Cloud scalability vs. elasticity: What's the difference?

Cloud scalability vs. elasticity

The National Institute of Standards and Technology (NIST) includes rapid elasticity as an essential characteristic of its definition of cloud computing:

“Rapid elasticity. Capabilities can be elastically provisioned and released, in some cases automatically, to scale rapidly outward and inward commensurate with demand. To the consumer, the capabilities available for provisioning often appear to be unlimited and can be appropriated in any quantity at any time.”

One key aspect of this definition is that it points out that elasticity relies on the function of scale. To be elastic, a system must scale.

So let’s start there. What is scalability?

Companies that need scalability calculate the increased resources they need, and plan for peak demand by adding to existing infrastructure with those resources.

Cloud scalability is not hampered by a company’s physical hardware resources. Whereas the physical nature of hardware made scaling a slower process, in the cloud, scalability is much more efficient and effective.

Cloud scaling allows for automation, which helps quickly scale systems to meet demand. Cloud scaling is generally performed in private cloud environments. For the most part, cloud scaling is either scaling up or scaling out.

Scaling up

Scaling up is also known as vertical scaling.

In the case of needing more processing power, a company moves from a smaller resource to a larger one that is more performant, such as moving from a virtual server with two cores to one that has three. While cloud scaling is automated and fast, often on the order of seconds for new containers and up to minutes for VMs, to bring up new hardware can take some time.

Containerized applications are a good fit for scaling out. In this type of scalability, virtual machines are spun up as needed to create new nodes that run containerized microservices. Think of it as adding the same type of services already running to spread out the workload and maintain high performance.

Scaling out

Scaling out is also known as scaling horizontally. This is used by companies that need high availability and little or no downtime with applications.

In horizontal scaling, companies add more of an equivalent function, to apportion the workload across multiple servers, keeping performance high and increasing available storage.

Containerized applications are a good fit for scaling out. In this type of scalability, virtual machines are spun up as needed to create new nodes that run containerized microservices. Think of it as adding the same type of services already running to spread out the workload and maintain high performance.

With scale, it’s possible to overprovision and pay for computing resources that are not necessary and stand idle. It’s also possible to under provision and suffer outages from having too little capacity for the workload. For these reasons, and others, elastic cloud systems are the right fit for some companies.

These are essential because they deliver efficiency while keeping performance high in highly variable situations. Companies that experience frequent, short-term spikes in workload demand are good candidates for elastic systems.

In elastic systems, resources are neither idle nor missing; instead, they are available. Using monitoring and automation, elastic systems respond dynamically to increasing or decreasing according to workload demands by making resources available in real time and removing them when they’re not needed. Elasticity goes hand-in-hand with rapid response to dynamic environments.

Companies that seek elastic computing are often smaller and use public clouds to achieve the elastic solutions they need. Under the elastic model, companies can add all the resources they need to meet peak demand — for example, for black Friday retail situations — without experiencing any downtime or significant delays. Companies can add all the necessary resources, such as RAM, CPU processing power, and bandwidth.

With scale, you add resources and keep them whether you use them or not; with elasticity, you have a base state and then use more of what you need, when you need it, and return to a ‘normal’ state otherwise.

Pricing models

Scalable systems and elastic systems both use a pay-as-you-go pricing model that helps companies achieve efficiencies in price and performance of their systems. For elastic scaling, there’s also a pay-as-you-grow aspect that denotes the added resource expansion for spikes, which, when they have passed, returns to the pay-as-you-go for use model.

How you choose to deploy has an effect as well. You may choose to deploy on an Infrastructure-as-a-Service (IaaS) platform, or as a platform-as-a-Service (PaaS) deployment. Each cloud service provider, such Microsoft Azure, Amazon AWS and Google Cloud, will have architectural decisions that affect how your deployment is priced.

Benefits of cloud scalability and elasticity

• Cost-efficiency: Cloud scalability enables companies to quickly have the systems they need and the compute power without the expense of purchasing equipment and setting it up. Since companies pay for only what they need and use, there is no waste on capacity. Other expenses such as storage and management resources are also avoided.
• Service availability: Companies can plan to meet their growing needs and can use elastic models to meet spikes, all without service failures.
• Agility: In contrast to the effort required for physical systems set up in cloud systems, administrators can quickly set up VMs to meet resource needs, which allows IT staff to devote more time to more value-added work.

Want to learn more about cloud and cloud development?

Take a look at our Cloud Application Development Guide to learn more about your cloud options, and how to successfully embrace the cloud and cloud-native development as part of your digital transformation.

Frequently asked questions