Update Nov 2017: selecting the most efficient AWS EC2 instance types using Pareto front analysis

AWS introduces the C5 family, the new AWS efficiency leader in CPU throughput per dollar

Back in May 2017, I posted an efficient frontier analysis of all EC2 instance types. AWS just announced the general availability of the new C5 instance family on November 7, 2017. The C5 family, which ranges from c5.large to c5.18xlarge is a compute-optimized instance family based on 3.0 GHz Intel Xeon Platinum 8000-series processors. In addition to network bandwidth and latency boosts, they claim:

“The new instances offer a 25% price/performance improvement over the C4 instances, with over 50% for some workloads. They also have additional memory per vCPU, and (for code that can make use of the new AVX-512 instructions), twice the performance for vector and floating point workloads.”

At dataxu, we’re running thousands of instances simultaneously. I’m always looking for a way to increase our system efficiency, so I decided to plug the raw numbers into my efficient frontier analysis from May 2017 and see if I agree with their claims. The updated chart is above. See the original article for how to interpret this chart. This is what I learned:

• The C5 family is 18% more cost effective in terms of compute per dollar (as measured by AWS ECU units) and 25% more cost effective in terms of memory per dollar.
• The C5 family is strictly superior in all aspects when compared to the equivalent instance types in the C4 family. The compute is cheaper, the memory is faster, and the network is faster. Notably, the C5 instance types offer huge boosts in EBS network bandwidth over C4: For example, the c5.large instance has up to 2250 Mbps of EBS bandwidth, whereas a c4.large is capped at 500 Mbps of EBS bandwidth.
• While there is no economic reason to use a C4 instance over a C5 instance, there may be a reason to use a less efficient C3 family instance. The C3 instance family has local SSD storage, should your application require this. C4 and C5 instances are EBS-only instance families.
• This is based on a simple spreadsheet economic analysis. Real-world performance tests of C5 vs. C4 instances may increase the gap even further, depending on your application as the C5 family uses hardware architecture and supports the AVX-512 instructions for enhanced floating point and integer performance.

P3 Introduction

In addition to the introduction of the C5 family, AWS also announced the GPU optimized P3 instance family in October 2017. The P3 family shows up in the chart this month in the bottom left corner. Since my chart only credits traditional CPU and Memory the P3 instance family presents poorly, which is not an accurate representation of this machine. I don’t currently have a method to compare GPUs efficiency, if you have some thoughts on how do this properly, please reach out or leave a comment.