Bill shock? The red ink of web services doesn’t come out of the blue
To avoid cloudy judgement, heed the weather forecast
Opinion Simple sums can pack a punch. When the CTO of 37Signals got his $3 million cloud bill for 2022, and after the red mist had cleared, he sharpened his pencil to see if that was kosher.
You can see more details here, but to give one part of the breakdown, he calculated that the AWS monthly compute component of the giant invoice was £63k a month. Buying equivalent hardware from Dell worked out at just $1.3k a month. A factor of more than fifty. Ouch.
Even taking into account the bleeding obvious – that buying a server gets you a metal box full of sand – that's a lot of spare headroom for the necessary software, power, aircon, people, networks, and big old server shed to put it all in, so it generates revenue. It seems a powerful argument against the public cloud in the eternal on-prem/private/hybrid/public service hosting wars that rage in many a CTO's head.
We've all heard sphincter-tightening tales of eye-popping web service charges that popped out of nowhere, as Douglas Adams noted, like a large drinks bill. Instances that run unchecked or storage that silently grows and grows can scale to career-threatening proportions if proper monitoring precautions aren’t taken. Not the case here, 37Signals insists; a large team spent a lot of time doing the smartest deals they could.
Going public on such matters is quite the thing in this business, so hats off to them for that. But it's a good excuse to look at what questions you could ask yourself in similar circumstances.
Deciding how to best spend your money to create your service is an engineering process like any other. You decide what you need to create, set the parameters it's going to operate within, and look at the options for making it happen. Once you've got a set of options, you cost them and use that to guide your final decision. And forecasting the up-front and recurring costs for services and software in IT is one of the easier parts of the job. Supplier quotes are contractual.
People and physics are harder to pin down. Development costs of your service will have all sorts of in-house, buy-in and outsourcing decisions, as will security and day-to-day operations staffing needs. Resilience and disaster recovery, well, how many nines do you need?
It's perfectly in order to do as much of the hard stuff yourself as you deem fit, provided you have a clear-eyed view of the competencies, costs and consequences involved. What are the single points of failure, and how do you fail over? How do you test it, and who picks up the pieces? If you're really good at this, wouldn't you be better employed selling those skills rather than being a CTO in charge of yet another B2B on-demand service? If you're not really good at this, should you be taking it on?
- Time to study the classics: Vintage tech is the future of enterprise IT
- Too big to live, too loved to die: Big Tech's billion dollar curse of the free
- ChatGPT has mastered the confidence trick, and that's a terrible look for AI
- Twitter is suffering from mad bro disease. Open thinking can build it back better
These are some of the decisions that define a CTO's right to wear their underwear over their trousers – but they are engineering decisions nonetheless, amenable to rational enumeration.
Very little of the above ends up with undecidable hardware, software or web service costs, which will be contingent on how you choose to do things rather than happenstance. If they're among the things that surprise you after deployment, it's worth asking yourself what you missed at the outset.
It's not that agreed supplier charges that verge on the criminal are unknown, or even always avoidable. In general, the better the stranglehold a company has on its market, the more outrageous the heist. From IBM through to Microsoft via Oracle, once competition is shut out, the standard contract can best be summed up as "open your wallet and repeat after me – help yourself."`
No company was ever forced at gunpoint to pay Oracle's server licensing fees –the option to write the database software themselves was always there – but it remained a logical decision to fork out the dosh instead of taking on those costs and risks.
That's not yet true of any cloud provider, not even with AWS's dominance. There are always alternatives, albeit all with implications about the ecosystem you choose, and you do get the option to pay a lot of money for a job that should cost a lot less. But you can also pay much less for a job that should cost a lot more, and that can end up even more expensive. It might not. Are you feeling lucky, punk?
This industry is full of unpleasantly exciting challenges, some surprising, some unavoidable. What should never happen is a whopping end-of-year bill you didn't bargain for. When it comes to what drops from the cloud, the least we can do is invest in an accurate forecast. ®