In St. Petersburg - transporting clouds

Text navigation:

→ The goal is clear, the outcome is clear

How did we get to this life?

A year ago, in the fall of 2023, we opened the site in St. Petersburg on the basis of the data center of the operator OBIT. This was the first experience of opening a location for expanding a public cloud. The launch carried commercial, product and marketing initiatives and meanings. We learned how to build the entire cloud infrastructure from scratch and organized new communication channels.

The launch did not go very smoothly, deadlines were missed due to delays in the supply of equipment, insufficient planning, quality of work performed, and many other unaccounted for issues. Nevertheless, we successfully completed this task and drew conclusions so as not to repeat the mistakes. Over the course of the year, several thousand of our clients have used the platform, and currently we have about 1,300 active services in this location.

And then we planned expansion into other regions, and it goes well.

At some point, it even began to seem that construction and moving would soon end. But as they say, if you want to make God laugh, tell him about your plans. Less than a year later, we learned about the upcoming development of the territory where the site with the public cloud location is located in the St. Petersburg region. I had to look for a new one.

And again I visited…

There was no agony of choice. We were one of the first to visit “Data Center No. 1” from OBIT (we talked about this Here). We realized that this is a modern site corresponding to the Tier III level: with redundant diesel generator systems according to the 2N scheme, redundant UPS according to the 2N+1 scheme and N+1 cooling systems for each turbine hall.

At the new site there were no problems connecting 7 kW racks, which is practically an industry standard. Equipment becomes more productive, the density of components placed in a single server increases – such equipment requires a responsible attitude to cooling, placement design and energy consumption. All this is fully respected here.

In addition, the OBIT management immediately met us halfway and provided us with all the necessary additional buffer capacity and communication channels, which greatly facilitated the immediate move.

The goal is clear, the outcome is clear

Our task was formulated simply – we needed to move the entire region online.

And that meant:

purchase and install a new communication node (routers, aggregation switches and TORs);
purchase and place buffer equipment to accommodate clients (compute) on the new site (all three of our types – standard, high-performance and high-frequency servers);
place a new control plane on a new site (why – see bonus track);
organize, configure, stretch L2 cloud segments into two data centers;
set up the process of continuous migration → movement → equipment → switching → refilling and so on in a circle;
switch telecom operators from one site to another.

As a result, moving an entire public cloud location took us two months of work at a relaxed pace, along with all the preparation. When we first heard about the need to move, we thought the task would stretch over more than a block.

All location services migrated online, without restarts or stops in sales. We tried very hard so that our clients would not notice the move and we will try to continue to do so =)