In This Chapter
The public cloud rules all!
"You'll move my apps out of my private cloud only after you pry them out of my cold, dead hands!"
It seems that people have chosen sides in the public cloud- versus-private cloud debate, and both sides have good arguments.
Instead of looking at this is a public-versus-private debate, it's useful to consider the individual outcomes that each side is trying to achieve and then architect around those.
Perhaps one of the biggest challenges facing IT departments is friction. As Chapter 1 shows, today's IT departments face competing needs. IT must support legacy applications that require constant uptime and low risk. At the same time, business leaders want IT capabilities that allow them to test new business models and roll out new services quickly.
Because of the need to ensure ongoing availability and performance levels for legacy and next-gen applications, fooling around with legacy local infrastructure is often frowned upon by the business, which often doesn't appreciate or understand the level of effort that it takes to keep systems going.
That's where the public cloud comes in.
But you need to understand the why behind all of this first. IT has generally done a good job supporting the business by running tools for enterprise resource planning (ERP), collaboration, and so on.
But so much more is at stake.
With the right technology environment and the right mindset, IT can move from a behind-the-scenes supporting role intended to reduce expenses into a revenue-driving role. IT can become an active enabler of the business.
Moving from a support mindset to an enabler mindset takes lots of rethinking of IT infrastructure and services.
Consider your legacy datacenter environment. You probably have a bunch of servers, a monolithic storage environment based on SAN technology, and various networks connecting it all together.
On the economics side, you likely prepay for all your resources — and you pay full price even if you use only onehalf of the resources during their life cycle.
This purchasing practice is far from efficient. Most organizations overbuy and overprovision resources so they don't run out mid-cycle. This approach provides good insurance, but it means money is being left on the table.
The public cloud offers enterprises a new consumption model for IT resources. Fractional consumption enables pay-as-you-grow economics, which allows you to buy resources on demand. This is one of the primary economic drivers behind cloud technology. CFOs and other financial decision makers love being able to buy only what is needed. However, fractional consumption in cloud also occurs as you spin down. That is, as your business needs change, you can reduce your resource utilization and, in turn, your monthly payment.
The new model shifts the cost paradigm from a CapEx-intensive activity to one based almost solely on OpEx. Cloud enables an OpEx focus because you have no need to buy a bunch of hardware up front. The upfront CapEx-intensive purchasing paradigm is one that plagues legacy IT.
Perhaps one of the biggest downsides of legacy infrastructure is the time needed to deploy new resources, which drives up project latency. The result is frustration from business users who have become accustomed to on-demand services.
Public cloud services enable this kind of agile deployment. As you or your business users want more services, your imagination is the limit.
You don't have to
Instead, with cloud services, you can spin up infrastructure and platform resources on demand with a single click to build, test, and deploy applications. Building blocks — databases, message queues, and so on — are available to deploy new applications with zero wait time. This innovation drastically reduces application development time and can massively increase time-to-value for new initiatives.
Nothing strikes fear into a CIO's heart like having to deal with the replacement cycle of the IT infrastructure.
Refresh cycles can be expensive and risky, and they often require downtime. Replacing monolithic structures such as a SAN requires big capital expenditures and off-hours work.
With the cloud, you have no operational overhead in deploying, managing, and refreshing infrastructure. That's the provider's problem. As a user, you simply key in a credit card number and, on a management console, instantly provision resources.
Security and trust in the cloud
Although still not 100 percent, people have far more trust and faith in the security of the cloud than they did in the past. Businesses are even starting to trust public cloud services for security, governance, and risk management. Organizations can focus on innovation and services without diverting resources to maintain security of the on-premises infrastructure stack.
With all these great benefits of the public cloud, you might be wondering why you haven't walked into your datacenter and set it ablaze. Well, for all of the good, the public cloud is not a panacea. It still has challenges to overcome.
For many organizations, operating a private cloud makes far more sense.
The term private cloud is shock- Further, some businesses are tryingly misused. Many believe that ing to build private clouds using getting close to 100 percent virtu- OpenStack/vCloud/Azure and a alized means they've successfully matching virtualization solution, but deployed their private cloud and they still relying on traditional scale-up can now enter the annals of cloud storage and three-tier architecture. history for their accomplishment. Not Although you get self-service proso fast! The word cloud carries some visioning and scale-out computing, implicit assumptions about architec- you still deal with the complex infrature, workload manageability, auto- structure life cycle of infrastructure. mation, and user self-service. Only You end up with silos for different after you've successfully deployed applications (for example, all-flash an infrastructure that has the right arrays and bare metal for high perarchitecture, at least some level of formance, virtualization with hybrid provisioning automation and user storage for VDI, and so on). This self-service can you start to con- setup may look like a cloud on the sider it a private cloud. Without surface but doesn't deliver many of those features, you're nothing more the benefits that public clouds like than a highly virtualized datacenter. AWS enjoy. Bear that in mind as you Virtualization is only one component read the rest of this book. of the private cloud.
The public cloud, for all its benefits, has limitations. While the public cloud is a viable, cost-effective option for elastic workloads where demand is highly variable or unpredictable, it is not as cost-effective as on-premises infrastructure for more predictable workloads.
In fact, managing predictable workloads is where IT shines. We've been doing that for decades and we do it really well. On the economic front, it's often less expensive to implement and maintain your own environment for predictable workloads than it is to pay monthly expenses for cloud infrastructure.
Here's why: Your predictable workloads often include such applications as ERPs, end-user productivity tools, and business intelligence and analytics suites. These applications often require consistently high levels of performance and, particularly for applications that use a legacy client/server model, the network connectivity between the server and the connecting clients must be very low latency and very high bandwidth.
With public cloud providers, you pay far more, for example, for all-flash storage in a public cloud environment than for spinning disk. For any applications that require consistently high levels of CPU, you pay monthly for that peak usage. On the network front, you pay far more for a very high bandwidth, low latency connection to the public cloud provider than you would pay to implement such a network in your own environment.
Consider this scenario: Pretend for a minute that each of your travels for work and for pleasure equates to an enterprise workload use case. So, that trip you took to the Caribbean might represent a VDI deployment. The business trip you took to London might stand for a CRM deployment.
As you undertake each of these journeys, you need transportation, which is analogous to infrastructure. Now, as you arrive at the destination airport for each of these trips, do you make your way to an auto dealership and buy a brand new car to use while you're there?
Of course not! Economically, that would be ludicrous and wasteful. You'd also catch the attention of your finance department, who would laugh at your audacity as security escorts you out of the building.
Instead, when you're at home with your predictable travel needs, you likely own a car, or maybe you lease one so that you can replace it every three years. When you travel, or you have unpredictable travel needs, you typically rent a car for the time you need it.
In essence, you're making an ownership decision based on each individual use case. Likewise, businesses want to balance owning and renting infrastructure, choosing between private (owned) and public (rented) infrastructure depending on application workload characteristics. In some situations, renting makes sense. In other cases, owning is a better choice.
Public cloud providers benefit from economies of scale in terms of lower costs, operational efficiencies through automation, and appropriate resource sharing, which they pass along as cost savings to customers.
Private clouds are better suited to predictable, well-established workloads. For these workloads, you decide that owning the infrastructure is a better economic decision.
Before I move on, you should consider one more possibility. Suppose you travel all the time to the Caribbean. In this case you might want to own a car at that location, even though it isn't your primary residence. Owning may be less expensive than continually renting cars. Translating this example to the cloud, many organizations are discovering that, once they've moved a certain amount of workload to the cloud, the economics begin to break down. A time comes when, regardless of the kind of workload, pulling some of the workloads back to the private datacenter makes the most sense.
At a certain point, the economies of scale tip back to favor on-premises private cloud environments. Make sure you have a deep understanding of your organization's needs so you know which workloads to run in which location.
Chapter 1 discusses situations in which people in certain regions of the world want to avoid having their data reside in certain other regions of the world. Public cloud providers have begun to address this issue by deploying new datacenters in new regions, but the need to maintain high-level economies of scale can make doing so somewhat challenging.
In addition, economies of scale stop providers from offering a more differentiated experience for individual customers and applications. One of the biggest roadblocks to public cloud adoption is that customers want control over where the data sits and how it is accessed. This level of control is not always possible with the public cloud.
With public cloud, customers may not always know exactly where their data resides. Is it in their state or even in their country? With a patchwork of data security and privacy laws worldwide, not knowing where data resides can create compliance and security issues for customers. Some organizations desire all sensitive data to be under their direct control, effectively eliminating public cloud as a locale.
However, with the private cloud, data locality and proximity are 100 percent in your control. You get to decide exactly where data resides and how close it sits to end-users and applications.
Although public cloud providers have become far more adept at offering granular service level agreements (SLAs), nothing compares to what you achieve with your own infrastructure.
When you're considering SLAs, be sure to consider these two points:
Today's businesses want carefully tailored performance and availability SLAs for their mission-critical applications, but they also want some choice — soft SLAs for less critical applications, and stringent SLAs for business critical applications. For example, for a particularly important application, the business may require that at least three copies of data be maintained separately — public cloud services cannot offer this level of granular control.
Once upon a time, not very long ago, business users simply accepted whatever IT gave them, whether that service was fantastic or poor.
Not any more.
These days, businesses want to use the public cloud where appropriate — for example, for backup, disaster recovery, and applications with highly unpredictable IT requirements — and switch between private and public easily.
They want three things:
Here's a quick recap of what today's businesses demand.
Businesses want the public cloud for
But they still want
Today, enterprise IT offers control, which you need for many applications, but when business users need frictionless agility and ease of use, they are going to the cloud. The two worlds are segmented, and bridging them is difficult.
The vision for the future is to have hybrid environments where the boundary between private and public disappears. You may have an application that has some parts/components in the cloud (for example, deep storage) and others on-premises. You can also have situations where the cloud is used as a backup/DR target while the production environment is on-premises. Finally, an application can be on the public cloud early on in its life when demand is unpredictable, but as demand becomes more stable, it may be migrated back to the on-premises environment.
To meet all these requirements, you need a new paradigm for the enterprise datacenter that can deliver on both sets of requirements and provide a seamless experience between on-premises infrastructure and public cloud services.
That's where the enterprise cloud comes in.