Home > Articles

Learning AWS

  • Print
  • + Share This
This chapter is from the book

Trying to Define the Cloud

The roots of public cloud computing are not new; the public cloud providers Amazon Web Services and Microsoft Azure have been established for well over a decade with strong IaaS and PaaS offerings around the world. The Google Cloud Platform (GCP) and the IBM or Oracle Cloud are other viable alternatives. Gartner’s Magic Quadrant (www.gartner.com/en/research/methodologies/magic-quadrants-research) in Figure 1-1 shows four types of technology provider a company can align their goals and strategies with. In 2018, IaaS market penetration dominated two of those categories. Under the Leaders quadrant, Amazon Web Services led in that area, followed by Microsoft and then Google. Google also aligned closely to the Visionaries Quadrant. Alibaba Cloud, Oracle, and IBM fell in the Niche Players quadrant.


Figure 1-1 Top public cloud providers. Gartner, Magic Quadrant for Cloud Infrastructure as a Service, Worldwide, Dennis Smith et al., 23 May 2018. (Gartner Methodologies, Magic Quadrant, www.gartner.com/en/research/methodologies/magic-quadrants-research)1

When I started my career as a computer technician back in the 90s, most corporations that I supported used several computer-based services that were not located on premise. Accounting services were accessed through a fast (at the time) 1200 baud modem that was connected using one of those green-screened digital terminals. The serial cable threaded through the drop ceiling to connect the terminal was strong enough to pull a car.

A customer of mine at the time was utilizing a mainframe computer for accounting hosted locally in town. However, he couldn’t access his accounting services any time he liked; he had his allotted slice of processing time every Tuesday, and that was that. Payroll services were provided by another remote service called Automatic Data Processing, or ADP for short. Both service companies and their services are still around today. IBM is continuing to release versions of its z series mainframe, and ADP payroll services was one of the first software as a service (SaaS) companies but remains popular today.

In 2015, IBM bought a cloud provider based in Texas called SoftLayer and merged it into its public cloud offering, today called the IBM Cloud. The z mainframe has ended up being hosted in the IBM cloud providing hosted mainframe services; in April 2018, IBM announced it was launching what it called a “skinny mainframe” for cloud computing built around the IBM z 14 mainframe.

If you work for a bank or financial institution, IBM mainframes probably provide 50% of all your computing services. This could be great news for companies that don’t want to have a local mainframe environment to maintain.

Fifty years since the launch of the IBM mainframe, many companies’ mainframes are continuing to be relevant and are now part of the public cloud landscape.

The reality is that more than 90 of the world’s largest 100 banks, the top 10 insurance companies, a majority of the 25 largest retailers, and most of the world’s larger airlines still rely on mainframe computers from IBM.

If you didn’t use mainframes, you probably lived through the deployment cycle of Novell NetWare and Windows and Active Directory, and virtualization using VMware or Hyper-V. You likely have a private cloud in your own data centers. You may be wondering why your company is moving to the public cloud.

The reality these days is that it is expensive to build and maintain data centers. Certainly, building a data center is going to cost millions or billions of dollars. Maintaining an existing data center over the long term is expensive as well. Because of virtualization and the rise of the Internet as a useful communication medium, cloud services have replaced many local data centers and will continue to do so. Figuring out the capital costs of hosting your applications in the public cloud instead of running them in your own data center is sometimes categorized as renting instead of buying, as defined in Figure 1-2.


Figure 1-2 No long-term capital expenses

Operational expenses (OpEX) are all you pay for using cloud services. The capital expenditure (CapEX) of building a data center does not have to be borne by a single business. Now let’s be clear: operational expenses are still expensive. You might say to your boss, “I don’t need $800 million for data center construction, but I will need $2 million a year forever.”

The reality is that the cost of running and hosting your applications in the cloud is cheaper once you add in every expense; however, operating in the cloud is only cheaper if your services being hosted in the cloud are properly designed. Services and applications don’t run 24/7; they are turned off or reduced in size when they’re not needed. A concept that you may not yet be familiar with is automation. Public cloud providers use automated procedures to build, manage, monitor, and scale every cloud service. By the end of this book, you will understand how automation is the secret sauce for successful cloud deployments. Automated procedures will save you money and allow you to sleep at night.

Let’s start by defining the public cloud. The cloud is just a collection of data centers. There is no ownership from the customer’s point of view; the cloud provider owns the services, and you rent each service as required. You may be thinking that the cloud is all virtual resources, yet the AWS cloud can provide you bare-metal servers. If you want, Amazon will happily host your applications and databases on bare-metal servers hosted in its data centers. Of course, more commonly, AWS will offer you many virtual servers in well over 150 different sizes and designs. Amazon is also quite happy to allow you to continue to operate your on-premise data centers and coexist with cloud resources and services operating at AWS. Microsoft Azure will offer to sell you a copy of its complete Azure cloud operating system to install on your servers in your data centers. As you can see, it’s hard to define the public cloud these days other than as a massive collection of compute and storage resources hosted on a network stored in the collection of data centers accessible across the Internet, or by using private connections.

Anything that you host in the public cloud is using compute and storage resources to execute your software application. And anything that used to be a hardware device, such as a router, switch, or storage array, can be replaced by a third-party software appliance or an AWS-managed software service composed of virtual computers, storage, and networking components. This doesn’t mean that many companies aren’t still using hardware devices. Hardware devices such as routers and switches have incredible speed and can operate much faster in most cases than a software router and switch. But what happens if you can run hundreds or thousands of virtual machines in parallel performing the function of a hardware switch or hardware router device? Perhaps we don’t need any hardware devices at all. Most of the AWS-managed cloud services are hosted on virtual machines (defined as EC2 instances, or Elastic Cloud Compute instances), with massive CPU and RAM resources running in massive server farms with custom-designed applications, providing the storage arrays, networking services, load-balancing, and auto-scaling services that we depend on at AWS.

  • + Share This
  • 🔖 Save To Your Account