Using the large cloud computing providers is convenient, but not exactly cheap. Ubicloud, a new startup founded by the team behind Citus Data, which
Ubicloud aims to provide an open source alternative to AWS by offering core cloud computing services on affordable bare-metal servers.
The focus is currently on compute, PostgreSQL database service, networking capabilities, with plans to add block storage and Kubernetes-based container service.
Co-founders have experience with Citus Data and Azure, and the company recently raised a $16 million seed round.
I was just about to say, isn't this just OpenStack?
I don't even think OpenStack is needlessly complicated.
Yes, it is complicated, but who thinks operating a cloud environment the equivalent of AWS is trivial?
And OpenStack is a mature open source project, tried, true, used in all sorts of data centers large and small, supported by multiple vendors. I'd take it any day before 10 VC-funded guys' project. I mean good for them for skinning that cat and if it gains a real community, I might bite.
Openstack is an orchestration platform, but doesn't deal with the lower-level things directly like IAM, fabric management, resource federation...etc. It sounds like they want to release FOSS tools which do that work. The CoreOS team was actually going to do something similar before being acquired.
Ubicloud seems to be reselling Hetzner with a 60% profit margin.
Each VM has a monthly egress quota of 0.625TB per 2 vCPUs. If you exceed this limit, we will charge you $3 per TB of egress traffic. We don't charge for ingress traffic or traffic between your resources within the same cloud region.
Hetzner has a limit of 20 TB, this means Ubicloud has an initial potential profit margin of 15 TB (45$) in the worst case (standard-16 plan), and even better margins for the cheaper plans. Hetzner charges €1.21 for additional traffic outside their limit, this means Ubicloud still has a margin of 56% for additional traffic outside of the Hetzner limit.
It seems like this could be quite profitable if they manage to create an easy-to-use platform that attracts companies. The 16m investment might be worth it.
What an awful name: I immediately thought it might be related to Ubisoft and had to look it up to sanity check my feeling of absolute revulsion. Such a good idea but the sentiment of that name won't do it any favors.
There's a lot more to a tier 1 hyperscaler like AWS, Azure, and GCP compared to the lower tier providers and its not ONLY on price. However here's the pricing bit I was interested in.
"Each VM has a monthly egress quota of 0.625TB per 2 vCPUs. If you exceed this limit, we will charge you $3 per TB of egress traffic. We don't charge for ingress traffic or traffic between your resources within the same cloud region."
AWS is software. Just not something you can self host.
There already exist alternatives to AWS, like localstack, a local AWS for testing purposes, or the more mature openstack, which is designed for essentially running your own AWS at scale.
AWS offers a collection of services made from many systems running all sorts of different software working together. Which piece of software do they plan to make? Are they trying to make provisioning management software? Cause that’s called AWS Management Console. Are they trying to make compute resource provisioning and scaling software? Cause that’s called Amazon EC2. They can’t possibly think they’re going to recreate everything AWS offers.
Most of cloud service provider revenue comes from basic services, like storage and basic compute. But
vendors like AWS, Microsoft, and Oracle figured out a long time ago the smaller, niche services are what differentiates them and makes the services sticky. If you need just a Linux instance, it's easy to provision it using Terraform or Pulumi and jump to the cheapest lowest-common-provider.
But with services like IOT, AI/ML, business forecasting, robotics, etc. Once you tie your business to those services, it's a lot harder to leave.
This includes a managed service and an open source version that allows developers to build their own cloud on bare-metal providers.
Citus Data was part of the Y Combinator summer 2011 cohort and after his time at Microsoft, Cubukcu went back to the accelerator as a visiting partner in 2023.
The company is also seeing a number of users who spin up machines for long-running workloads on the platform, as well as solid demand for its PostgreSQL database service.
Erdogan stressed that a Kubernetes platform is also on the way, which in turn will allow the team to offer more features on top of its infrastructure-as-a-service layer.
Ever since the dawn of hyberscale clouds, we’ve seen projects that aimed to provide the core AWS services, mostly for on-premises use cases.
It is open, yes, but then it supports 10 different flavors of operating systems and hypervisors — and in order to make it work, you really need an army of people.
The original article contains 781 words, the summary contains 161 words. Saved 79%. I'm a bot and I'm open source!