Amazon EC2 T3 Instances: Selecting Best-Match Workloads for AWS T3s

calendar August 24, 2018

What Are The Advantages of Amazon T3 Instances?How to Determine When T3s are Right for Your AWS Workloads

Amazon Web Services (AWS) has introduced Amazon Elastic Compute Cloud (EC2) T3—burstable, general-purpose instances that are available across twelve regions.

AWS EC2 T3 instances offer a balance of compute, memory, and network resources, and are designed to provide a baseline level of CPU performance with the ability to burst above that baseline when needed.

AWS EC2 T3 instances

T3 instances run on next-gen hardware, powered by custom 2.5 GHz Intel Xeon Scalable (Skylake) processors, and provide a higher level of performance and efficiency. Today, T3 can be launched in seven sizes:

Instance Name vCPUs Baseline Performance/vCPU Memory Price/Hour (Linux) Price/Hour (Windows)
t3.nano 2 5% 0.5 GiB $0.0052 $0.0098
t3.micro 2 10% 1 GiB $0.0104 $0.0196
t3.small 2 20% 2 GiB $0.0209 $0.0393
t3.medium 2 20% 4 GiB $0.0418 $0.0602
t3.large 2 30% 8 GiB $0.0835 $0.1111
t3.xlarge 4 40% 16 GiB $0.1670 $0.2406
t3.2xlarge 8 40% 32 GiB $0.3341 $0.4813

When to Move Workloads to T3 Instances

The introduction of T3 makes burstable instances types even more attractive for running workloads that don’t require high level of sustained CPU performance. T3 instances are unlimited by default, meaning that, while they won’t be throttled, they can rack up additional costs if you run CPU-hungry workloads. Densify analyzes the base performance requirements of each of your workloads to decide whether a specific AWS T3 instance is a good match for it, or if it would be better served by C5, M5, R5, or other instance families.

Here is an example of a good cloud workload candidate to migrate to burstable T3:

CPU utilization for a workload that is a good match for Amazon EC2 T3

And, here is an example cloud workload that’s not a good candidate for burstable T3:

CPU utilization for a workload that is a poor match for Amazon EC2 T3

Densify’s complete knowledge of available services and configurations—across public clouds and on premises—plus our patented machine learning gives you the best answer for where to place your apps and workloads.

Request a demo of Densify resource management

What to Consider When Moving Workloads to T3 Instances

T3 instances are powered by the Nitro system which means that they support network and EBS bursting. This also means that you must watch out for things like HVM virtualization exclusively (no Paravirtual) and the need to launch within a virtual private cloud (VPC) using an AMI that includes the elastic network adapter (ENA) driver.

Densify performs comprehensive checks utilizing our patented, rule-driven engine to compare the requirements of your workloads against the capabilities of different cloud instance types based on metadata from AWS. We make sure only suitable workloads are considered as migration candidates for T3, and any additional considerations are highlighted and presented to our users.

Here are examples of rules that are evaluated as part of multidimensional analysis:

Workload analysis rules

One of the rules in action:

Details of a workload analysis rule

Cost Comparison of T3 vs T2, M5, & R5 InstancesShould You Use T3 for Your Workloads?

Let’s compare EC2 instance types with 16GB RAM:

Instance vCPUs ECU Baseline Performance per 1 vCPU Memory Price/Hour (Linux) Price/Hour (Windows)
t3.xlarge 4 Variable 40% 16 GiB $0.166 $0.24
t2.xlarge 4 Variable 22.5% 16 GiB $0.186 $0.227
m5.xlarge 4 16 n/a 16 GiB $0.192 $0.376
r5.large 2 8 n/a 16 GiB $0.126 $0.218

Assuming the following:

  1. Instances running @ 100% for a 1-month period
  2. Baseline performance for t3.xlarge is 1.6 vCPUs (40% * 4 vCPUs)
  3. Baseline performance for t2.xlarge is 0.9 vCPUs (22.5% * 4 vCPUs)
  4. Additional charges of $0.05 (for Linux) and $0.096 (for Windows) per vCPU-hour for T2/T3 bursting beyond credits

Here are the monthly costs for the various instance types:

Instance Linux Windows
t3.xlarge (throttled) $120 $173
t3.xlarge (unlimited) $206 $339
t2.xlarge (throttled) $134 $163
t2.xlarge (unlimited) $246 $378
m5.xlarge $138 $271
r5.large $91 $157

Amazon EC2 T3 Break-Even Points

So, what is the break-even point for the amount of time a T3 unlimited can run at 100% CPU versus the equivalent M5?

For Linux
($138-$120)/(0.05*2.4) = 150 hours out of the month or 21% of the time
For Windows
($271-$173)/(0.096*2.4) = 425 hours out of the month or 59% of the time

Key Insights about Running Workloads on AWS T3s Versus Other Instance Types

Amazon EC2 offers a wide menu of instance families and types to select between. These guidelines can help you make the best choice for your T3-compatible workload:

  • If you can fit on less CPUs, R5 is the best value of the bunch
  • A T3 unlimited running @ 100% can cost 25-50% more than the equivalent M5 – the extra charges for unlimited bursting can be significant
  • With the default setting of unlimited for T3 instances, customers will often inadvertently leave it as unlimited – and rack up extra costs
  • Densify can detect when unlimited bursting charges exceed a threshold – e.g. breakeven point with M5 costs – and recommend different instance type or turning off unlimited (e.g. dev/test workload)
  • T3 are a notably better value than T2 – more CPU credits and better baseline performance, faster CPUs – but just need to deal with migration effort
  • The cost difference of the instance types between Linux and Windows is significant, and results in different optimal instance types:
    • For Linux, T3 is cheaper than T2, but for Windows T2 is cheaper than T3
    • The breakeven point for unlimited bursting of T3 vs M5 instances is significantly lower on Linux vs. Windows (21% vs 59%)

EC2 T3 Reserved Instances Considerations

As with the introduction of any new service, the announcement of T3 instances has left some AWS customers with pre-existing T2 reservations publicly expressing their buyer’s remorse. With the help of Densify and expert insight from our Densification Advisors, you will be able to plan and execute comprehensive RI purchase strategy that will help to utilize your existing RIs and safeguard against any disappoints by reserving for the Optimal state rather than Current state, as well as avoiding purchasing long-term unconvertible reservations for older instance types.

Get Automatic Matching of Candidate Workloads to T3s

Densify is the only technology with complete knowledge of available services and configurations—across public clouds and on premises. See how we deliver the best answer for where to place your apps and workloads!

Request a demo of Densify resource management