Cloud cost optimization fallacy

With interest rates soaring, there's been a rising trend where even traditionally "growth focused" companies are taking a very, very hard look at their operational costs. I have seen renewed interest in cloud cost optimization at a lot of companies that I talk to. A recurring theme is how their organization or company saved x% by doing things like right sizing VMs, adding spot instances, and buying the right amount of reserved instances, amongst other things. All very valid things. Some companies, like 37signals, even went so far as to give up on cloud entirely to move back on premises in their search for cost savings.

I kind of get it. There's a limit to how much you can optimize, and there's also a breaking point for when it no longer makes sense to spend time (=money) on optimizing further. Spend $10 to save $8? You ultimately run into negative marginal returns.

However, I also believe there's a false premise that's been built up over time in the industry as a whole. One example would be wildly expensive egress rates often associated with cloud computing, touted as the price of admission to get the benefits of a hyperscaler, such as resiliency, scale, elasticity and fast time to market. Sure, you do get those benefits by going to a cloud provider. However, is that really a good reason why you have to overpay for things like egress?

Let's say you run on AWS and GCP, like 37signals did, and you look at your egress charges and start to optimize for cost. You probably make sure you use a CDN aggressively and start caching as much as possible to reduce your origin egress. You start thinking about cache hit rates and cache invalidation. You might look into refactoring your application to be less chatty. Maybe you try to re-encode video to lower bitrates and more efficient codecs. You might even try to renegotiate your contract to bump up the discounts on egress.

In other words, you have made your cloud usage as cheap as possible. You have optimized as far as you can. It's still too expensive. You are Ray Ewry in the 1908 Olympics. You have worked really, really hard to optimize as much as is possible.

However, if you actually were Ray, you couldn't have known that Dick Fosbury would revolutionize the sport by doing something unconventional...the Fosbury flop. All world records since 1978 have been done with the Fosbury flop. You will never achieve optimal results trying to optimize something inherently inefficient.

Sorry, I know I shouldn't mix engineering and sports history. There's a reason we engineers always use car analogies I guess...

I believe many cloud consumers have fallen victim to a false premise without knowing there's a better way. By our very nature I think we like to optimize, to tinker. I know I do. But what if, in our egress scenario, I could just give you 90% lower price from the get go, before ever starting to cost optimize. Even if you would have cut your bill in half by optimizing cost, you would still be multiples more expensive than my non-optimized environment. Why put in all that effort if it only results in a bigger bill?

OCI has 90-92% cheaper egress than AWS, GCP or Azure. In other words, this is not a fictitious thought experiment. Many executives I talk to have done their due diligence comparing offers from AWS, Azure and GCP, and many times that is where they stop. If you have looked at the top three, and they all look eerily similar, would you really spend the time to look at a 4th? Maybe not. This also goes to reinforce the misconception that all cloud services are the same.

I strongly believe Oracle has built something different, not only from a technical perspective, but also how it approaches things like compute and network cost. Compute costs on OCI are roughly 55-60% lower cost, ceteris paribus (everything else being equal), and networking is close to 90% lower cost. Imagine having spent a year cost optimizing your cloud spend, just by realizing you could have ended up with something much cheaper ... without having done all that work.

I've been thinking about what to call OCI. Many ideas, such as "financially feasible cloud", but maybe I should just start calling it "Fosbury Cloud". Oracle has set a new bar for cost effective cloud computing; let's just not call it a "flop".