AWS Reserved Instances (RIs) are a good cost-saving option. However,  it’s usually not the best option for non-production workloads given the usage rigidity and cost commitment. Companies under pressure to quickly cut costs sometimes still jump on it even for QA or development machines only to later regret it. Here’s one such example from a Lightwing PoC.

We recently completed a successful deployment with a company in Bangalore that provides a platform for enterprise clients for their internal administration needs. They boast of marquee customers serviced by an able management and technology team. The only problem brief we received was that their AWS bill was too high and that it needed to come down, fast.

Before meeting with us, they had procured Standard RIs for their production workloads across multiple families including t2 and t3.  Trouble is, they have several non-production workloads on t2 and t3 family machines as well. Because of this, their RI utilization crept into their non-production machine clusters too. This wasn’t detected until we analyzed their spends and wondered why there was no reduction in the monthly bill even though Lightwing showed that their bill should have come down by 72%–82% based on the idle time when the machines were shut down.

Lightwing AutoStopping

A quick mention here on what Lightwing AutoStopping does. Lightwing’s AutoStopping gateway feature helps eliminate excess cloud spends on non-production (QA, staging, development, demo, R&D) machines using something we call as consumption-based automation. In a nutshell, this just means that your non-production resources are always there when you need them (100% availability), but you are only charged for them by the cloud provider when you are actually using the resources, never when they are idle. On Lightwing, you set up what we call as a Gateway that monitors network traffic into your  non-production resources. When there is traffic, the resources are started. As traffic continues, the resources stay alive. When there is no traffic beyond a certain pre-set idle time, the resources are shut down (or terminated after a snapshot is taken, if it's a spot instance) by Lightwing. This way, you always have access to these resources at any time, while also ensuring that you are never charged when they are unused. Setting up Lightwing AutoStopping is a simple one-time effort that typically takes just a few minutes, and required no on-going manual intervention whatsoever. The key advantage, we have learned, is that it also requires absolutely no change in the way your developers operate. In fact, most developers in the company usually aren’t even aware that Lightwing is being used for optimizing their cloud spend. On average, companies see their non-production bills come down by 70%-90%, with just this simple setup. AutoStopping works with both on-demand and spot instances.

So why is utilizing RIs for non-production machines not recommended? If  you read the paragraph above on how Lightwing works, you will see that  you can save up to 90% on your monthly cloud bills, which give you complete flexibility on usage and payments. No capex of any kind, like with reserved instances. RIs, on the other hand, cannot typically match these type of savings, not to mention you have payment commitments of 1-3 years or more. RIs also do not give you much flexibility on usage, in that you are charged the RI fee whether or not you actually utilize your purchased RIs.

RIs do not give you much flexibility on usage, in that you are charged the  RI fee whether or not you actually utilize your purchased RIs.

Standard RIs, for example, will give you about 38% in savings if you commit to paying for a year or more every month. No flexibility in changing the machine family, etc. So, for non-production instances, in comparison to RIs, Lightwing AutoStopping gives you the full usage and payment flexibility of on-demand or spot instances, and double the savings. This was our exact business case to this particular company.

In comparison to RIs, Lightwing AutoStopping gives you the full usage and payment flexibility of on-demand or spot instances, and double the savings.

No marks for guessing that the company proceeded to onboard all of their non-production machines onto Lightwing AutoStopping, and utilized those RIs for some of their production workloads. They even had some excess RIs eligible for releasing. All this, with a significant reduction in their monthly AWS bill. Today, you can even run production workloads seamlessly on spot instances with Lightwing ComputeGroups.

Try Lightwing AutoStopping for your non-production workloads and Lightwing ComputeGroups for your production workloads with a 14-day free trial. No credit card required. See detailed analytics as your bill reduces.