The General Purpose AWS instance family welcomed a new member last week with the introduction of M4s to the ever-growing list of instance options for AWS users. Featuring a custom Intel Haswell processor “optimized specifically for EC2” and the impressive performance of Enhanced Networking, these instances balance compute, memory and network and are available in five sizes.
We were curious upon learning of this new M4 instance type to see how it would compare to other instance types from both a performance perspective, and a cloud cost perspective. What we found was exciting—and there are some interesting nuances to the breakdown.
The M4 features the following specs across five usage types:
These specs are notable in two ways. First, when compared to equally sized M1 and M3 instances, M4 performs equally or better across every dimension:
If you have specific storage requirements for your General Purpose instances, you’ll select whichever type best suits those needs. But if not, M4 is clearly the highest performing General Purpose instance.
M4 instances are also notable in their scope. The M4.10xlarge offers compute, memory, and networking capabilities significantly greater than what was previously available for the largest M1 instance (M1.xlarge) and the largest M3 instance (M3.2xlarge):
These specs are so impressive that they even draw near to what can be achieved with the largest instances of other types: the Compute Optimized C4.8xlarge (which features a slightly greater ECU and slightly lower vCPU and RAM), the Memory Optimized R3.8xlarge (which has significantly more ECU, but less vCPU and RAM), or the Storage Optimized D2.8xlarge ( which has greater RAM, but less vCPU and less ECU).
When you factor in the fact that of these three instances, only C4.8xlarge costs less than an M4.10xlarge (and that a D2.8xlarge costs over twice as much), the versatility of the M4 becomes clear.
While an M4 instance will certainly perform better than its fellows in the General Purpose category, whether said performance is cost-effective is a potentially different matter. Fortunately, when you look at the hourly cost per performance dimension, M4 is the most cost-effective General Purpose instance type for every category when you’re paying on-demand:
Choosing a General Purpose instance to pay for on-demand? M4 is almost definitely your most cost-effective bet. And this doesn't appear to change when you introduced Reserved Instances to the equation. In this example, we’ve applied All Upfront 1-year Reserved Instance prices to the same instances:
Upon introducing Reserved Instances, M4.xlarge remains your cheapest and most cost effective option across all specs. As such, M4 will likely be the best choice for most General Purpose needs, both on-demand and reserved.
It’s clear with the release of M4 instances that AWS is eager to continue improving its offerings from both a performance and cost perspective; these instances will represent a powerful and reasonably priced resource for many users. However, while migrating to the most cost-effective instances available is always a great practice in AWS cost management, introducing M4s where applicable is only one piece of the puzzle. Log in or sign up for a free 14-day trial of Cloudability to see how you can monitor, optimize, and govern your AWS spending today.