There are different types of cloud computing and different opinions on how they should be categorized. Personally, I find four types most logical:
- Private: Everything hosted internally, in our own data center.
- Hosted: Something between a private and public cloud; the service provider creates a separate environment in their data center and offers us an isolated cloud for our use only.
- Public: The service provider offers a service available to everyone—publicly available. There is still tenant isolation but we'll talk about this later.
- Hybrid: A combination of private and public cloud. Some services are used in the public cloud but some stay in our local data center with direct connection between two or more environments. From my experience, this is the most common form of cloud computing. Again, we'll explain more about this later.
In the private cloud, all resources are located on-premises, in our local data center, and no internet access is needed to access resources. The internet and resources are accessed separately as shown in the following diagram. Building your own private cloud previously required large-scale investment, both materially and in terms of knowledge. First, you needed space and needed to consider other elements like cooling and power. Then, you needed to invest in hardware like firewalls, routers, network switches, servers, and storage.
You needed licenses for a virtualization layer, operating system licenses for virtual machines, and then licenses for different kinds of software. In the end, all material investment was in vain if you didn't have the right people to set everything up and maintain it in the years to come. Once everything was in place and you had your private cloud running, it required new investment every few years as you needed new versions of software (virtualization, operating systems, and other software) and hardware needed to be replaced as well:
The hosted cloud came as the first step in the transition from the private cloud to the public. As creating and maintaining your own private cloud demanded large-scale investment, some companies took advantage and started offering services where you could rent part of their data center and use it as your own private cloud. They specialized in this kind of offer; it was cheaper for them to buy hardware and software as vendors offered discounts on mass purchases. So, creating an environment in the hosted cloud was cheaper then creating an identical environment in the private cloud.
There is also the question of upfront investment; using the private cloud requires that all hardware and most software licenses be paid for upfront, so many companies have decided to use the hosted cloud as they don't have to make an upfront investment but monthly or yearly subscriptions instead. Also, it's easier for data centers to provide experts to maintain systems as a single expert can take care of multiple customer environments. For the private cloud, you need a network engineer, a storage specialist, a virtualization specialist, and so on, and this is for a single data center.
In the case of a hosted cloud, all personnel are still required but a single specialist can set up and maintain environments for multiple customers and the price of maintenance is lower than for a private cloud. Note that to access the hosted cloud, usually some sort of Virtual Private Network (VPN), either site-to-site or point-to-site, is required. We access resources located outside our own network and located in another hosted network as shown in the following diagram:
In the next step of cloud evolution, the public cloud emerged. Large service providers offered large amounts of resources for on-demand use. Similar to the hosted cloud, resources you used were still outside your local infrastructure and hosted by service providers who specialized in this kind of offer.
There are two key differences. The first difference is that in a hosted data center the amount of resources available I usually predetermined and to get more resources you need to wait for new resources to be configured, if this becomes available at all. In the public cloud, providers have a large amount of resources available for on-demand requests and you can get them whenever you need them. You can create any kind and any amount of resources when needed. All you need is to create a subscription and access to the internet to start deploying. This also means you have highly scalable environments and you are not limited by the initial size of the resources created. For example, if you create a virtual machine with four CPUs and 16 GB of RAM and find out over time that the virtual machine can't handle the workload you have, you don't need to create a new virtual machine; you can use a scale-up option to change size. Scaling up is explained later in more detail. This works other way around: If you find out that the size of the virtual machine initially created is too large for your workload, you don't need to keep that size and pay for something you don't need. Simply scaling down will do the trick. In this case, we access resources over the internet as shown in the following diagram:
The other difference between a hosted cloud and a private cloud is pricing. In a hosted cloud, you would get an agreed amount of resources and pay a monthly or yearly subscription no matter in what capacity these resources are used, whether 10% or 100%. In the public cloud, pricing is based on usage and the model of payment is such that you pay for only things that are used. So, in the public cloud, if you create a virtual machine, you will be paying for that virtual machine for the time you actually use it. If you stop or delete this virtual machine, you will not be paying for it. The payment model is different for different cloud providers and can vary by per-day, per-hour, or per-minute usage. As we'll talk about Microsoft Azure, it's important to mention that Azure is using a per-minute billing system. So, for example, if you create a virtual machine in Microsoft Azure and delete it after 12 days, 11 hours and 13 minutes, the amount you pay will be calculated for that exact amount of time. In a per-hour billing system, you would pay for 12 days and 12 hours. In a per-day billing system, you would pay for 13 days.
Another difference is multitenancy. Even the public cloud is available to everyone; creating your own subscription creates your own tenant. By using special fabric, this tenant separates your resources from other tenants, and resources created in that tenant are available only to people with access to that specific tenant.
To sum up, the key concepts of the public cloud are:
- Access over the internet
- Multitenancy
- Resource pooling
- On-demand consumption
- Highly scalable
The term cloud or public cloud wasn't forged with modern IT but the term started in the 1960s with the concept of resources being time shared. The concept did evolve in the 1990s with the private cloud. However, the cloud did evolve and shift further to a modern form in the 2000s.
It all started with Amazon Web Services, a subsidiary of Amazon, when they released their Elastic Cloud Compute (EC2) in 2006. Google followed with Google App Engine in 2008. Microsoft announced their version of the cloud in October 2008 and it was publicly available in February 2010. Other service providers followed and many companies such as IBM or Oracle have their own public cloud offering. Looking at market shares and the pace at which they evolve, we can put only two cloud providers at the top of this list: Amazon Web Services and Microsoft.
