At vshosting~, we run a huge number of physical servers, both for our cloud infrastructure and for dedicated client servers. The most used brands in our data center include SuperMicro and DELL. This year, however, in connection with the preparation of the project of a new data centre, we started to develop our own hardware. Other large companies such as Google, Facebook, OVH, and dozens of others have taken this step ahead of us.
What exactly will the hardware look like?
It will be the development of its own chassis of special dimensions with its own power supply from 48VDC to the standard ATX power supply. For maximum reliability, we will use ready-made memories, motherboards, and processors.
1. Maximum adaptation for the new data centre project
The new data centre does not count in most areas with the standard 19-inch racks for servers, as it is in other data centres. Special multi-row shelves will be created for the data halls, with hundreds of pieces of servers placed on each shelf, which will be arranged for maximum efficiency.
Each shelf will be air-separated and sealed and will have its own redundant fans to cool otherwise passively cooled servers. The shelves will always have two central 48V DC sources (branch A and B) with distribution to each server.
2. Cost reduction
The savings in acquisition costs are significant, as there is no need to pay the manufacturer for their "branded server". Savings are also on the side of racks and server resources. From the point of view of operational efficiency, there will also be huge savings in operating costs due to minimal resource losses and a combination of indirect free-cooling and adabiate direct free-cooling, which will ensure operation without the need for less economical compressor cooling.
The selection of the data centre location with an ideal ratio of the lowest summer temperatures and at the same time the lowest humidity, as well as the construction and orientation of the building, will be adapted to the maximum operational efficiency.
3. Simple (= fast) service
Remotely shutting down a faulty server and pulling it off the shelf without the need for tools will be a matter of minutes. We are also toying with the idea of the future connection of robots for automatic assembly and replacement of defective servers for the possibility of unattended and fully automated operation of the DC2 data center.
4. No vendor lock-in
We will not be limited by any manufacturer who would dictate to us until when the given type of server is supported, we do not have to adapt to the service programs or the list of supported additional components. We will adapt everything to our needs.
Damir Špoljarič
CEO