Choosing the Right Cloud Compute Option

This was something that I’ve had a definite interest in. Simply put, there is an intimidatingly large amount of compute options in the cloud to choose from. When you’re looking to migrate into the cloud or building a greenfield environment, what should you consider, and why.

Defining Compute in terms of IT.

Let’s first define what compute is from an operational IT perspective. Compute is a transitive verb that is defined as deriving something by means of calculation. The definition is straight-forward, accurate and definitely overly simplified from the information technology perspective. By that definition, everything in the information technology stack is a compute device. However, when you speak to an IT professional and mention the word compute they instinctively know, I would hope, that the conversation is focused on a specific type of device.

To that end, I would define it as any computing device that can be configured perform multiple related or unrelated computational tasks. Let’s take your basic, Windows or Linux server. By default, these devices do not serve any specific purpose outside of a user desktop. However, with little modification and effort, theoretically, they can be configured to perform a specific of multiple related or unrelated tasks. A router for example is purposely built to route layer 3 traffic. A server can also be configured to route traffic, but additionally it can provide mail, storage and database services. There are a several reasons why this would be a terrible idea, but it’s not impossible.

Choosing the appropriate computing resource

Now that we’ve established a baseline definition, let’s start discussing cloud computing options. Cloud service providers (CSP) provide a variety of ways to implement solutions in the cloud. Most major cloud providers offer their version of virtual machines, container services, and web apps. All of which will provide a viable migration path into the cloud, but there are pluses and minuses to consider with each offering.

Will most likely be the easiest path to migration. Architecturally speaking most IT organizations are very familiar with type-1 hypervisor architecture.  Additionally, services such as Azure VMware solutions (AVS) and VMware Cloud on AWS allow organizations to extend their existing VMware presence into the cloud.  This reduces the requirement to retrain their current virtualization teams. 

Virtual Machine Pros

• The learning curve is very low.
• Most CSP offer appliances that can be deployed within your environment that will analyze your infrastructure and provide the equivalent VM sizes necessary to support a similar workload in the cloud.
• More control over the environment.
• Dedicated hosts are also available.

Virtual Machine Cons

• Virtual Machines will most likely be the costliest solution of the bunch
• High Maintenance Requirements – cloud based virtual machines still require the same maintenance tasks
• Most operational maintenance tasks will still need to be performed.

Virtual Machine Use Cases

• If you require absolute control over the hosting operating system
• The workload is supported on a legacy or outdated programming languages and/or frameworks.
• If there are agents or 3rd party security mechanisms not native to PaaS or SaaS offerings that must be installed dictated by regulatory, compliance or organizational policy.

Additional Thoughts

Although virtual machines provide a relatively easy way to migrate existing workloads into the cloud, it doesn’t necessarily mean it would be the appropriate path. When considering a migration into the cloud, consider modernizing your infrastructure. If virtual machines are to be leveraged, then consider reserving compute or spot instances.  

Containers are next level virtualization. It decouples the application from the hosting operating system. This extraction layer introduces new possibilities for IT workloads.  More organizations are considering containers as it adds additional scaling and decoupling capabilities to their infrastructure.  Additional, many orgs leverage containers within their virtualization environment. 

Container Pros

• Containers much lighter than a virtual machine as it only contains the dependency of the designated application. This also makes them very portable.
• The container ecosystem is quite robust.
• Public registries will contain popular pre-packaged applications such as web servers, database servers, DNS server and more.
• Containers can respond to performance fluctuations much fast than standard virtual machines.
• Ease of maintenance – containers can be destroyed and redeployed within minimal downtime.
• Orchestration adds vertical and horizontal event-driven scaling capabilities.

Container Cons

• Learning curve is probably the biggest hurdle to leveraging containers. Many established IT departments are very comfortable with traditional virtualization that containers have not been prioritized.
• Although portal, containers are still bound to a specific operating system. In other words, a windows-based container will not run on a Linux based host.
• Cost still needs to be closely monitored since containers are still supported by virtual compute.

Container Use Cases

• When you want to run applications without managing the operating system.
• Your application requires very responsive horizontal or vertical scaling
• You want to decouple the applications from the operating system.

Additional Thoughts

Containers are a great alternative to virtual machines. They are extremely portable and very performant. By decoupling the application from the operating system tasks such as Scaling in/out, software updates, and high availability operations take less time to complete.  Cost saving opportunities are similar to virtual machines.

Web app are purposely built to for hosting web applications, REST apis and mobile back-ends. Web apps support most popular coding languages such as .Net, Java, Ruby, Node.js, PHP, and Pythons.

Web App Pros

• Multiple supported languages and frameworks
• Operating System and Frameworks maintenance are managed by the cloud provider
• DevOps integration ready
• Native Blue / Green deployment architecture
• Dev / Test environment can be spun up easily and cheaply.
• Affordable alternative to traditional compute running web services.

Web App Cons

• Similar to containers, the operations team may require training
• Aging Home grown application may not be supported by modern languages and frameworks.
• Management of web apps is primarily performed via source code management tools such as visual studio code, visual studio and Atom IO.
• Lack of a GUI interface may be challenging for some.

Web App Use Cases

• You require DevOps Integration.
• Migrating or rearchitecting on-premises web front end users
• Need an environment that is Blue / Green Deployment capable.

That’s just a small subset of computing offerings in the cloud that this article doesn’t touch.  In addition to the information provided above, I’ve created a workflow diagram that attempts to pictorially aid you in your decision quest. It’s something that I know isn’t perfected, but I plan to keep updating it.  I leveraged a diagram created by Microsoft and expanded it to highlight related solutions from AWS and Google Cloud.