AMD Epyc 4 - The Perfect Processor for Virtualization?
Transcript
(logo booms) (graphic chimes) >> Hey, it's Keith Townsend, Principal, The CTO Advisor. Thanks to AMD for sponsoring this CTO talk about selecting a processor for your cloud infrastructure. We've been pretty consistent that we're in a hybrid world. You're going to have multiple providers. You're going to have data sets living on premises, in the cloud, and anywhere in between. Fortunately, the industry as a whole is making this much easier for us than it has been in the past. We're starting to see certain use cases shape, especially around cloud native.
If we look at what the big cloud providers are doing, they're taking ARM processors putting it in their data centers, driving down the per wattage use for performance inside of their data centers and then passing those savings onto you. We're seeing this effect in the enterprise as well with some folks in the cloud native community trying to follow suit with the cloud providers. What have they learned? They've learned that VMs are still an amazing platform to deploy, not only containers, but applications directly.
I've been watching a lot of content around Firecracker VMs, which are micro VMs that spin up almost as fast as a container, but all the things we love about VMs, separate kernels, the ability to have dedicated TCP/IP stats, so the networking nightmare of needing to have sidecars, eBPF, all of these workarounds we find in Kubernetes ecosystems we're not seeing that and we can still use Kubernetes as the orchestrator. But infrastructure truths are still consistent, as the cloud providers are proving out, low wattage ARM processors are great for microservices and other application designs which require just hundreds if not thousands of VMs on a single set of host.
However, one of the challenges that enterprises have that cloud providers don't, is that you have to rewrite your software stack for the ARM processor. This is why I'm excited about AMD Epic's fourth gen processors, in which they're giving us something like 128, no, it's not up to 128 cores, efficiency cores in a single CPU package. What does that mean practically? 8 terabytes of RAM can go against all of these small micro VMs to give you the density and that performance that you need without sacrificing the need to re-platform all of your x86 base code for ARM.
Again, the industry is moving forward. We're going to be in a hybrid nature world where we're running workloads in the public cloud, reclaiming some of those workloads that are running on premises, but what's consistent is the desire to have a cloud-like operating model, at least a cloud-like interface. And I'm happy to see companies like AMD answering the bill and giving us these processor options that fit these new models of compute. com. I'm on most social media platforms as @ctoadvisor and my DMs are open.
Talk to you next, CTO Advisor, Studio Chat.
