From conception to the first cluster in Amsterdam: How the vStack platform came to be

In the economics of cloud solutions, licensing fees usually account for a significant share. The founders of vStack decided to reduce these costs – thus the idea of a new hyperconverged platform was born.

Selecting the technology stack

July 18, 2018 can be considered the birthday of vStack: then the developers first sketched the structure of the future platform on the board.

“It was clear from the beginning that we couldn’t just make something on the spot. We needed a system that could provide IaaS in both service and on-premise models,” said Evgeny Gavrilov, project manager.

It is worth mentioning the technology stack that formed the basis of vStack. The founders of the project were guided, first of all, by their own experience and decided to refuse KVM: it doesn’t work well with the Overcommit CPU and is distributed under the “inconvenient” for the commercial product GPL license.

In addition, the developers cared about the compactness and “stability” of the technology. For example, if you take Red Hat virtualization, a cloud with redundancy and redundancy would require a huge number of components to be deployed. OpenStack, on the other hand, in the hyperconverged vein, is actually a hack that is not supported by anyone.

As a result, FreeBSD, ZFS and bhyve were chosen as the stack.

Implementation stage

August and September 2018 passed in the search for the first team member, and in December the design and implementation of the cluster layer of the solution began.

In terms of Software Defined Storage, there were no problems, but the networking part of the solution turned out to be much more complicated. It took a lot of time to make changes in the OS, which would have allowed to build private virtual networks with the right characteristics.

To implement the API, the team chose JSON-RPC.

“Over the course of the operation, we’ve seen that it was the right decision. A good half of the requests in our practice ceased to be relevant after the request to provide the request ID: the author himself found the cause of the problem on his side thanks to the end-to-end uniqueness of each request,” added Evgeny Gavrilov, project manager.

Entering the market

Then came the long-awaited moment of release. At this stage, the product was greatly helped by an internal company, Serverspace, a public cloud provider focused on small and medium businesses.

After only 10 months from the start of the first phase, Serverspace was able to use the first cluster on vStack in a production operation.

The first cluster was hosted in Amsterdam and sold for about half the price of VMware. So the project immediately got a pool of early customers and, based on their feedback, began its development.

It continues today: vStack expands its team of specialists and works on creating even better solutions for customers.