AWS Virtual Machine – Understanding the solution options

AWS Virtual machine, present their services on varying SKUs (product codes) or naming conventions and therefore it can be difficult to decipher what exactly constitutes general purpose compute, compute optimised, memory optimised, storage optimised or any of the other options. This blog hopes to explain and help distinguish the different naming elements, hopefully enabling you to be more informed when choosing what solution suits your application or service best. Obviously, if you are still not sure speak to one of our team for some guidance.

To ensure this helps to reveal some of the complexities of the naming conventions set out by AWS we look to highlight useful trends or patterns to understand the scale of their offerings. Once understood, it becomes simpler to decipher the iterations available.

AWS Virtual Machine – EC2

AWS Virtual Machines (AWS EC2) tend to follow a pattern whereby each of the virtual machine iterations uses the last letter of the virtual machine to portray the CPU. With virtual machines ending in the letter g (for example C7g) is powered by Arm-based Graviton3 processors, the letter a (for example T3a) is running AMD EPYC 7000 Series processors, the letter i (for example M6i) running Intel Processors.

The first letters for AWS Virtual Machines constitute the ability of the virtual machine and potential workload it is designed for. We have heard of FIGHT DR McPIXIE as a good reference point to follow, albeit it is slightly different with it ultimately being FIGHT DR MCPXZ. As seen below, each first letter of the Virtual Machine portrays its ability.

F – FPGA, I – IOPS, G – Graphics, H – High disk throughput, T – Cheap general purpose, D – Density, R – RAM, M – Main choice general purpose apps, C – Compute, P – Picture (graphics), X – eXtreme memory, Z – Z-factor. Extreme memory and CPU, A – ARM based instances, U – Bare metal

AWS Virtual Machines provide scalable architecture aligning both the vCPU to RAM count, for example, an increasing quantity of 2vCPU with 8Gib RAM, 4vCPU with 16Gib RAM, 8vCPU with 32GiB RAM and so forth. These tend to follow increments of 2x the underlying resources on each upgrade until the larger instances are required, whereby it scales at a rate of 1.5x the underlying resources, starting from 32vCPU and 128Gib and above.

As part of this guide, we will focus on two cloud architectures specifically surrounding General Purpose Compute and Memory Optimised Compute. There are further instances across the platform covering storage optimised, graphics optimised and further computing infrastructure but to keep it relatively focused we are focusing on two of the most popular cloud services.

AWS Virtual Machine General Purpose Optimised Compute – Most widely used for a wide variety of applications

Under the Amazon Web Services arm, general purpose machines sit under a number of instance types (currently 14x high level types) however we are going to review a few of these;

  • Lower vCPU count sitting in the T3 instance for burstable compute (providing a baseline CPU alongside the ability to burst at any time for as long as required) – T3 Series tend to start around the lower vCPU count of 1vCPU to 8vCPU.
  • If the need for 8vCPU or more is required, this branches into the M4 instance, M5 instance and M6i instance types – all of these do have varying iteration types.
  • M4 instances run on Intel Xeon E5 processors
  • M5 Instances run on Intel Platinum processors
  • M6i instances run on the 3rd generation Intel Xeon processors.

Additionally, with the AWS Service, a few alternatives are also offered Mac Instances built on Apple Mac Mini Computers. Both T4g and M6g instance types are powered by ARM-based Graviton2 processors with A1 instances running custom built AWS Graviton Processors. For customers looking for AMD processors within AWS, both T3a, M6a and M5a provide AMD EPYC 7000 processors.

AWS Virtual Machine Pricing – General Purpose Compute

A comparison table below provides AWS Virtual Machine pricing with varying vCPU, RAM and Amazon Elastic Block Store (EBS). These are based on the underlying Operating System being a Windows Server instance, but doesn’t take into account egress charges, speeds of SSDs and any snapshots/backups.

T2.xlarge – Intel Xeon

  • 4vCPU
  • 16Gib RAM
  • 3x 64GB Amazon Elastic Block Store (EBS)

Monthly Cost On-Demand

 

Approximately  

£201.00 per month 

M5.2xlarge – Intel Xeon Platinum  

  • 16vCPU
  • 64Gib RAM
  • 3x 128GB Amazon Elastic Block Store (EBS) 

Monthly Cost On-Demand

 

Approximately  

£467.00 per month 

M5a.4xlarge – AMD EPYC Processor 

  • 16vCPU
  • 64Gib RAM
  • 3x 256GB  Amazon Elastic Block Store (EBS) 

Monthly Cost On-Demand

 

Approximately  

£889.00 per month 

AWS Virtual Machine Memory Optimised Compute – Suited for memory intensive applications

As a secondary comparison, we will be looking at memory optimised VM’s. Just like General Purpose, there tends to be the most options to chose from here.

Amazon Web Services provide a variety of memory optimised environments designed to deliver fast performance to workloads that need to process large data sets in memory.

  • The initial offering sits within the R4 instance type, providing an Intel Xeon E5 processor and progresses to R5 and R6 instance types.
  • R5 AWS Instances provide a blend of Intel Xeon Platinum processors
  • R5a Instance is for AMD EPYC 7000 Series processors
  • R5b instance is for EBS-optimised instances based on Intel Xeon

The upgrade path which provides a rough 20-40% better price and performance, is the R6i instance type based on Intel and is a direct upgrade to R5 instances. Also under the R6 instance types is the R6g instances which are Arm-based AWS Graviton2 processors and provide better performance over R5 virtual machines and is EBS optimised by default.

  • X instance types from AWS focus more enterprise-class databases and in-memory applications.
  • X1 relies on the Intel Xeon E7-8880 processor and has one of the lowest prices per GiB RAM in AWS and provides up to 1,952GiB of DRAM-based instance memory.
  • If the requirement needs to exceed 1,952 GiB then X1e instance provides up to 3,904 GiB with the same underlying infrastructure.

There are some further AWS memory optimised environments coming soon based on 2nd and 3rd generation Intel Xeon processors with up to 55% better price performance when compared to X1 instances alongside a 16:1 and 32:1 ratio of memory to vCPU on all sizes – these are yet to be released but is part of AWS ongoing plans to increase capability in this field.

AWS Virtual Machine Pricing – Memory Optimised Compute

A comparison table below provides similar virtual machines with varying vCPU, RAM and managed disks. These are based on the underlying Operating System being a Windows Server instance, but doesn’t take into account egress charges, speeds of SSDs and any snapshots/backups.  

R5.2xlarge – Intel Xeon  

  • 8vCPUs
  • 64Gb RAM
  • 2x 128GiB Amazon Elastic Block Store (EBS)

Monthly Cost On-Demand

 

Approximately  

£485.00 per month 

R6i.12xlarge – Intel Xeon Platinum 

  • 48vCPU
  • 364Gib RAM
  • 4x 512GB Amazon Elastic Block Store (EBS)

Monthly Cost On-Demand

 

Approximately  

£2951.00 per month 

R5a.8xlarge – AMD EPYC Processor 

  • 16vCPU
  • 64Gib RAM
  • 3x 256GB Amazon Elastic Block Store (EBS) 

Monthly Cost On-Demand

 

Approximately  

£1828.00 per month 

As previously mentioned, there is a variety of different instance types of AWS Virtual Machine with this guide scratching the surface due to only focusing on two particular service types. It is interesting to see that AWS Virtual Machines understand their customer platforms providing a balance of varying Intel and AMD processors for different environments providing the customer choice.

It also amplifies the focus on innovation of these areas and upgrading their underlying infrastructure to encompass changing customer workloads and applications. The pricing tables help to provide a basic level of AWS Virtual Machine pricing, although considerations still need to be made for networking/egress bandwidth charges, speeds of storage, connectivity between on-premise and data centres through to backups/snapshots that are part of a fully cloud based service.

More Cloud Computing Resources: