Cloud Storage

Cloud Storage: The cloud began as a cheap proposition. Servers are expensive, and so are the air-conditioned rooms and extensive internet connections that servers need to thrive. Why not rent them and stop paying the huge bills for the hardware and the staff who care for it? Large cloud data centres can save at scale and then pass these savings on to the rest of us. More than a decade later, those cuts have started to add up.

Companies are easily expenditure millions of dollars on these small instances. The list price may be a few cents an hour, but the developers use so many that the total has the CFO clutching at heart medication.

Here are 12 different options to start saving. Some involve replacing cheaper parts. Others imply the total suppression of several pieces. Many involve improving communications so developers and the development team can make smarter financial decisions. None is perfect. They all involve trade-offs. None of them works in all situations. But they all deserve to be considered.

Turn to Commodity Cloud Providers

Big cloud companies offer a bewildering array of options, including some top-notch services for mundane jobs like data storage and exotic jobs like machine learning or computer vision. But if you’re not going to use the best APIs and services, you can find essential cloud brokers that sell pretty much the same thing for much less.

Find the Top Cloud Discounts

No one needs to pay the list price at the big cloud companies. List prices are intended for people who want instant gratification and machines that start with a snap of the fingers. Planning and looking for discount options is one of the easiest ways to pay less for top clouds. AWS, for example,  offers discounts for long-term commitments through Reserved Instances. Your spot market can be a good place to run background processing when prices drop at odd hours. Google offers automatic discounts that are activated the longer the machine operates.

Create Hybrid Clouds

Not everyone has friendly machines with cycles to spare in their office. Own devices can be much cheaper and even a little faster. Still, those can take advantage of increasingly sophisticated software packages to build secure subnets connecting them. Tailscale,  Submariner and  Traefik are some of the options to make this fast mesh from a mix of public cloud machines and whatever else is gathering dust. Some are building Kubernetes clusters. Others host back-end services that don’t need low latency. It’s not a solution for client-facing workloads, where clients count milliseconds while they wait, but it’s easier than ever to improvise something useful for background processing.

Explore ‘Bare Metal’ Servers

Not everyone needs the flexibility of modern cloud instances. Some providers rent entire servers without the additional layers of hypervisors and virtualization. Your code stays with the whole machine, with fewer security concerns. You have the actual root password. You may have to do small more work on preserving the OS layer, but sometimes that allows you to tune the mechanism better. IBM,  OVHcloud,  GTHost,  plus  Vultr are just four of the many companies that rent out bare-metal servers. Or you can buy your own.

Break down the bill and note the parts

Some cloud users melt when they get the monthly bill because their apps consume so much data. Some clouds charge for data exfiltration, which can add up, especially if your content goes viral. Other companies like Cloudflare deliberately build networks that don’t charge for exfiltration. So look carefully at your app profile. Do you use the web, data storage, or a particular line item on your bill? Chances are, some other cloud company has a better formula for you.

Stop paying more for Data Storage.

Some cloud companies, such as  Backblaze or  Wasabi, specialize in low-cost data storage, boasting that they charge up to 80% less than leading clouds. The savings can also be higher because they have different fee structures for data movement, allowing you to save on data exfiltration. These companies also offer an Amazon S3-compatible API that makes switching easy.

Go to Static

If your app has a lot of information that doesn’t change,  apps built from static web pages may be the cheapest way to reach a broad audience through a content delivery network. They’re not an excellent solution for tasks that require a lot of computation and customization on the back end, but perhaps that processing could be split into a separate service. Developers often talk about cutting costs by 10 or 20 when they can move everything to a CDN. Good, still frameworks include Jekyll,  Gatsby,  Middleman,  and  Hugo.

Go Serverless

The cloud is often better for applications with significant spikes in demand. If your app sits down for hours or days, rewriting it to run in a serverless framework like  Cloudflare Workers or  AWS Lambda means you won’t pay for all that downtime. You also won’t pay as much when the load is low.

Go to Tiny

Arm’s number of options is growing rapidly and sometimes surprisingly. We learned to hack with tiny  Raspberry Pi or Jetson machines that cost nothing. They are ideal for building embedded devices, Internet of Things applications, or just for experimentation. Now they can also be part of your cloud infrastructure. Raspberry Hosting and minisodes, for example, rent you a tiny Raspberry Pi or Jetson Nano board in a professional data center with a fast internet connection so that you can run essential software cheaply.

View Default or Default Settings

Many cloud instances come with default sizes that make them easy to configure. Our code doesn’t need all of the storage space in many cases, but we still pay for it. Several months ago, my cloud bill nearly doubled for local storage that I wasn’t using. To make matters worse, cloud companies often make it difficult to reduce this unused space. You can usually start with just enough space to run your application by being vigilant during the build.

Make the costs Evident to Everyone.

I still remember the manager’s face when I asked him for my test cluster because I constantly bumped into another programmer. His mouth said he would study it, but his face said it was an expensive and unreasonable request. But how was he to know? The developers on my side never saw any of the costs because the development team took care of all those details. Cost-sharing helps engineers understand that pennies per hour add up quickly when any Kubernetes cluster is scaled up to support a specific load. It may not be easy, or still possible, for developers to cut costs, but they’re less likely to suck more machines into the cloud if they know the price.

Also Read: What are Non-Financial Tokens NFTs – and why are they so Valuable