When I first started as an enterprise developer in the mid 1990s, everything ran in our on-premises data centers, and procurement was slow. Purchasing servers, racking them and installing all the base software on them would take around three months, sometimes longer.
A side effect was that when we were sizing a project to forecast how much hardware was needed, we would typically overprovision. I would work out what I thought the maximum load was, run load tests to simulate that, and then, once I was confident I knew the requirements, double them just to be on the safe side. I was never challenged. I assume this was because overprovisioning is such a common — and generally successful — risk-management strategy.
In theory, the cloud changed this entirely. Elastic compute means you can get the hardware you need, when you need it. For developers, this means shifting workloads to the cloud has the potential to save you money and reduce your carbon emissions.
However, as I explain in The Developer’s Guide to Cloud Infrastructure, Efficiency and Sustainability, an upcoming ebook from The New Stack, Google and AMD, “moving to the cloud is only potentially more carbon efficient [than a typical enterprise data center]. It heavily depends on a wide range of factors, including your choice of regions, server density and CPU, as well as your use of managed services and spot instances. Of course, you can still over-provision, even in the cloud.”
In other words, the cloud can be a lot more sustainable than on-premises infrastructure — but that isn’t guaranteed, and simply lifting and shifting your existing apps won’t cut it.
Unpacking CPU Performance in the Cloud
A topic that tends to get less attention than it deserves regarding sustainability is choice of CPU. I think this may be because it is, to a large extent, abstracted away in the public cloud. But CPUs need electricity to run, and they have embodied carbon.
The good news is that CPUs are getting better at energy proportionality. Building Green Software explains that CPU design improvements mean they are getting closer to “achieving perfect proportionality.”
These improvements are part of what cloud vendors have been able to exploit to increase their profitability. But you can take advantage of them too. In The Developer’s Guide to Cloud Infrastructure, Efficiency and Sustainability, I explore:
- How developers are helping build a sustainable future.
- Developer best practices for choosing a cloud VM.
- The positive and negative environmental impact of cloud infrastructure.
- The factors that can make cloud computing “green.”
- How to measure carbon emissions from your cloud infrastructure.
I believe that sustainability should join cost, performance, security, regulatory concerns and reliability as one of the top-level considerations when optimizing your cloud workloads. If you want to learn why, and what you can do to measure and improve the “greenness” of your cloud deployments, please preregister for early access to this free ebook, which will be released on April 1.
The post How to Choose the Right Cloud CPU for Your Workload appeared first on The New Stack.
