AWS Instance Types Comparison and When to Use Each
Share this on:
Cloud computing services play a crucial role in today’s business environment. AWS (Amazon Web Services) is a fundamental player in the game, being used by many companies wanting the support that cloud computing can give. Given the prominent role played by AWS in this sphere, any businessperson wishing to avail themselves of their services must have a sound grasp of what’s on offer. We’ll start by assessing what we mean by ‘instance’, in this instance.
Instances and AWS
The scalability and cost-savings that cloud computing can deliver, make it a must for a growing business. It is why it’s so widespread.
And when dealing with cloud computing providers, AWS is very popular – offering a wide range of instances for business use.
So, what’s an instance? Simply put, it’s a virtual machine you can use to run applications on AWS’s cloud computing platform. Because they’re virtual rather than physical, you can scale up or down at pace to match the fluctuations in the operational activity your business will probably experience over time.
Just as physical computers differ in terms of being best for different applications, instances can exhibit varying RAM size, CPU speed, storage characteristics, and networking resources. So, whatever your particular application, there will be an AWS instance perfect for it. However, their very heterogeneity can be a problem. There are over 50 of them, and to some, this can be a tricky field to pick.
Their naming doesn’t make things any easier. Try this one for size: R5ad.24xlarge. No, not much to get a handle on there. But, with a modicum of background data, you’ll understand these names in no time. They contain information on the instance family, generation, capability, and instance size.
As far as the preponderance of instances is concerned, they group rather handily into five broad families, which we’ll now explore.
If you need computing support for a business involved in a variety of non-specialist areas, then a general-purpose instance is what you’ll need to be looking at. They tend to feature a balance between performance and cost, and offer generally good CPU, RAM, and networking features.
A hosted VoIP system may well use a general-purpose instance. What is hosted VoIP? It’s cloud-based telephony commonly used in contact centers.
The general purpose subsets are as follows:
M (Generic) – these are a good, stable choice for applications such as small and medium databases.
T (Burstable) – these provide elasticity to cope with sudden rises in demand, ideal for particular website usages.
A – these are specifically for ARM-based servers and support enhanced networking.
Mac – these allow you to run macOS in the cloud.
Free to use image sourced from Unsplash
When you need full-on intensive computing processing power, compute-optimized instances will deliver. Note – these are for CPU-heavy applications. They don’t give as much in the way of memory or storage, so use them wisely. Examples of applications include video rendering, machine learning (such as SageMaker), and scientific modeling – all of which would also benefit from being run through a crowdtesting platform.
Types of compute-optimized instances comprise:
C – router, firewall, and network traffic applications
Hpc – seriously high-performance computing, powered by 3.6 GHz Epic 7003 processors. With this option, up to 96 virtual cores are at your disposal.
Memory instances deliver a high RAM factor, as well as high levels of input/output (I/O) memory rate. They prioritize this over CPU speed, so are suited to memory-intensive applications. Types comprise:
R – high levels of RAM (768 GB), ideal for real-time big data analytics.
X – even more memory-intensive (3.9TB).
Z – a combination of pretty good RAM (384GB) and 4.0GHz CPU, plus up to 1.8TB of SSD storage. Good choice for large enterprise database use, such as Lakehouse for Retail.
As you might expect from the name, these offer the best storage possibilities on AWS. Types of storage-optimized instances are as follows:
D – up to 48TB of HDD storage per instance makes this variety suitable for data warehousing.
I – good I/O memory performance combined with optimized storage, ideal for transactional databases.
H – good disk throughput and 16TB of HDD storage. Suitable for distributed file and network file systems, plus various data processing tasks.
Free to use image sourced from Unsplash
The use of hardware accelerators delivers a turbo boost, giving a sustained high performance for compute-intensive applications. Accelerated computing often uses powerful video adapters and graphical processing units.
P – with speeds of up to 400 Gbits/s these are the most powerful of the accelerated computing family of instances. Machine learning, speech recognition, and seismic analysis are common applications.
G – used in graphics-intensive applications, such as 3D visualization and video encoding.
F – offers customizable hardware acceleration, for tasks such as genomic research and real-time video processing.
DL1 – mainly used in training.
VT1 – used for real-time video transcoding.
So, now you have an idea of what’s out there and the sorts of uses to which they’ve been put. It’s not a bad idea to bear the following in mind when you’re deciding on which instance to go for, especially if you’re thinking about the links between virtual and physical in a hybrid integration platform.
Fixed v burstable performance
Is your requirement fairly static, or does it tend to vary from baseline performance over time? Most applications are in the latter camp and include smaller databases and web servers. So, make sure that whatever instance you select, can deliver the support you need when you need it.
A lot of money is wasted on having an instance size that is way out of proportion with what you need. For example, M5dn.24xlarge provides 384GB of memory. Do you need that much? Hope so, because you’ll have to pay for it, and it’s not the cheapest. Speaking of which…
There are four paid pricing tiers and a free trial available. You can pay by the hour or even by the second so that you do only pay for what you use. However, be warned: AWS instances run until you stop them. So, you’ll be charged until you do. There are also options available wherein you can pay to use AWS computing power during its downtime, which results in the lowest cost.
So, there are seemingly unlimited instances available but don’t feel overwhelmed. Similarly, with deciding which of the different types of testing to use, a little sober consideration of your requirements and a methodical analysis of what’s being offered will get you there.
About the Writer
Matthew Cooper is the Marketing Automation & Operations Manager at Global App Testing, a best-in-class software testing company that has helped top apps such as Facebook, Google, Microsoft, and Craigslist deliver high-quality software at speed all over the world. Matthew has over 14 years of experience in the I.T Networking, Software & Services Industries. He is highly skilled in Search Engine Optimization (SEO), Content Marketing, Digital Advertising, Social Media Management, WordPress, Email Marketing, Marketing Automation, CRM, and People Management. You can find him on LinkedIn.
A not-for-profit organization, the Institute of Electrical and Electronics Engineers (IEEE) is the world's largest technical professional organization dedicated to advancing technology for the benefit of humanity.