- Software as a Service (SaaS)
- Infrastructure as a Service (IaaS)
- Platform as a Service (PaaS)
- Broad network access
- On-demand self-service
- Resource pooling
- Measured service
- Rapid elasticity
The best way to start
with that is to compare it to traditional IT computing. Where on-premises on
our own networks, we would at some point have a capital investment in
hardware. So think of things like having a server room constructed, getting
racks and then populating those racks with equipment. With things like telecom
equipment, routers, switches, servers, storage arrays, and so on.
Then, we have to
account for powering that equipment. We then have to think about HVAC,
heating, ventilation and air conditioning, to make sure that we've got optimal
environmental conditions to maximize the lifetime of our equipment.
Then there's
licensing. We have to license our software. We have to install it, configure
it and maintain it over time, including updates. So with traditional IT
computing, certainly there is quite a large need for an IT staff to take care
of all of our on-premises IT systems.
But with cloud
computing, at least with public cloud computing, we are talking about hosted
IT services. Things like servers and related storage, and databases, and web
apps can all be run on provider equipment
that we don't have to purchase or maintain. So in other words, we only pay for the services that are used.
And another part of the cloud is self-provisioning,
where on-demand, we can provision, for example additional virtual machines or
storage. We can even scale back on it
and that way we're saving money because we're only paying for what we are
using.
With cloud computing,
all of these self-provisioned services need to be available over a network. In the case of public clouds,
that network is the Internet. But
something to watch out for is vendor lock-in. When we start looking at cloud
computing providers, we want to make sure that we've got a provider that won't
lock us into a proprietary file format for instance. If we're creating
documents using some kind of cloud-based software, we want to make sure that
data is portable and that we can move it back on-premises or even to another
provider should that need arise. Then there is responsibility.
This really gets
broken between the cloud provider and the cloud consumer or subscriber,
otherwise called a tenant. So the degree
of responsibility really depends on the specific cloud service that we're
talking about. But bear in mind that there is more responsibility with cloud
computing services when we have more control. So if we need to be able to
control underlying virtual machines, that's fine, but then it's up to us to
manage those virtual machines and to make sure that they're updated. The hardware is the provider's responsibility.
Things like power,
physical data center facilities in which equipment is housed, servers, all
that stuff. The software, depending on what we're talking about, could be
split between the provider's responsibility and the subscriber's
responsibility. For example, the provider might make a cloud-based email app
available, but the subscriber configures it and adds user accounts, and
determines things like how data is stored related to that mail service. Users
and groups would be the subscriber's responsibility when it comes to identity
and access management.
Working with data
and, for example, determining if that data is encrypted when stored in the
cloud, that would be the subscriber's responsibility. Things like data center
security would be the provider's responsibility. Whereas, as we've mentioned,
data security would be the subscriber's responsibility when it comes to things
like data encryption. The network connection however is the subscriber's
responsibility, and it's always a good idea with cloud computing, at least
with public cloud computing, to make sure you've got not one, but at least two
network paths to that cloud provider.
A cloud is defined by resource pooling. So, we've got all this IT
infrastructure pooled together that can be allocated as needed. Rapid elasticity means that we can quickly
provision or de-provision resources as we need. And that's done through an
on-demand self-provisioned portal, usually web-based. Broad network access means that we've got connectivity available
to our cloud services. It's always available. And measured service means that it's metered, much like a utility, in
that we only pay for those resources that we've actually used. So, now we've
talked about some of the basic characteristics of the cloud and defined what
cloud computing is.
For non-techies out
there, the cloud might be an intimidating and nebulous concept. We hear about
cloud computing all the time, but what exactly does it mean?
The National
Institute of Standards and Technology (NIST) describes the basics of cloud
computing this way:
Cloud computing is a model for enabling ubiquitous,
convenient, on-demand network access to a shared pool of configurable
computing resources (e.g., networks, servers, storage, applications, and
services) that can be rapidly provisioned and released with minimal management
effort or service provider interaction.
Still confused?
In short, the cloud
is the Internet, and cloud computing is techspeak that describes software and
services that run through the Internet (or an intranet) rather than on private
servers and hard drives.
Cloud computing is
taking the world by storm. In fact, 94% of workloads and compute instances
will be processed through cloud data centers by 2021, compared to only 6% by
traditional data centers, according to research by Cisco.
So why companies are moving from traditional severs
and datacentres to cloud ?
With
cloud computing, it allows companies to have more time to focus on their
business, instead of on the IT infrastructure. One of the reasons for this is
because most of the IT infrastructure is the responsibility of the cloud
provider.
With the
cloud we have metered usage, whereby all of the usage of cloud resources is
tracked and that's how we get billed. So, what that ends up meaning at the
accounting level is that we're dealing with a monthly recurrent operating
expense versus a capital expenditure. Which might be less frequent, but
nonetheless requires capital for things like the acquisition of hardware to
run on-premises networks. So, it's a monthly ongoing expense which often might
include a subscription plus a usage fee. But also the nice thing about it is
it can facilitate budgeting over time as we set a baseline for our usage of
cloud resources in our environment.
So, you
pay for what you use similar to a utility like electricity or water.
We also
have to bear in mind that the IT needs of an organization to support its
business processes will change over time. Things like the amount of disk space
that's needed, or the underlying server horsepower, maybe to perform Big Data
analytics. We have to think about lost productivity and downtime from failed,
or misconfigured, or even hacked IT systems. Then we have to recognize, in the
cloud we've got the ability to rapidly provision and also deprovision
resources. That can even be automated in some cases. Bear in mind that
deprovisioning, for example scaling in, that would be removing virtual machine
instances to support an app, saves money. Because we have less virtual
machines running in the cloud at the same time.
In the
cloud it's also very easy to enable high availability across different
geographical regions. For example, to make sure our data or database replicas
are highly available. The cloud also has a global reach with points of
presence around the world with Google Cloud Platform. So, if we want to enable
some content to be made available to users in Europe. Even though the web site
that really hosts the content is in North America, we can do that using things
like a cloud content delivery network,
or CDN.
We should
also plan to monitor the network link between our on-premises network and the
cloud. Assuming that we have connectivity from on-premises to the cloud,
whether it's for data center purposes or that's where end users are. We need
to make sure that's done properly because we want to make sure that we have
redundant links to the cloud. That's essentially the single point of failure
between a company that has an on-premises environment as well as a cloud
environment. The Internet links us between them and so, we need to make sure
that it's trustworthy.
Then
there's cloud economics. Consider this
example, where we might have a return on investment, or ROI, of 5 years. Now
that could mean that we have a payback period of 5.5 months. So that our total
cost of ownership, or TCO reduction could potentially be upwards of 64%. Now
these numbers will certainly vary depending on your usage of the cloud. Like
the use of any type of commodity so to speak, it has to be used efficiently.
That doesn't take care of itself in the cloud. We still have to manage it
properly. It still requires proper IT workload and IT system governance.
So, what
are some factors, then, that could feed into some of these potential cost
savings? Well, one is less time spent updating and managing IT resources,
because the bulk of that is the responsibility of the public cloud provider.
Also high availability. We're more likely to have systems left up and running
over the long term if they're hosted in different data centers owned by a
cloud provider than we are on-premises. Then there's the rapid provisioning
and deprovisioning or the scalability of resources. So, if we offer some kind
of retail services online for shopping, maybe during holiday seasons. We have
it configured to auto-scale to add other virtual machine instances to support
the peak demand. But then when that disappears, so do the extra virtual
machine instances. So, we're only paying for the peak demand usage that's
required. And the lessened risk of downtime, besides high availability, is
also something we consider in the SLA, the service level agreement. The
service level agreement is a contractual document between the cloud provider
and the cloud consumer. And one of the many things it will cover is things
like expected uptime, and the consequences of that not being met.
Public Clouds
There are
varying cloud models, one of which is the public cloud. The public cloud is
accessible to all users over the Internet, potentially. Now, of course,
they're going to have to sign up, either initially for a trial account, or
initially or after the trial is expired for a paid account. But nonetheless,
the public cloud is accessible to individuals. So, individual consumers as
well as organizations, government agencies, academic institutions, and so on.
So, this means then that with a public cloud environment, the public cloud
provider has data centers around different geographical locations in the
world. And as a result, they also have multiple cloud tenants or customers. Of
course, each cloud tenant or customer, their configurations, and their data
are kept completely separate from other cloud tenants. Otherwise, nobody would
use the service, whether it be individuals or organizations.
Cloud
provider hardware is owned by the cloud provider and exists in data centers in
different regions around the world physically. So, what we end up having then
is a large scale of pooled resources. Things like servers with a lot of
horsepower that can act as hypervisors and run virtual machine guests and
applications. Things like Docker containers. Also, we have things like pooled
resources in the form of network equipment. That can be configured at a higher
level by cloud tenants. Things like routing paths and firewall rules. Another
example of a pooled resource, of course, would be storage that customers or
tenants can provision at a moment's notice, or deprovision at a moment's
notice, as well. So, the cloud provider, then, benefits from economies of
scale. Due to the large number of cloud tenants or customers, it allows them
to be able to acquire all of these facilities containing all of this physical
computing equipment. And to assure that they meet specific security standards,
so that they are trustworthy, and customers will actually use their service
and pay for it. And so as a result, the economies of scale mean that users can
end up perhaps with a small subscription fee but really, they only pay for
what they are using. And that is in quite a stark contrast to the purchase of
equipment on premises, and licensing in software, where once you've purchased
it it's yours.
Then we
have to consider network connectivity to and from the public cloud. Now,
normally this is done over the Internet. Certainly, for individuals this is
the case and even for most organizations. So, we have access to the public
cloud provider services through the Internet. In the corporate environment,
user devices like mobile smartphones, laptops, desktops will maybe reside on a
corporate on-premises network that connects over the Internet to the cloud.
Now, that can also happen even through a VPN if required, if we need that
encrypted tunnel or if we want to link an on-premises network directly with a
cloud-based network. Now, also, there's the option of a private dedicated
connection.
With
Google Cloud Platform, that's called Google
Dedicated Interconnect. We'll get into the details later. But for now,
essentially, it's a dedicated network link from your on-premises network or
data center to the Google Cloud without going through the Internet. Now, if we
are going to use network connectivity such as the Internet, we should consider
redundant network paths. And ideally, that would be network connections from
different Internet service providers. We don't want a single point of failure.
So that if there's a problem with one Internet service provider, we can fall
back on our Internet link to the Google public cloud to our secondary Internet
service provider network link.
Private Clouds
Besides
the public cloud type, we also have private clouds. Private clouds are
different in that they're accessible only to a single organization. Whereas
public clouds are accessible to anyone that wants to sign up over the
Internet. So, it's organization-owned hardware infrastructure that the
organization is responsible for configuring and maintaining over time.
However, a private cloud still adheres to the exact same cloud characteristics
that a public cloud does such as self-provisioning. Whether it's through some
kind of a web user interface and the rapid elasticity of pooled IT resources,
such as the ability to use a web interface to quickly deploy virtual machines
or additional storage, maybe for use by a particular department within the
organization.
So,
you'll see private clouds in larger enterprises are often used for
departmental chargeback. So, another common characteristic of a cloud is, metered usage. So, all the usage of resources
is tracked and then, in this case, it's charged back to departments that use
those IT resources.
A private
cloud can also be extended into the public cloud. Now, that's where we get
into what's called a hybrid cloud, and
we might use that for something called cloud
bursting. Cloud bursting essentially means that, once we've depleted
our on-premises resources, whether that be the number of virtual machines we
can run, or we've consumed all of our storage capacity on-premises. We've got
a link to the cloud as a secondary plan B that can be used to provision, in
our case, maybe additional virtual machines or additional storage. And in some
cases, it won't even be known to the end user, this is actually happening into
a different environment, into the public cloud.
The
private cloud is sometimes the only real choice we have, because regulations
might prohibit or limit the use of public cloud providers. So, if we're
dealing with some kind of government sensitive data, we might be prohibited
from that leaving a specific network under military control if it's that type
of information.
In a
private cloud, the organization has way more IT responsibility. Now, that
comes in the form of the cost of acquiring all the equipment and paying
personnel to configure it and maintain it over time. But also, fault tolerance
is also the responsibility of the organization. So, the organization is responsible for the installation,
configuration, and maintenance of IT equipment for the private cloud.
Now,
organizations can have internal SLAs or service level agreements that
guarantee uptime, for example, to departments that pay for their usage of
those private cloud resources. And fault tolerance might be achieved by using clustering on-premises. Multiple servers that
offer an app, if a server node goes down, then another one can pick up the
slack and continue running the app.
The
organization, in a private cloud, also has full
control of that environment in how it's configured, how it's maintained
and updated.
And, of
course, there's definitely a further degree of
privacy depending on the configuration. But generally, this is true
because we don't have any data that is leaving the on-premises network. That's not to say that a private cloud is
always more secure than the public cloud. Not necessarily true at all. But
there is more privacy.
Community Clouds
We've
already discussed public and private clouds. Another cloud type is a community
cloud. A community cloud uses cloud provider IT infrastructure and pooled
resources. So, really that's just cloud computing unto itself. However, what
makes them a little bit different than a public or a private cloud is that
there are same needs, or the same requirements exist across multiple tenants.
Now, of course, that means that if the public cloud provider also offers
public cloud computing to anyone over the Internet, that there is isolation
from those public tenants.
In some
cases, a community cloud would be the perfect solution to remain compliant
with certain industry regulations. So, a
community cloud must ensure regulatory compliance.
So, for
example, for regulatory compliance, in terms of security, we might use G Suite
for Government. Now, there also is FedRAMP compliant, where FedRAMP stands for
the Federal Risk and Authorization Management Program. Where US government
departments have to have certain security controls in place. Then on the
medical side there's HIPAA compliance. HIPAA stands for the Health Insurance
Portability and Accountability Act, and again, as usual, it's really about
data governance on the security side. Making sure that data is protected with
reasonable encryption. Whether it be data at rest, that would be stored data,
or data being transmitted by configuring things appropriately like firewalls
and intrusion detection systems. Other possible regulatory compliance items
would deal with things like long term data retention policies and archiving
and making sure that it's immutable. In other words, that something that's
archived cannot then be modified.
Hybrid Clouds
You can
think of a hybrid cloud as really being the best of both worlds. When we talk
about both worlds we're talking about on-premises
computing environments as well as cloud
computing environments working together at the same time. Now this
would be something that's common with public cloud adoption for an
organization. Where we might run a private and public set of IT workloads in
parallel or together as we migrate over time. Now that would mean taking data
or systems or virtual machines. Even converting on-premises
physical servers to virtual machines in the cloud. That physical to
virtual conversion is called P2V. We can
also use a hybrid cloud as a way to extend our private network into the cloud.
And often that would be done for things like cloud bursting to meet peak
demands. So that when we've depleted the IT resources available on-premises,
we simply seamlessly extend into the public cloud to use additional resources
there.
The
migration of our on-premises systems and data can take a long time. So for
example, we might have data that's stored on-premises and in the cloud. And
maybe that is our backup solution, it's an off-site backup. So we might have
an appliance on-premises that allows users to access data. But at the same
time maybe it synchronizes or replicates data to the cloud for backup
purposes. Could also be the other way around. We might also have cloud data
presented to an on-premises appliance, which then presents itself to backup
software on-premises as something like a virtual tape library. So our
on-premises backup software just thinks, oh yeah, it's another regular backup
source.
Now, in
the case of the migration of very large amounts of data into Google Cloud. It
might be unfeasible to think of doing that over and even the highest speed
Internet connections. So instead, we might look at something like the Google Transfer Appliance. And what this
really means is that we've got an on-premises storage appliance that we
essentially copy data to. And that appliance is then physically shipped to a Google Cloud Platform data center where
the data is then copied onto their infrastructure.
Another
possible option to consider when we talk about hybrid clouds is the Google Dedicated Interconnect feature. We'll
focus on its details later. But what it really lets us do is bypass the
Internet by having a private dedicated network
link from our on-premises network directly into the Google Cloud. Now, that
normally means that we have more predictable
throughput than we would by using a shared connection over the Internet. And
one might argue it might offer more security since it's private. Now of
course, over the Internet we could also achieve security by using a VPN tunnel. But Google Dedicated Interconnect
is yet another consideration when we think about hybrid cloud computing.
Service Models in Cloud Computing
One type
of cloud computing service model is software as a service, which is otherwise
simply called SaaS. Now, really what this is, is application or office
productivity software that's hosted on provider equipment. And it's accessible
over a network such as the Internet. So normally, it requires only a web
browser. It might require some plugins. But usually, the web browser and a
plugin would be the extent to what's required to access SaaS services. And
often they'll offer some degree of customization. Where you can configure
maybe the visual appearance or skins or certain security settings related to
the specific SaaS tool you're using.
For
example, if we're using office productivity tools like word processors through
a cloud provider. So with software as a service because it is cloud computing,
we pay for the usage of a service. For example, we have Gmail and G Suite,
which are considered software as a service offerings through Google. And in
the case of Gmail, depending on whether it's being used by an individual or a
large enterprise determines on the cost of it, in terms of the storage space
for archived messages. Or the number of messages in and out.
Software
as a service relies on virtualization. And this also means that we've got
tenant isolation. Now the virtualization doesn't necessarily mean virtual
machine operating system virtualization. It could also be application
virtualization using what are called application containers. In other words, no control of underlying virtual
machines, also known as VMs. In the end, as a cloud consumer, we
don't really care as long as the isolation is in place. So for example, if I'm
using a cloud Gmail solution, I want to make sure that my organization's Gmail
accounts are not in any way accessible to other Google cloud tenants. So when
we talk about software as a service, yes, it does rely on virtualization of
some kind at some level. However, that's not visible and certainly its
configuration is not available to us as cloud consumers. That configuration,
including things like making changes or applying patches is the responsibility
of the cloud provider. Software as a service also removes the need for things
like installing software and licensing it. So really, it transfers the IT
management responsibility from our organization to the cloud provider instead.
With
software as a service, one business consideration is vendor lock-in. So we
might take a look at the service level agreements for the various cloud
services we use from a specific cloud provider. Service level agreements are also known as SLAs. Because we
want to make sure that if we decide that we want to stop cloud computing, we
want to go back on-premises or switch to a different public cloud provider,
that we can. And that might mean that we have to ensure that data can be
exported in a standard format from one provider and then imported to another.
Or even live migrated over the Internet.
For example, with Google, we have the option
of actually migrating Amazon Web Services storage data from Amazon Web
Services directly into Google Cloud Storage. We then have to think about our
network connection. Because when we depend on software as a service, that
network connection to the cloud is our single point of failure if we've only
got one network connection. So, there is a
dependency on connecting to the cloud provider. We should also make
sure that we acquire sufficient bandwidth in that connection. And that would
depend upon, for instance, the number of users that would concurrently be
using software as a service through that network link. And again, due to
having a single point of failure, we should really consider having redundant
network paths. In the case of Internet connections, that would be multiple
Internet connections ideally in a perfect world through different Internet
service providers.
Now we
will look into infrastructure as a
service. Just like software as a service is a different type of cloud service
model, well, so is infrastructure as a service and it's normally just called IaaS. It relies on virtualization. For
example, working with virtual machines. Virtual
machines are also known as VMs.When we deploy our virtual machine
instance, that's infrastructure as a service. When we provision new storage in
the cloud, that is infrastructure as a service. So is things like network
configurations, like firewall rule sets. It's all part of our network
infrastructure. Physical equipment that makes all of this possible is actually
housed in Google Cloud data centers around the globe. Now because that
acquisition of hardware, that expense in powering it and cooling it and so on
is done by Google, it’s their responsibility. So for us, the cloud customer,
it reduces capital expenditures on our end. All we do is pay for what we use,
which is really just an operational expense.
Infrastructure
as a service can be provisioned using the Google Cloud console, so using the
web GUI. Or of course it can be provisioned and managed programmatically
through various APIs. Or even using the GCloud command line tool. We provision
and manage infrastructure as a service over the network. And it also supports
scalability, on-demand where we pay for the usage. So if we decide, for
example, we want to provision four new virtual machines, we can do that in a
matter of minutes with a few clicks. That same thing is true if we want to
provision additional storage. And also deprovisioning
resources that we no longer need to save on costs. This might even be
automated. So we might even configure something like a VM instance group in
Google Cloud Platform and configure it with a specific number of instances to
support things like autoscaling. So that
when we have a peak demand in an app, we scale out horizontally. We add more
virtual machines to support the app and when that demand declines, so does the
number of virtual machines we scale back in.
Google Cloud Storage is definitely considered
infrastructure as a service. All of the network configurations are also part
of infrastructure, and this uses something called software defined networking. Now, that's not specific to Google
Cloud computing. It's a standard cloud computing term that's often referred to
as SDN. And what it really means is that
we, the cloud customers, have an easy way of configuring network resources.
Deploying
cloud virtual networks and IP ranges and firewall rules without actually
communicating directly with the underlying physical network hardware that
makes all of that possible. Of course, we would then deploy virtual private clouds or VPCs. In
other words, a VPC is a virtual network in the cloud. These are
simply virtual network configurations that we define in the cloud. And then we
deploy our resources like virtual machine instances into the VPCs. We
mentioned that we have firewall rule sets that can be configured for a VPC or
even specific VM instances that control inbound or outbound network traffic.
So what
is infrastructure as a service normally used for?
So, infrastructure as a service would commonly be
used as follows. Well, often it will be used for testing purposes, because it's so quick and
easy to provision, for instance, virtual machines that are isolated from
others in a VPC. So it's sandboxed, maybe to test an application quickly. Or
to host a web site, whether it's for
private organizational use, or it's a public-facing
web site. We might use infrastructure as a service for cloud storage and/or backup. Even for high
performance computing or HPC, where we can configure clusters for the purpose
of parallel computing, which you might use for things like Big Data analytics.
Platform
as a service is yet another type of cloud computing service model, and it's
often simply called PaaS. Now it's related more to applications, the
development of applications, and then the deployment of them. Whether that's
in a staging or test environment or deploying it into production. So the focus
then is on the application rather than the underlying supporting resources
like network configurations, virtual machines, and storage. Or virtual servers.
Platform
as a service-specific resources would include things like development tools.
Now that would include also APIs that expose cloud functionality to developers
in a variety of different languages. And that's definitely the case with
Google Cloud Platform. Many of the cloud services are available to developers
in languages like C#, Python, Java and so on. Also platform as a service deals
with databases. So if we want to deploy a MySQL database, for example, in the
cloud the underlying complexities like virtual machines, installing the
software and so on, that's already taken care of in most cases. We simply tell
our environment, our GUI environment, or whether we're doing it through the
command line, that we want to deploy some kind of a database in the cloud. A
couple of parameters, such as, like the number of replicas, whether we want to
use our own license, that type of thing. And it just happens very quickly.
Other
types of PaaS resources would include things like business intelligence tools
or Big Data analytical tools. These are essentially like extensions that can
take data that's been massaged through our platform as a service use. And then
we can gain insights from them that otherwise wouldn't readily be apparent.
Some examples of Google Cloud platform as a service offerings would include
MySQL or even using the Docker Container
Registry. Where Docker containers are essentially a way to isolate
applications from one another, like virtual machines can isolate operating
systems from one another.
Disaster Recovery as a Service (DRaaS)
Even
though it doesn't sound like a good idea, planning for failure is a good idea.
It's really related to disaster recovery. In the cloud, we're really talking
about Disaster Recovery as a Service, otherwise called DRaaS. Now with traditional on-premises disaster recovery,
otherwise simply called DR, we have things like off-site backups to protect
data. Failover clustering, so that if we've got a critical app running on a
host that fails, that app can then run on another host within the cluster, and
in some cases, with zero downtime. We then have machine imaging that we can
then use to quickly get a system back up and running in the event of a
disaster. And then on the facility side, we've got power generators in case the
power grid goes down. There are battery-powered lights and so on. Now that's
with a traditional on-premises disaster recovery mechanism.
Now in
the cloud, with Disaster Recovery as a Service or DRaaS, we have a much smaller capital expenditure than traditional
disaster recovery. Actually there's no capital expenditure really because in
the cloud, we have an ongoing operational expense instead. So that means
dealing with things like data backups, maybe from on-premises, to the cloud,
and also long-term data archiving to the cloud. We might even have a hybrid
cloud solution where we extend our on-premises workloads into the Google
Cloud. So often that might be done in the case of disaster recovery so that if
we have a problem with an on-premises IT workload, it can failover into the
cloud as required.
So
Disaster Recovery as a Service then will vary from one public cloud provider
to the next. But generally speaking, it allows us to utilize cloud provider
infrastructure in case we have a failure of some kind with our on-premises
environment.
Summary
The public cloud is available to all over the
Internet. It might require signing up for an account, but potentially anyone
over the Internet, individual or organization, can sign up for public cloud
computing services.
A private cloud is different because it has all
of the cloud characteristics that a public cloud would have. Things like rapid
elasticity, self-provisioning, metered usage and so on. However, it's all
under the control of a single organization. So it's an organization's
infrastructure on-premises. They've paid for it, they're responsible for it,
and they manage it. Hence, it's a private cloud. But take note that just
because you're using virtualization on-premises does not mean you have a
cloud. You have to meet the cloud characteristics of metered usage,
self-provisioning, rapid elasticity, broad access, and so on.
A community cloud means that we've got the
infrastructure available but really for special needs, so organizations that
have similar computing requirements. Often it's usually in the case of
isolation and security, and maybe even certain types of connections or certain
app availability. Those similar computing requirements are best served through
what's called a community cloud.
A hybrid cloud uses on-premises as well as cloud
resources in the public cloud environment. So for example, we might have a VPN
link between our on-premises network or data center and the Google Cloud VPC
or a virtual network in the Google Cloud. We might have on-premises clients
that access Google Cloud virtual machine instances, that might also happen
through that VPN link. Virtual machines
are also known as VMs. Now, in some cases a hybrid cloud could be a
temporary solution during cloud adoption and the migration of data and IT
workloads from on-premises to the cloud. But also a hybrid cloud could also be
a more longer term solution for cloud bursting,
which means that we have depleted our on-premises IT resources and now need to
use public cloud resources. It might also be used for high availability where
we use it as a disaster recovery solution
if we have a failed workload on-premises we might have it configured to fail
over to the cloud.
Software as a service or SaaS is often in the
form of end-user productivity software where the cloud provider is responsible
for the maintenance of this software. The installation, the general
configuration, updating and so on. So in the Google Cloud we might use
software as a service offerings such as G Suite, or even using Gmail.
Infrastructure as a service, or IaaS, is
really based on IT infrastructure components. So that would be things like
virtual machines, or cloud storage, or cloud network configurations such as
VPCs, or IP address ranges we configure for certain VPC subnets, or firewall
configurations, or VPN connections, and so on. So in the case of Google Cloud,
infrastructure as a service would include virtual machine instances, VPCs,
which are virtual networks in the cloud and Google Cloud Storage offering.
Characteristics of cloud computing.
What makes a cloud a cloud? 5 defining
characteristics are,
This excerpt is from
'The Economics of Cloud Computing:
An Overview For Decision Makers,' by Bill Williams. Coming in July from Cisco
Press (ISBN: 1587143062).
With
traditional IT computing, we have an on-premises computing environment, which
requires a capital investment in hardware. We then have to power the
equipment. We have to assure that we have proper HVAC, that's Heating,
Ventilation, and Air Conditioning to make sure that equipment runs smoothly.
Then we have to license software. After we acquire the software, it needs to
be installed, configured, and maintained, including upgrades and even updates.
Now in
the cloud, we're talking about hosted IT services on the Google Cloud
Platform, that's on Google's equipment in their data centers around the world.
That includes servers, storage, databases, web apps, and so on. And just like
a utility, such as electricity, with cloud computing we pay for the services
that we use. There's also self-provisioning on demand, where we can either
programmatically or from a command line or using a GUI, we can provision cloud
resources or deprovision them if we need less. Also, the services are all
accessible over a network such as the Internet. But we should always be
careful with cloud computing to avoid vendor lock-in. And we want to make sure
that means that we have the ability to export data to standard formats or to
migrate data to other providers should we have a need in the future.
In the
cloud, there is often shared responsibility of taking care of the IT workload
running in the cloud. But the degree of that responsibility depends on the
specific cloud service that we're talking about. For example, if we're talking
about a hosted database in the cloud, maybe the underlying virtual machine
operating system and updates are taken care of by the cloud provider. So more
responsibility does, however, mean more control. Hardware is the provider's responsibility, whether we're
talking about network routers and switches, storage arrays, physical
hypervisor servers, and so on. The software, in some cases, will be the
provider's responsibility, such as in the case of something like Gmail.
However, the subscriber has some responsibility in terms of using it and
configuring it to serve their business needs. So it's a little bit of both, in
that case, with software.
When it
comes to the creation of users and groups, and in the Google Cloud Platform
that's called IAM, identity access and management, that's the subscriber's
responsibility. It's not Google's responsibility. Just like the creation and
management of data is the subscriber's responsibility. Security, in some
cases, can be the subscriber's responsibility. But the provider also has a
role in that, in making sure that their data centers, the equipment, and the
staff are all falling under the appropriate security recommendations to run a
proper IT cloud environment. The network connection, of course, is the
subscriber's responsibility. And that's why often it pays to have redundant
links, ideally through different Internet service providers, to assure that we
can get to the Google Cloud.
Google
manages the physical data center facilities that house all of the racks of
equipment. So they deal with things like hardware physical security, such as
locking up racks of equipment. They deal with the physical network
infrastructure links and the virtualization infrastructure, which really means
having the hypervisors that can run the virtual machines on them.
The
subscriber has a responsibility with the virtual machine instances that they
might deploy. [Video description begins] Virtual
machine is also referred to as VM. [Video description ends] How
they configure and create policies, such as how to control firewall rules and
also how to control data retention. That is under the subscriber's
responsibility, in this case. Dealing with user credentials, whether for example,
multi-factor authentication is configured, is the subscriber's responsibility.
Such as whether data at rest or data in transit is encrypted. And also what
type of data stores are being used to hold cloud-based storage information.
The cloud
really has a number of characteristics, such as resource pooling, having all
of these resources available for cloud consumers. Broad network access allows
access over the network using any type of device. The network in this case
being the Internet. Cloud computing also has a characteristic of having
on-demand self-provisioning. Usually, that's done through a web interface. So,
for example, if a client requires more storage, they can simply click a few
buttons and it's done very quickly. It's a measured or metered service, and
that's how we pay for our usage of cloud computing. The rapid elasticity is a
result of both resource pooling and self-provisioning, where we can rapidly,
in our example, deploy more storage or even more virtual machines as required.
