It used to be the norm to store data on your own servers – now businesses are increasingly migrating to the cloud. But any changes, especially for large enterprises, are not easy, although they benefit the company. Here’s an example: the online hypermarket 21vek.by abandoned hardware servers and migrated to the cloud in two weeks.
What does the digital transformation of a large e-commerce company look like from the inside and what does it give? Sergey Kontorovich, director of IT infrastructure at 21vek.by, shared his experience with MTS in a partnership project.
– It is important for us to constantly develop and improve our product. And the speed of the site and the convenience of shopping is one of the main factors affecting the loyalty of 21vek.by customers. This approach gives its results, including increased load on the system. The total audience of 21vek.by is 2.2 million people, and, for example, during “Black Friday” last year, our site had more than 500 thousand visitors at the same time – the load reached 30 thousand requests per minute.
We used to fully rely on the support of the previous hosting provider and their fleet of physical servers, but they could no longer provide such power and scalability. In 2020, we decided to move all infrastructure to the “cloud”. I will share my experience as it was.
Why did we decide that it was time for the “clouds”?
Over time, the provider stopped responding to our needs, especially with the outbreak of the pandemic. There were cases when the duration of server expansion reached six months, and the business cannot wait that long.
We faced many problems:
The entire infrastructure was functioning on “iron” servers, and this is already the last century. The servers were unstable. The download could take 1 or 20 seconds. This caused delays in the site
Technical support often did not meet our requirements. There was no server administration as such, although the contractor must monitor failures and fix them. All problems were solved only at our request and not as quickly as required by modern infrastructure
The previous provider did not have the technical ability to develop and adapt the infrastructure to our request.
We are constantly expanding our assortment, now it is 600 thousand items. It became clear that if we want to remain market leaders, it is time to change something.
We began to research the market and look for a solution where we could manage the infrastructure ourselves, set up the required capacities, and not contact support for every issue. The solution was as follows – to move the infrastructure to the “cloud”. We wanted to find a cloud platform that runs on modern equipment – by analogy with such foreign giants as AWS and Google Cloud.
Why did we choose MTS Cloud?
We analyzed existing solutions on the market and selected the best in terms of “price-technology”. The providers provided us with test access so that our team could try and evaluate the convenience of the interface, the speed, competence and proactivity of the support service with their own hands.
MTS was pleasantly surprised by the approach – we were immediately offered solutions for setting up and transferring the site infrastructure, the support responded to all requests quickly and competently, communication takes place without unnecessary ceremonies and redirects.
How did we migrate to the cloud?
We prepared for the global data transfer for about three months and conducted experiments. This included partially letting users into the new infrastructure and analyzing the results.
First of all, our product images migrated, it was the easiest thing to do, there was no need to disable the site. There are hundreds of millions of pictures, the total size is about 1 Terabyte. MTS Cloud provided a 10 Gigabit channel for free, and synchronization took three days.
Then we had a phased transition, which was divided into 3 stages of testing:
Infrastructure under load.
The compatibility of our applications with the work of the infrastructure.
Support teams from MTS.
The MTS team prepared a trial infrastructure in advance, we “raised” one of our ten servers to the MTS Cloud and tested the work by our specialists for a week, and then tested the work on real clients.
At the end of this stage, we jointly eliminated the identified comments and planned a large transfer of the remaining data. After that, we performed a database migration in one night with a site shutdown, transferred 50% of our infrastructure to MTS. This process took only 6 hours. The rest was already invisible to users.
Of course, we could not foresee all the nuances, since the transfer took place from the “iron” infrastructure to the cloud. We had to rework everything we had: containerize applications, adapt them for the “cloud” and make edits to the site code. It was two weeks when my team and I worked around the clock.
As a result, moving towards heavy loads has turned into dynamic infrastructure scaling in real time. At any time, you could change the number of virtual servers and resources (processors, memory, etc.) – very flexible and convenient. We were completely ready for Black Friday: we could immediately track how much resources we need, add, decrease, and adjust to the load.
And the best part is that we pay only for the resources that we use at the moment. Previously, we were forced to buy hardware and continue to work with it, significantly overpaying.
For example, on Black Friday we scaled up three times (using 800 processor cores and 1200 GB of memory), and then rolled back to our usual configuration. For us, this is an opportunity to boldly develop the company and keep up with the times.
We now plan our business without worrying about infrastructure constraints, and our customers have the ability to shop quickly, conveniently and securely.
How 21vek.by moved to the “cloud” – a view from the MTS Cloud
The main requirements of the customer were high performance, fast scalability, data storage security, and the availability of qualified technical support. After evaluating them, we offered to place the IT infrastructure on the Huawei cloud platform, since this solution has high hardware performance, advanced free functionality and meets customer requirements.
The MTS Cloud team together with 21vek.by conducted preliminary tests, we drew up a migration plan and allocated areas of responsibility.
As a result, it has improved:
Infrastructure scaling speed and limits have grown
External channel speed and communication between VMs increased
Support response time is now 5 to 15 minutes
Convenience of the control panel when creating virtual machines on Huawei – 9 out of 10.
Load and network stability graph before moving (no load):
Load and network stability graph before moving (under load):
What do you need to know when migrating to the cloud?
MTS Cloud recommendations:
Plan your migration ahead of time. At least a few months will take you to prepare and test
Perform an IT infrastructure audit and test load ahead of time to help prepare and identify bottlenecks
It is the transfer of IT infrastructure to outsourcing that gives the business flexibility and dynamics – you can quickly increase or decrease capacity to suit your needs. Cloud IT infrastructure will enable the launch of new services and the development of existing products without capital expenditures.
The more you delay using any modern infrastructure, the more technical debt accumulates.
Decide what you want from the infrastructure, and only then start looking for a contractor. Once you have a plan and a vision, the rest is just tools.
Pay attention to how your partner is accommodating, whether you feel comfortable working together. People decide everything.
While updating the infrastructure is a laborious process, it is absolutely worth it. Every step towards the future makes your product stronger, we have tested this on our own experience. We wish you courage, strength and good partners on your digital transformation journey!