My name is Leonid, I am a website developer , therefore, due to my activity, I am interested in the history of the formation and development of various companies from the field of hosting services. Today I want to present an interview with Danil and Dmitry - the creators of hosting . They will talk about the structure of the infrastructure, organization of work and their experience in developing a virtual server provider in Russia.
Please tell us a few words about yourself. How did you get into hosting? What were you doing before?
Until 2016, both Dmitry and I worked in the Enterprise sector, including companies such as Dell, HP, EMC. Analyzing the cloud market in Russia, we realized that it is actively growing, and decided that we could make an interesting offer to the market. A team of people who had already worked with each other on other projects got together and together they began to develop their own virtualization platform, aimed at large businesses with their specific needs. Since 2018, we have simultaneously launched cloud hosting “for everyone”, allocated for the project a team of five people.
Rack storage and preparation before the start
Is this project for business already working or still in development?
Yes, it works in parallel - there is already a larger team, and we are talking more about software and hardware solutions for IT infrastructure, not about hosting.
You now have quite a lot of different services. When you started, was the list smaller or the same? Moreover, all these services are actually a regular virtual server, but there is a certain separation.
We started with the classic IaaS: we provided “bare” virtual servers with closed ports and virtual networks for them, so that the user himself could create a full-fledged infrastructure for himself. But after the launch, it turned out that most users did not understand why they needed such opportunities, and we decided to introduce a new product for ourselves - a standard VDS / VPS, which the market is already familiar with. For us, it was, in fact, a stripped-down version of the product, but users immediately understood what it was, and we began to get the first customers. Apparently, our experience with large companies forced us to immediately develop a more complex and customized solution, and the mass market wants simplicity. And further on the basis of VPS, we began to develop new services, based on what customers most often ask for. And we are still developing.
Where do you place the equipment? Does it belong to you or do you rent it? How did you choose a DC for placement? Have there been relocations?
All equipment is ours, we only rent space in two data centers. We started with three DCs: we wanted to implement three-way fault tolerance, but the demand for it at that moment was too small to invest in it, so the third data center was abandoned. We had one move: we were just moving from the third data center to one of the remaining two. They were chosen according to this principle: DCs should be known on the market, reliable (Tier III), so that both are geographically located in Moscow, in areas remote from each other.
And in which DCs are you currently located and which one have you abandoned?
Now we are located in DataSpace and 3Data. Abandoned one of the 3Data data centers.
Leaving the third data center
Do you rent or buy IP addresses?
We rent.
And for what reason did you choose this approach, and not a purchase?
By and large, to grow quickly. We provide clients with a virtual infrastructure that does not require immediate capital investment and can be broken down monthly. We ourselves adhere to the same philosophy as our clients - we strive for expansion and rapid scaling.
What do you think about IPv6?
So far, we haven’t noticed significant demand, so we haven’t added it to ourselves, but the architecture for the conclusion has been worked out, we are ready to “roll it out” in a short time, as soon as we understand that there are requests.
You are using KVM virtualization. Why was she chosen? How does she perform at work?
That's right, but we are not using a "bare" KVM, but a full-fledged, modified KVM-based virtualization system that our "big brother" developed, including a storage system (SDS) and a software-defined network (SDN). We chose on the basis of building the most fault-tolerant product without single points of failure. It shows itself well, so far there have been no global problems in the production. At the stage of alpha testing on the market, when we provided services to the first customers for bonus points, we tested the technology and caught a number of unpleasant moments, but over the past two years we managed to understand and solve a lot.
Do you use overselling? How do you control server load?
We use overselling only for processors, but in no case for RAM. Even in the case of physical processors, we do not allow their load to exceed 75%. By disks: we work with “thin” allocation of capacity. We have centralized monitoring of the entire environment, which allows us to control the load. Two engineers are responsible for supporting the entire infrastructure, so we are trying to automate as much as possible and collect all kinds of information on the system. Any deviations from the normal mode are immediately visible, while we periodically evaluate and rebalance the load within the infrastructure. Rebalancing always takes place online, transparent to customers.
How many physical servers do you currently have? How often do you add new ones? What servers do you use?
Currently 76 servers, new ones are added approximately every four to five months. We use QCT, Intel, Supermicro.
Have there been times when a client came and occupied all the remaining free resources, and you had to urgently add servers?
There were no such resources. While growing more or less evenly. But there was a case when a user came and wanted 50 IPs, each in a separate block. Of course, we don't have anything like this yet 🙂
What are your most popular payment methods? What is the least used?
The most popular are a bank card and QIWI. The least of all is payment by bank transfer under an offer for legal entities, but such transfers are the most voluminous (companies usually pay for solid resources for several months). PayPal is also lagging behind: at the start, we did not count on foreign users, but they began to appear.
Self-written billing in Boodet.online. Why did you choose to use this solution? What are the pros and cons? Was it difficult to develop?
In general, we have the entire system - our own design. The existing platforms did not seem very convenient to us in terms of UX, so we decided to create and develop our own. Billing is just one of the microservices that are part of the system. It turned out to be more difficult to develop than we thought at the start. Even at some point, I had to postpone the launch of the project in order to have time to prepare a working product that would not be embarrassing for in alpha testing. Subsequently, they “overgrown” with competencies in the development methodology for a long distance and product management. Now it is easier to add new functionality and new products to the system.
And how many people developed all this? What did they write on?
We have five people for the whole project, two of them are developers (frontend and backend). Back is written in RoR/Python. Front is JS.
How is user support organized? Is it open XNUMX/XNUMX or only during business hours? How many support lines are there? What is the most frequently asked question?
We have three entry points: chat, phone and application system from your personal account. Two lines of support: if the engineer on duty could not solve the problem, the technical director or development team is connected. If the problem is in the main platform, which happens much less often, then the technical director turns to the support of the "big brother". At night, we only respond to calls from customers who buy separate maintenance, or to platform failures reported through a specially written Telegram bot.
Top Questions:
- Are our IPs available in Turkmenistan (this is the very first in popularity - apparently, the country has a strict blocking policy).
- How to install this or that software.
- How to get root access (even a special memo was made in the interface when creating machines, but it does not always help).
Do you verify clients? Do spammers and other bad personalities often appear?
Verification by mail and phone (if the user activates 2FA). Spammers and other abusers appear periodically. We are forced to respond by temporarily blocking compromised servers, as we do not want IPs to be blacklisted. But we always write to the user in advance that a complaint has come to him, please contact and discuss the problem. If the user does not respond, or there are repeated complaints, we block the entire account and delete the servers.
Are there frequent DDoS attacks on clients? What do you do in such cases? Have there been attacks specifically on you, on the site or infrastructure?
Clients are rarely attacked. But often we ourselves are a website, a personal account. Sometimes they hammer the network at different IP addresses. We do not undertake to judge who it is and why, there may be several options. There are even attempts to attack us from within. Previously, when verifying by phone, we gave out a bonus hundred rubles so that normal users could test any configuration. But one day a user came with a “package of SIM cards” and from under one IP began to create dozens of accounts, receiving bonuses on them. Therefore, we had to remove the automatic accrual of test points. Now you need to submit a request to technical support for testing, and we consider each case separately.
How is work organized, is there an office, or is everyone working remotely?
There is an office, but with the start of restrictions due to the coronavirus, everyone went to work from home / dacha / hometown.
Our office
What is your current course of development of the company?
We are going to add new services. We have an extensive roadmap, we do not interrupt development, and every two weeks a new release of the personal account is released. We add functionality and services that are in demand among colleagues, we add what customers ask for.
How do you find clients? Is there a large influx and outflow of customers per year? What is the average "lifetime" of a customer?
Channels for attracting customers in our field are what the whole business rests on if there is a working good product. Therefore, they are not ready to share.
Churn rate, LTV and life cycle are also quite important indicators that we use only for internal analytics, but not for disclosure.
Could you give readers any advice on choosing a web host? What should you pay attention to before buying?
The most important thing is to choose a hosting with the letter “B” at the beginning of the name.
Seriously though, there are a few factors to consider:
- To understand the quality, you can take an average configuration and try to solve your applied problems on it. Choose hostings that have an hourly rate - you can test the servers without losing a lot of money if the quality does not suit you.
- Look at the data centers where the host has physical servers. They can be roughly judged by the quality of services.
- We do not advise you to pay attention to prices: there are both super-cheap solutions, but they show themselves well, and super-expensive ones that do not differ in anything special.
Tell us about your most memorable work moments.
Project start. For the first month and a half, we worked 24/7: we looked at how registrations were going, if something was broken in the interface of the personal account, how users behave, whether it was convenient for them to order services. Many things had to be decided right on the go, up to replacing one product with another. Changes were made immediately in the production, bypassing the test environments. It was a stressful period, but I managed to survive and not give up this business.
Users who came for vulnerabilities in logic. It was interesting to catch them and close vulnerabilities. For example, when we still did not work for money, but accrued bonuses so that users could order servers, a link to us was posted on one of the hacker forums with the comment: “they give free servers for 500 rubles.” Of course, we were immediately filled with miner guys hungry for freebies.
Could you make a short timeline of the company's history?
- The first half of 2017 - we began to develop the platform, website and personal account Boodet.online.
- 2018 - went into alpha testing, provided capacity to customers for free and received detailed feedback and test results in return.
- Mid 2018 - launched a beta version with money. The first hundreds of customers, running technical support.
- 2019 - began to attract legal entities as clients, work on custom solutions.
- 2020 - everyone goes into self-isolation, the demand for virtualization is growing. We feel it ourselves - there is an increase in customers, which makes it possible to work on a large number of additional services.
Source: habr.com