:00:06.120 --> 00:00:07.960
characteristics of the cloud.

00:00:08.540 --> 00:00:10.810
Let's look at them in a little more detail.

00:00:10.810 --> 00:00:14.790
And for this, I'm going to rely on the NIST special publications,

00:00:14.800 --> 00:00:17.810
800‑145 and 146.

00:00:18.440 --> 00:00:21.730
Those are free documents, and they give us a good,

00:00:21.740 --> 00:00:25.090
fairly well‑recognized standard we can use.

00:00:26.540 --> 00:00:31.110
The first of those essential characteristics was broad network access.

00:00:31.530 --> 00:00:36.990
That means we have network‑based connectivity between the cloud consumer and

00:00:36.990 --> 00:00:41.340
cloud service provider that is essential for this all to work.

00:00:41.340 --> 00:00:46.240
Now the idea it was called the cloud in the early days was because

00:00:46.240 --> 00:00:50.460
whenever we drew network diagrams back in the 1990s,

00:00:50.840 --> 00:00:53.260
we always drew the internet as a cloud.

00:00:53.740 --> 00:00:59.060
And so the idea of cloud services was really just the marketing

00:00:59.190 --> 00:01:03.260
of saying we're going to use internet‑based or internet types of

00:01:03.260 --> 00:01:07.640
connectivity in order to be able to communicate with then the

00:01:07.640 --> 00:01:09.660
data hosting organization.

00:01:10.240 --> 00:01:14.360
So the term cloud really referred to the internet.

00:01:14.360 --> 00:01:18.550
So we know that for many organizations today,

00:01:18.550 --> 00:01:22.180
we very much benefit from being able to use really

00:01:22.190 --> 00:01:26.360
inexpensive connectivity rather than leased lines,

00:01:26.360 --> 00:01:27.940
for example.

00:01:27.940 --> 00:01:32.900
One of the things about the internet is that it is fairly redundant.

00:01:32.900 --> 00:01:34.350
It's fairly robust.

00:01:34.630 --> 00:01:37.920
Certainly there have been some outages and so on,

00:01:37.920 --> 00:01:40.520
but usually they're only in a small region,

00:01:41.440 --> 00:01:45.470
but there's always still that risk of a single point of failure,

00:01:45.470 --> 00:01:50.320
the connection from the last point of the telecom

00:01:50.320 --> 00:01:53.740
providers building to our office.

00:01:53.740 --> 00:01:57.730
It's sometimes called the last mile or the subscriber loop.

00:01:57.880 --> 00:02:01.230
The idea is that that may be a single point of failure.

00:02:01.530 --> 00:02:04.800
So a large cloud service provider will usually make sure

00:02:04.800 --> 00:02:07.150
they at least have redundant connectivity.

00:02:07.740 --> 00:02:09.910
But even that can be difficult to maintain.

00:02:09.910 --> 00:02:11.780
Things get moved around.

00:02:11.780 --> 00:02:15.270
And there's been cases where companies believe they had

00:02:15.270 --> 00:02:19.240
redundant fibers only to find that the fibers,

00:02:19.240 --> 00:02:21.900
even though they are from two different telecom providers,

00:02:21.900 --> 00:02:25.460
were actually in the same cable in the same trench.

00:02:26.340 --> 00:02:29.170
The other thing is to make sure we have enough capacity.

00:02:29.370 --> 00:02:34.310
We certainly have seen a great increase in the amount of data we transfer,

00:02:34.310 --> 00:02:40.720
and we have to ensure that the cloud carrier will be able to

00:02:40.720 --> 00:02:44.560
support peak demands of traffic as well.

00:02:45.540 --> 00:02:48.810
The other challenge we have is that we have many

00:02:48.810 --> 00:02:51.750
different types of endpoint devices, well,

00:02:51.750 --> 00:02:55.970
even different types of mobile phones, everything from,

00:02:55.970 --> 00:02:58.710
of course, different operating systems, but you know,

00:02:58.710 --> 00:02:59.650
that's not all.

00:02:59.760 --> 00:03:03.680
We have many devices connecting to the cloud now from all

00:03:03.680 --> 00:03:06.350
sorts of Internet of Things devices,

00:03:06.350 --> 00:03:10.890
and each of those can have its own proprietary languages,

00:03:10.890 --> 00:03:12.950
its own proprietary needs.

00:03:13.050 --> 00:03:18.140
And so when we look at the cloud carrier and broad network access,

00:03:18.510 --> 00:03:22.220
we have to be able to ensure we can support all those

00:03:22.220 --> 00:03:25.260
various types of endpoint devices as well.

00:03:27.040 --> 00:03:31.030
The idea of on‑demand self‑service is the ability of

00:03:31.030 --> 00:03:36.470
the cloud consumer to change their, should we say,

00:03:36.470 --> 00:03:41.660
contracted usage of the cloud without going through all kinds of

00:03:41.660 --> 00:03:45.360
administrative negotiations and contracts.

00:03:45.970 --> 00:03:49.620
The idea now is that the cloud consumer can increase or

00:03:49.620 --> 00:03:53.790
decrease their requirements for compute,

00:03:53.790 --> 00:03:57.860
processing, storage, and network services automatically.

00:03:58.340 --> 00:04:00.920
And we all know that many of us have done this.

00:04:00.920 --> 00:04:02.740
You can easily add something on,

00:04:02.740 --> 00:04:07.640
and you don't have to negotiate with a salesperson or whatever or

00:04:07.640 --> 00:04:11.060
in any way interact with cloud provider staff.

00:04:11.540 --> 00:04:16.220
And that's much better than the old mainframe days where we had to

00:04:16.220 --> 00:04:19.240
negotiate for months to get a few more MIPS,

00:04:19.240 --> 00:04:25.860
or millions of instructions per second, or a new logical partition and so on.

00:04:27.940 --> 00:04:33.080
One of the greatest advantages, of course, to the cloud is resource pooling.

00:04:33.080 --> 00:04:36.550
The fact that we used the example before,

00:04:36.550 --> 00:04:40.080
we can have 15 companies that each had their own data center,

00:04:40.080 --> 00:04:42.520
now they use one data center.

00:04:42.520 --> 00:04:46.440
That gets rid of 15 air conditioning units,

00:04:46.450 --> 00:04:49.250
diesel generators, backup batteries,

00:04:49.260 --> 00:04:53.180
15 companies all managing their own physical access,

00:04:53.190 --> 00:04:55.450
each one having their own spare parts.

00:04:55.840 --> 00:05:01.640
No, we now share one facility in a multi‑tenant model.

00:05:01.650 --> 00:05:03.990
Now this, of course, brings in a concern.

00:05:04.380 --> 00:05:08.650
My data is in the same building as maybe my competitors data.

00:05:09.240 --> 00:05:12.120
But the advantage is we have pooled resources.

00:05:12.120 --> 00:05:17.830
Instead of me needing to have enough spare capacity to meet peak needs,

00:05:18.240 --> 00:05:26.510
I am able to access a pool of spare capacity that I can use when I need,

00:05:26.510 --> 00:05:28.760
and I release when I don't need.

00:05:30.440 --> 00:05:31.720
So this, of course,

00:05:31.720 --> 00:05:36.660
includes everything from equipment such as memory and processing.

00:05:36.860 --> 00:05:39.620
But the other thing that many companies struggle with today,

00:05:39.620 --> 00:05:40.940
of course, is staff.

00:05:40.940 --> 00:05:47.060
Where do we find knowledgeable staff to support all of our IT needs?

00:05:47.340 --> 00:05:49.230
Well, if I can use a cloud provider,

00:05:49.230 --> 00:05:52.670
they have staff there that is dealing with different

00:05:52.670 --> 00:05:56.010
types of IT support for many companies.

00:05:56.010 --> 00:06:01.700
They have experience, and, of course, we have economies there of scale and so on.

00:06:02.140 --> 00:06:07.540
We also have, of course, the network where we're able to use a shared network,

00:06:07.540 --> 00:06:11.270
but this is, again, where we have to look at enough capacity and so on.

00:06:12.240 --> 00:06:13.990
We share memory.

00:06:14.080 --> 00:06:17.780
The problem with that is I want to make sure that there's no

00:06:17.780 --> 00:06:22.740
residual data left on a piece of equipment that could then be

00:06:22.740 --> 00:06:25.580
accessed in property by somebody else.

00:06:26.740 --> 00:06:32.650
We can, of course, have a private type of cloud off‑site.

00:06:32.790 --> 00:06:37.010
Of course, onsite we can have one, but when we deal with private offsite,

00:06:37.010 --> 00:06:42.040
we're renting some floor space from a large cloud service provider,

00:06:42.040 --> 00:06:46.960
and we have our own equipment in there that may be in a cage of some type.

00:06:47.440 --> 00:06:53.540
And the cloud provider provides all of the physical facility,

00:06:53.540 --> 00:06:57.720
network power, heating, ventilation, air conditioning,

00:06:57.730 --> 00:07:03.170
physical access, but we don't have resource pooling.

00:07:03.170 --> 00:07:06.740
In this case, the hardware, the memory,

00:07:06.740 --> 00:07:10.140
and so on that we have purchased and installed at the cloud

00:07:10.140 --> 00:07:13.500
service provider's premises is for us only,

00:07:13.500 --> 00:07:18.360
and we don't have easy access always to additional resources.

00:07:19.240 --> 00:07:21.860
So the idea here is we still have pooling,

00:07:21.860 --> 00:07:26.550
but only the infrastructure, the building level we could say.

00:07:26.550 --> 00:07:28.920
Now they, of course,

00:07:28.920 --> 00:07:33.650
do have the advantage that they can provide staff and support for us.

00:07:35.740 --> 00:07:39.590
Another of the advantages, of course, is rapid elasticity.

00:07:39.590 --> 00:07:45.850
Rapid elasticity refers to the ability of the consumer

00:07:45.980 --> 00:07:48.780
to be able to access or provision,

00:07:48.780 --> 00:07:53.200
but also then to release service capabilities automatically,

00:07:53.200 --> 00:07:57.460
part of this self‑service according to demand.

00:07:57.840 --> 00:08:03.450
So the idea here is that an organization is able to acquire greater

00:08:03.450 --> 00:08:10.630
bandwidth processing and storage capabilities very easily,

00:08:10.630 --> 00:08:13.050
but also release them when they don't need them.

00:08:13.050 --> 00:08:18.420
So we can meet peak needs at certain times of the year and not pay for

00:08:18.420 --> 00:08:21.850
those pieces of equipment the times we're not using them.

00:08:23.340 --> 00:08:27.550
We also have then the ability to do things like cloud bursting.

00:08:28.040 --> 00:08:29.710
Maybe I have a private cloud,

00:08:30.240 --> 00:08:36.059
but I realize I have a peak demand that I need additional resources,

00:08:37.539 --> 00:08:42.640
so I would be able to actually migrate some of my services

00:08:42.640 --> 00:08:47.030
temporarily onto a public cloud type of environment.

00:08:47.030 --> 00:08:49.760
Measured service is important.

00:08:49.930 --> 00:08:51.760
We pay for what we use.

00:08:52.240 --> 00:08:54.650
That means we don't pay for what we don't use,

00:08:54.650 --> 00:08:59.000
but it also means that quickly we can end up paying for a lot of

00:08:59.000 --> 00:09:03.250
stuff we're using that maybe we shouldn't be.

00:09:03.740 --> 00:09:08.440
So the idea here of building and monitoring is that I don't have the

00:09:08.440 --> 00:09:11.610
initial capital expense of some of the equipment,

00:09:12.040 --> 00:09:15.850
but I'm able to only pay for the resources that I'm

00:09:15.850 --> 00:09:18.090
actually using at any one time.

00:09:18.470 --> 00:09:18.930
Of course,

00:09:18.930 --> 00:09:21.470
this doesn't apply to a private cloud because that's

00:09:21.480 --> 00:09:23.460
my equipment that I've purchased.

00:09:24.040 --> 00:09:28.030
The one advantage sometimes of that though is that if I'm

00:09:28.030 --> 00:09:30.680
working with a large cloud service provider,

00:09:30.680 --> 00:09:33.200
they're probably getting a far better price on a

00:09:33.200 --> 00:09:34.880
lot of their hardware than I am,

00:09:34.880 --> 00:09:38.640
and it could be we're even using them to provision equipment

00:09:38.640 --> 00:09:41.170
for our private cloud if it's an off‑premise.

00:09:41.940 --> 00:09:44.850
Again, there's so many different types of deployments.

00:09:45.530 --> 00:09:49.320
The one thing that you'll see later on that becomes a

00:09:49.320 --> 00:09:52.340
factor in this is what we call VM sprawl.

00:09:52.340 --> 00:09:59.270
And this is where, as an organization, we spin up a virtual machine.

00:09:59.270 --> 00:10:01.030
When we spin that up,

00:10:01.030 --> 00:10:04.630
a certain number of resources are allocated to us for

00:10:04.630 --> 00:10:08.280
our use of that virtual machine, and we start paying for those.

00:10:08.280 --> 00:10:12.500
Let's say that a few moments later we don't need that and

00:10:12.510 --> 00:10:17.910
the technician walks away and they leave that still as

00:10:17.910 --> 00:10:20.600
dedicated resources that we pay for.

00:10:20.600 --> 00:10:24.680
And this is a very common problem that companies have.

00:10:24.680 --> 00:10:26.560
It's called VM sprawl.

00:10:26.690 --> 00:10:31.340
We end up and I've seen cases of hundreds of virtual machines that have never

00:10:31.340 --> 00:10:36.240
been used or haven't been used in years that company is paying for because

00:10:36.240 --> 00:10:39.160
they're not cleaning up and releasing those as well.

00:10:39.340 --> 00:10:43.050
So that's one of the disadvantages of measured service.

00:10:43.050 --> 00:10:47.880
It's too easy to use too many resources without realizing

00:10:47.960 --> 00:10:50.210
what we're using and what it's costing us.
