Credit: IDG
Everyone knows the pain that follows after a CFO takes a look at the cloud computing bill. The products and services are priced in fractions of a cent, but somehow those fractions all add up.
The good news is that more options are appearing as smaller clouds compete directly on price. In most cases, it's not quite accurate to use the word smaller because these cloud competitors are often quite big — they just don’t have the immense size and visibility of the major clouds. That’s why some prefer to use the word independent.
These cloud companies concentrate on the most popular services, from basic instances running Linux to commodity block and object storage. Then, they slap on a price tag that could be 40 per cent or even 80 per cent lower.
In many cases, the deal is even better. Some offer hidden discounts on extras like bandwidth, which are often overlooked. Many have different pricing models for features like data exfiltration or block storage, which can help users save even more money.
These cloud services aren’t for everyone. In many cases, independent clouds don’t have big research teams building sophisticated tools for artificial intelligence (AI), machine learning, and other cutting edge technologies. They rarely support as many operating systems or architectures as the big clouds do. Some don't provide time-saving managed tools like databases.
Yet, some savvy independent clouds turn all of this into a sales point. They point out the ways that simpler systems can make life easier for developers, devops teams, and even managers.
If a customer's team can work with basic instances, if the idea of something simpler sounds refreshing, or if the CFO is on their back to bring down cloud costs, then one of these cloud options might be right.
Backblaze
Many people know Backblaze as the company that markets an unlimited back-up product to individual users. Not everyone knows the company also offers an API compatible with Amazon Simple Storage Service (Amazon S3), that will squirrel away data for developers, too.
This service, called B2 Cloud Storage, is priced at US$5 per terabyte per month, an amount that’s about 70 per cent lower than the list prices from Amazon Web Services (AWS). The savings get better for projects that don’t just let the data sit there. Exfiltration may be one fifth of the standard price, a savings of 80 per cent. The service-level agreement promises 99.9 per cent availability, and the cloud is certified for many compliance standards like HIPAA.
Cloudflare
Cloudflare made its name offering a smart content-delivery network with a broad footprint that can help reduce bandwidth costs for delivering static content. Lately, Cloudflare has been expanding its services at the edge to do more than just keep unchanging copies in a cache.
The Workers product, for instance, runs arbitrary functions written in JavaScript, Rust, C, or C++, and bills according to how many resources are consumed when it's running. The serverless model can yield dramatic savings for functions that only run occasionally. The Pages product is tightly connected to Git repositories, so all customers need to do is commit some Jamstack code and Pages does the rest.
There are too many services to list here because Cloudflare is rapidly becoming a full-service shop with plenty of options for building out cloud services, networks, and APIs — all running on the edge of the internet.
Wasabi
As data proliferates, so do data-focused services. Wasabi is another independent service offering cheaper data storage using an API modelled on Amazon S3. Wasabi also offers extra features like reserved storage, which allows bigger customers to lock in a discount at a rate that could be dramatically cheaper than hosting their own disks.
Companies that need to worry about ransomware can turn to Wasabi's air-gapped, immutable storage. In the right combinations, Wasabi estimates that its service could cost one fifth of using Amazon S3 — and it doesn't charge for data exfiltration.
Vultr
Vultr's basic cloud instances using the older generation hardware begins at $0.004 per hour, which comes out to $2.50 per month. Prices go up if users want to use some of the newer, faster machines, but not by much. Vultr estimates that the increased performance of its newest AMD hardware more than makes up for the slightly higher price.
The vendor also offers Kubernetes clusters, bare-metal machines, and some object storage — all at lower rates. Vultr's network could be as broad as anyone’s. It is also proud of supporting 23 data centres around the globe, so there’s a good chance it has one close.
DigitalOcean
While it's not as small as it once was, now that it has gone public, DigitalOcean continues to target developers with a simple, no-nonsense interface. The DigitalOcean cloud offers a mixture of raw compute power and newer managed services.
Users can either set up their own database from a free image in the marketplace or let DigitalOcean manage a machine running MySQL, MongoDB, PostgreSQL, or Redis instead. The newer application platform offers a serverless option for Node.js, Python, Django, Go, PHP, and static sites.
Linode
Dedicated CPUs, shared CPUs, GPUs, or bare-metal machines: Linode offers them all, along with a wide range of managed services for tools like databases. Block or object storage is readily available and the company goes out of its way to emphasise that its menu doesn't hide pricing landmines. Linode includes free inbound transfer and “up to 1TB outbound.”
It's been building out a marketplace filled with prebuilt images for many of the most popular applications and some of the more obscure ones, too. Setting up a Jitsi server, for instance, takes just a few seconds. If users have more specific needs, they can spin up a StackScript that will configure new instances to their specs.
SSD Nodes
Projects that require plenty of RAM will find SSD Node's pricing model attractive. The vendor doesn't charge as much to increase the amount of working memory as others do. While some cloud providers will double the monthly rental price for twice the amount of RAM, SSD Node only adds a moderate charge that’s closer to the real price of the RAM.
The company emphasises virtual private servers with long-term contracts, with substantial discounts for 12-month and 36-month contracts.
Hetzner
If customers are looking for a bare-metal machine at low prices, Hetzner has plenty of options in a variety of configurations. The emphasis is mostly on classic servers with all of the RAM and drives bundled in one package, but Hetzner also has some cloud instances available in data centres in Germany, Finland, and the United States.
Hetzner bills with an hourly rate but includes a monthly price limit that caps the bill for long-running instances. Some users who like the old model of managed servers for basic websites like WordPress will also find options that are reasonably priced.
Upcloud
Fast servers with high performance are the focus for Upcloud. Indeed, Upcloud calls itself the “world’s fastest cloud” because its block storage offers superior performance; enough to be called “faster-than-SSD." The MaxiOPS technology is the default selection, but projects that don’t need super performance can be switched to the hard disk-based HDD option.
Upcloud also emphasises trustworthiness, offering up to 50-times payback for any outage lasting more than five minutes. Upcloud's product line includes the usual cloud instances, and some that can be partitioned into a private cloud with dedicated machines in a separate, physically isolated environment. Projects running on Microsoft Windows can also find a home here; Windows is an option, alongside the usual Linux distributions.
Big clouds look to go faster and cheaper
Not all the low-rent options are in independent clouds. Some of the big providers are exploring using ARM CPUs and many are reporting that they’re faster and cheaper. AWS, for example, is now deploying the third generation of its Graviton processors, which are custom built to deliver faster performance at a lower price in the Amazon cloud.
Progress has been rapid, and the third generation chip is said to be three times faster than the second generation, at least when measured on basic floating point operations commonly used in machine learning algorithms. Google, Facebook, and Microsoft are reportedly following a similar path.
