Interview Civo, a cloud provider based in Hertfordshire, has made its flavour of K3s Kubernetes generally available, with the claim that a usable cluster can be fired up in 90 seconds.
K3s is a lightweight Kubernetes distribution from SUSE's Rancher, which claims to be "easy to install, half the memory, all in a binary of less than 100 MB."
Civo is among the smallest of Cloud Native Computing Foundation Certified Kubernetes providers and picked K3s for its speed and efficiency.
CEO Mark Boost was the founder of UK hosting company LCN where he remained until it was sold to Team Blue in early 2019, and also of ServerChoice, which runs three data centres for co-location services.
Boost's pitch is that the giant cloud providers such as AWS, Microsoft Azure, and Google Cloud Platform are both expensive and unpredictable in cost.
"I think there's this misconception in the industry that hyperscalers are cheap, which they're not," he tells The Register. "They're more expensive than a lot of other providers out there... the phrase cloud repatriation was coined four or five years ago, where people realise that running physical tin is a lot more cost-effective."
The difficulty with such claims is that comparing like with like is hard. Removing some of the burden of operational costs as well as providing a vast array of features that can be hard to match on-premises keeps customers hooked on the big cloud providers.
Boost says that it is not only the absolute cost, but the challenge of cost management, that is a problem.
According to Civo's own survey of cloud developers conducted in July, around 45 per cent of developers reported surprise bills and difficulties working out the cost of running containers, and 15 per cent said the opaque pricing of hyperscalers was a primary frustration with the technology.
"There's so many services on AWS, everything charged by the hour, these small charges that look meaningless. You have no concept of what that looks like once you start scaling up, until you start to use these services in anger," he says. "I find it a really weird concept that there's companies that just do consultancy around how to understand your AWS bill and how to reduce your spend."
Are the profit margins of the big providers excessive? "I've been involved in the data centre industry since 2008," Boost told us. "That's from a different company that I run called ServerChoice. I've seen the data centre costs and bandwidth, and we've bought from Tier 1 carriers and understand what the wholesale costs of buying bandwidth are, versus the data transfer charges that the hyperscalers charge out, are so far out it's quite incredible" – a point which CloudFlare has made repeatedly.
Boost regards high data egress costs as a kind of lock-in. "It makes it a challenge to move from one provider to another, should you wish to do so," he says. That in turn undermines a benefit of Kubernetes. "Kubernetes is meant to be standardised. If I'm not happy with this host, I can move to another host. It should be down to customer service, performance, reliability."
Data transfer fees are easy to measure, but what about the vast array of features on the big cloud platforms, and multi-region resilience, that are hard for a small provider to match? "We're a newer player, we haven't got the scale or the resources that they have access to, but we have the same level of resilience built into our platform," he says.
"There's an opportunity to take a lot of business from Amazon, and the likes of Google and Microsoft." He adds: "We're not trying to compete directly with Amazon, we're trying to focus on five or ten services. We're not trying to be everything to everyone."
- Jamstack research: Typescript and serverless are the winners
- Judge rejects claims Cloudflare should be held responsible for customers' copyright infringement
- Cloud Foundry insists Kubernetes transition still alive despite VMware's retreat
- Celonis and Software AG target IoT and streaming process mining as firms grapple with data overload
What about things like reserved instances or spot prices that enable lower prices from AWS and others? "We haven't done that so far," said Boost. "Our feeling is that we price competitively from the word go. There's not a huge amount of movement for us to go cheaper."
There is one aspect of the Civo platform that could be a good or bad thing depending on perspective, which is the amount of custom code used to provide the service. The company used to use OpenStack, but "we haven't used it for over a year. We've got one legacy region that we've got some customers on. The problem with it, it wasn't built for cloud native or Kubernetes, they've built Kubernetes into it as an afterthought. It wasn't really working for us, it's very complicated, lots of different modules, so we took it back to the drawing board, built from scratch, and architected our own platform called CivoStack. It's optimised for Kubernetes, we still offer VMs but it's through Kubernetes because we use KubeVirt. Everything is cloud native, things like GPU-as-a-service we're going to be offering soon, but it will be done with Kubernetes."
Is there a risk in all that custom code that has not had the scrutiny of open-source software? "It's proprietary code but we do use open-source components within that," says Boost. "Then we extend those, and we've built our own custom operators to build and manage and deploy and scale clusters.
"The approach used to be 'no one got fired for buying IBM' and maybe that expression is now 'nobody got fired for buying AWS'. With a smaller provider, he says, "you can get a much more personal service." The general availability of Civo Kubernetes, he says, "is a chance to take a little bit of business away from hyperscalers, to demonstrate that there's alternatives that offer enterprise functionality at a good price."
Civo claims that its pricing for an equivalent Kubernetes cluster is less than one third of AWS, Google, or Microsoft Azure. That does not take into account reserved capacity discounts and so on, but that is part of the point. ®