Choosing an AWS EC2 Instance: Where to Start

Choosing the best IT infrastructure depends on a host of factors, and it’s getting more complex by the day

Many enterprises are looking to the cloud as a key part of their modernization puzzle, but as they explore their options, it’s important to remember: The cloud is not a monolith. There is no single cloud, and there is no single way to configure the silicon that powers it.

Instead, the cloud contains a multitude of hardware configurations assembled for a host of specific needs. IT professionals tasked with moving workloads from mission-critical platforms like IBM Power or Z to AWS’ Amazon Elastic Compute Cloud (EC2) service have numerous decisions to make, and one of the first is choosing the right instance.

What’s an EC2 Instance?

EC2 instances come in pre-built configurations tailored to specific workload types, with varying combinations of CPU, memory, storage and networking capacity. The problem (and advantage) is that they come in more than 1,000 configurations. Some products, such as Amazon Lightsail, are specifically designed for non-developers and have a more intuitive management console than others. Options include bundles designed to meet the needs of startups, as well as instances preconfigured with Linux or Windows application stacks so that organizations can get off to a fast start.

Having so many choices means that AWS can serve organizations from small businesses to global corporations in every industry, including highly regulated sectors like healthcare and finance. But it can also make choosing the right instance for the right jobs feel like being lost in the jungle.

Start With General Purpose Instances, Then Adjust

Before anything, it’s important to think carefully about what you’re trying to accomplish. A few of the reasons customers have moved from a mainframe platform to AWS are performance, cost savings and unlocking innovation, says Martin Yip, head of product marketing in compute, enterprise workloads and migration and modernization at AWS.

Whatever your needs may be, Yip urges against paralysis by analysis.

“What we tell customers is to just get started,” he says, noting that migrating from one instance to another is relatively simple. “I advise customers to start with general purpose instances, then based on testing and benchmarking—if they need more memory, or if the application is graphics-intensive, or they’re doing AI/ML—then adjust accordingly.”

One of the reasons customers tell Yip they’d like to migrate to AWS is that cloud-native platforms offer benefits like scalability, elasticity and cost savings. Companies can plan for things like spikes in usage without having to over-provision. Workloads can scale up with a larger server or scale out by automatically adding more servers of the same type—and scale back down when the traffic subsides.

Tools like AWS Compute Optimizer help customers identify underprovisioned and overprovisioned resources to make sure they’re optimizing for both performance and price. If you begin with a general-purpose instance and run the workloads, patterns in demand emerge.

Compute Optimizer monitors the workload, recognizes the patterns and makes recommendations regarding instances. You may find a memory- or storage-optimized instance to be a better option than a general-purpose instance. Since moving from one instance to another is not difficult, making such adjustments means your operations become more efficient over time.

General EC2 Options

Even for general-purpose instances—like those comprising AWS’ M series, currently in its seventh generation—the choices are plentiful. AWS Companies want to tailor their workload to get the best performance for the lowest cost. The issue is that the millions of applications in existence all have dramatically different characteristics, Yip points out.

“Some workloads, like HPC applications, are compute intensive, while others, like SAP Hana, are memory intensive,” he says. “There are endless types of applications out there and new ones being developed all the time.”

Some considerations in choosing a general purpose instance:

• Processor type – These include AWS’ own ARM-based flagship Graviton chip, x86-based processors from Intel and AMD and specialized accelerators for AI training, such as AWS’ own Trainium and Inferentia as well as chips from NVIDIA.
• Price-performance ratio – Per dollar spent, certain instances deliver as much as a 25% better performance than others. For instance Graviton-based instances perform significantly better than x86 equivalents.
• Memory-to-compute ratio – Depending on the organization’s needs, some general-purpose instances offer more memory per vCPU compared to others.
• Network and storage performance – For applications that require fast networking and low latency, AWS offers networking-optimized general-purpose instances.

As enterprises’ requirements change, they can tweak EC2 instances on the fly. Since moving from one instance to another is not terribly labor intensive, testing and making adjustments as needed is the best way to find the instance best suited to the needs at hand.

Understanding EC2 Instance Naming Conventions

With so many possible permutations, understanding the naming conventions of EC2 instances may be helpful. The name of each EC2 instance indicates its series, generation, options (processor and optimizations) and size.

For example, here’s how the instance type M6in.xsmall breaks down:

“M” – General-purpose series
“6” – Sixth generation
“i” – Intel processor
“n” – Network- and EBS-optimized
“xsmall” – Extra small

Instance series are organized under specializations including:

M – General purpose
C – Compute optimized
R – Memory optimized
I – Storage optimized
T – Burstable
G – Graphics/GPU

Assessing EC2 Viability

It’s not always simple and straightforward to move from a mainframe architecture to a cloud-native service. For example, a COBOL-based application that’s been running for so long that the source code is no longer accessible may not be an easy “lift and shift.” That doesn’t mean it’s impossible. AWS offers tools that can be helpful, but “it might require some reimagining of what the application can do,” Yip says.

Migration feasibility isn’t the only factor in whether to move a workload to the cloud. A common concern among IT professionals considering transitioning to a cloud infrastructure is security. For that reason, many companies have retained on-premises platforms for their most valuable and protected workloads.

Whether in the cloud, on-premises or a hybrid configuration, the modernization journey comes with many hurdles. But choosing an EC2 instance type doesn’t have to be a paralyzing task. Think of it as picking out clothes for your workloads.

“You can think of instances as pre-defined T-shirt sizes,” Yip says. “Once you choose the style, you just have to choose the right size for your workload.”

---