A RedMonk Conversation: Arm64 for the Best Price/Performance on AWS: Why You Should Take The Graviton Challenge

I enjoyed this conversation with Dave Brown, VP of EC2 at AWS about specialist processors generally and Graviton2 specifically. We start with some goodness about the early days of EC2 – you may not know a lot of the code was written in Cape Town, South Africa. We then talk about what’s up with ecosystem and developer toolchain support, and how will AWS attract more developers and builders to take advantage of these instances? One answer – the Graviton Challenge for workload migrations. If you’ve got a workload you’d like to optimise, AWS wants to see it.

One reason I am interested to see more folks try these instances is that their efficacy has been confirmed by the good folks at Honeycomb, an organisation with an opinionated, high performing engineering team.  Here is a great thread from Liz-Fong Jones and the associated blog post (plenty of good gory details in there) on their initial experiments. This was the TLDR experience after a week:

For the past several weeks, my side project has been exploring the possibility of running Honeycomb on ARM64 instances, specifically, AWS’s M6g instances, powered by Graviton2 processors. The performance gains and cost savings are real, y’all.

Here’s a graph from the experiment:

The outcome of our week-long bake running entirely on Graviton2. Notice the 27%+ decrease in instance count peak-to-peak, and 37.5% fewer instances for baseline load, which averages out to >30% fewer instances overall.

These pics tell a thousand words.

And here is the results after 12 months in prod:

A year out from our initial Graviton2 dogfood experiments, we can report that 92% of vCPUs in use by Honeycomb are arm64, spanning virtually all of our workloads and all of our environments. The investment we put into tooling for service standardization paid off, as we were able to push-button migrate multiple individual services over the year with a minimum of service-specific wrangling. The change to Graviton2 has been transparent to all of our developers, and in fact, having already deployed arm64 in production has ensured the transition to M1 Apple Silicon Macs has been smooth for our development team’s local testing.

Traffic to Honeycomb has surged in the past year, but the expense for serving and the complexity of our services hasn’t. Graviton2 is a large part of how we do it.

Sounds good, right? Here’s a link to to the AWS Graviton challenge. Do a migration and you could be talking at re:Invent this Fall. There are various prizes in categories from startups to enterprises.

disclosure statement: this video was sponsored by AWS, and Honeycomb is also a client.