Interview with Sergey Belousov: really big data / Company Blog Hacker Magazine

XakepRU February 13, 2014 at 09:44

Interview with Sergey Belousov: really big data

Interviewed: Stepan Ilyin.

I met with Sergey Belousov at eight in the morning in a Moscow hotel. The concept of a house is out of date for him: why do we need houses and apartments, if there are hotels? Perhaps it was precisely this sharp and innovative thinking that helped him build several technology companies at once in completely different fields: from developing a Linux distribution (ASP Linux) and virtualization technologies (Parallels) to data storage and backup systems. At some point, the main focus was Runa Capital Venture Fund, but now its focus has shifted to one of the long-established companies - Acronis.

Return to Acronis

For many years I did not deal with Acronis, but now I suddenly had to. Acronis has changed management teams several times over the years, but lately they have not done very well, and the company has not grown as fast as it could. So, in my opinion, the company can easily grow ten times or more. So I decided to return to Acronis.

Any project has a certain internal property, like human DNA. List of requirements, specifications, tasks. This comes out of context.

Facts:

Hard worker. It works on average 85-100 hours a week.
Investor. He invested in several dozen projects, being a senior partner of the Runa Capital venture fund, including Igor Sysoev, nginx.
Physicist. A complete set: a graduate of the Physics and Technology Institute, Ph.D., Honorary Doctor of MIPT and holder of more than two hundred patents.
Traveler. Since 2001, a citizen of Singapore, which is very convenient for traveling around the world on endless business trips: Singaporeans do not need visas in almost all countries of the world.
Businessman. He founded more than 20 companies, including Acumatica and SWSoft, which has turned into two world-famous companies - Parallels.

Context is the environment. For example, if we settle with our DNA on Mars, we will live there for a few seconds, or even less. Because there is a different context.

Management for a project is like intelligence or the central nervous system. If everything is in order with intelligence, then you will not fly to Mars.

Acronis was not in its context, could not function normally. Management brought her into such a situation. Now the project needs to be returned from Mars to Earth. Nevertheless, it is still a stable and profitable company.

One part of Acronis DNA has always been a fairly strong team.She stayed now, just a little “smeared”, but we almost gathered her back. This is a group of people who are well versed in storage systems. Acronis now needs ultra-sharpened technology: ultra-fast, ultra-efficient, ultra-reliable.

Do you know how big the Swiss watch industry is? 25 billion dollars. And Swiss watches always advertise on yachts, in space and so on. But you buy them and, of course, never drive a yacht with them in a storm and do not dive to a depth of 200 meters. But it's nice that you have such an opportunity.

To make this real, Swiss engineers spend time and effort.If they claim that their watches work at a depth of 200 meters or in space, that’s true. Acronis has also always made technology of this kind. But now this is just one of those DNA components that has broken down.

A project manager can break the DNA of a project.The team that was in Acronis before did not set out to sharpen technology for complex tasks. If we continue with the watch example, the manager can decide: “Oh, we are Casio, why do we have to have good technology? After all, we sell cheap watches, let's save! We will buy bad displays, quartz oscillators, we will not worry about software, we will not add any properties there. Who cares what kind of cases we have? ”That is, Casio is also a good watch, but it’s not perfect technology. We want to be a Swiss watch. Accordingly, now we are returning everything to the place.

Takeoff storage systems

The focus of innovation moves from time to time. Sometimes the focus shifts to web applications, sometimes it goes back to the platforms. Now this center has moved to data storage systems. That is, a huge number of startups, new technologies have arisen that are associated with the simple fact that the amount of data is growing rapidly.

This is not to say that we have missed any trends. Software companies, like living creatures, have this property - a very high level of survivability, a large margin of safety. Say killing a person is not easy. He is tenacious, he needs to be tormented for a long time.

When Acronis started, we developed very good storage technologies, but storage was not a trend.And now this is a very big trend. After all, the amount of data is growing infinitely fast. Personal data, backups, analytics - all this must be stored somewhere. In particular, that’s why it was interesting to me to return to Acronis and work on this task.

The whole world that we know turns into data: they become incredibly valuable and important. I think this is only a matter of time, when the value of the data that you generate on a computer in a lifetime will exceed your own value.

Singapore is going to sequence all of its inhabitants. The population of the country, taking into account the expatriates,
about six million people. One sequenced DNA, even after optimization, takes up 22 GB. Six million is some incredible amount of data: 6000 * 22 Pb. The storage system for such data is a very complex system. And this is only in small Singapore. And in the end, obviously, every person in the world is sequenced. And not only the people themselves, but also all their viruses and so on.

All storage systems have changed a lot. Both Makovskaya HFS and Windows NTFS will soon cease to be relevant. Some time ago, file systems seem to have stopped evolving altogether. And again, the trend begins to change.

Large amounts of data

Now we are launching Acronis Storage technology, which was previously used only in our own data centers. We developed it five years ago - and so far only used it inside the company. Its main properties are very high reliability, scalability and efficiency.

On large volumes of data, very strange effects begin to appear.For example, cosmic particles, I'm not joking. Each data system of some size has a section. From time to time particles come from outer space that can fall into bits. If they get to the very right place of the bit, it can be turned over. This happens quite rarely, but the more data you have, the more often it happens. On some systems, this happens almost all the time. For example, if you have a million petabytes, then this is a constant situation. The storage system must be resistant to such, it should not break. Therefore, reliability of a different kind is required there.

Scalability is also very important.Even trying to find a block or file on a very large system can be a very expensive operation. Say, in a million petabyte system, the file table itself can be the size of a petabyte. And to load it into memory to read is impossible. While there is no memory in petabytes in volume. Such a table needs to be read in a completely different way. And store too.

Acronis products are sold in more than 90 countries and translated into 14 languages.

Efficiency is just as important. With so much data, even small changes in efficiency are of great value. For example, a storage system can cost one hundred million dollars. Then a 10% change in its effectiveness is a gain of ten million.

The level of errors in the storage system should be ten times less than in operating systems. The OS may crash, something may go wrong, but it’s not scary - you can reboot, update. And if data is lost, everything is irreversible. Therefore - a low level of errors and a high level of quality.

HR issues

Acronis can preserve its DNA and evolve only by making high-quality technologies. And only super-advanced people who fit in well with the team are capable of this. The degree of reliability must be very large, and only very high-quality programmers can develop such systems.

All developers and R&D are located in Russia. Recently, we discussed with the Ernst & Young branch in Israel, which specializes in the audit of technology companies, are there any programmers in Russia and how many are there? But the whole idea is that a lot is not needed. It takes only a few - but extremely high quality.

The company has offices in 18 countries, including the USA (Boston), Germany (Munich), Japan (Tokyo), France (Paris), Great Britain (London), Belgium (Brussels), Singapore.

Of course, there are more programmers in India or China, but this is not the main thing - quality is important. Quality is key
, for example, the same nginx was written by one person - Igor Sysoev. And Acronis core technology was written by five people, and they are still developing it. We really need to find five more of the same. The Russian-speaking population is more than 200 million people, including the CIS countries. Surely five people can somehow scrape together.

In Russia, we are looking for people who will be architects, product managers, program managers, and mainly senior programmers.That is, people who will do complex pieces of code. Not from Russia, we attract people with business expertise. These are either managers of large teams, or those who have a good understanding of product or user experience (which are closer to the client).

If a person decided to live in Russia, then there are few teams capable of giving him the opportunity to write software that will be used by one hundred million people. On the fingers of two hands. Acronis is definitely one of the best teams, and now we with Stanislav Protasov, Acronis senior vice president of research and development, will make her the best team. It is more pleasant to write a storage system that is used by hundreds of millions of people than, for example, software for even the largest bank.

Now Acronis employs about 650 people, almost half of whom are developers with experience working in software giant companies.

About my role

For a long time I have not written code, but I am directly involved in the development of complex technologies. It's like conducting an orchestra. When you conduct an orchestra ... do you know how to play a good instrument yourself? You know how, but you never play. The main task is to somehow conduct a complex orchestra to get a consistent signal. Basically this is what I do.

Someone must conduct. In the most successful companies, this is always one person: at Microsoft it was Bill Gates, on Facebook Mark Zuckerberg. Mark does not write code, but he conducts. Someone should take this role. It's like a brain transplant. Probably, if it weren’t me, someone else should have taken the place. But then it was supposed to be his company. Since my company is mine.

The other person will not have so much motivation and knowledge of the details so that he can conduct the orchestra. Therefore, technology has to climb. But it’s in sophisticated technology, only this is really interesting to me.

Trends

From my point of view, for the technical audience there are now five promising areas in complex technologies.

Large, mass storage systems.This is a powerful trend that many startups pick up. Many of those who previously developed older platforms have moved on to developing storage systems. Yes, these are rather boring systems, but this is a very large area: there is a lot of data and they are gradually moving to the cloud. As I said, the focus of technology is shifting. Now it has shifted again, because security, various communication algorithms, and reliability have become important. Reliability is imposed on everything, even on data transmission. When a very large amount of data is transmitted, the errors present in any protocol begin to emerge stronger. If you have one hundred thousand users and 0.01% of your users are unhappy, then this is not a problem - it is only a hundred people. And if you have one hundred million users, and not one hundred thousand, then not one hundred people are dissatisfied, but one hundred thousand.

Big Data Storage systems and the Big Data area are connected, but not directly. If you store a lot of data, it’s not necessary to conduct some kind of extremely complex analytics with them. However, this is usually done.

Virtual worlds and systems for their creation.A huge number of people live in virtual space more and more. Today's systems for creating virtual worlds are games, interactive entertainment or learning systems, virtual worlds as such. But for now, all this is handicraft development. There are some platforms like Kinect, but they are not quite suitable for virtual worlds. They are all proprietary. Reminds me of the times before the Internet, when there was AOL, CompuServe, MSN - each vendor has its own platform. For a developer who wants to create his own virtual world, it will take a lot of platforms and knowledge. This is not very convenient. An ideal way to create a virtual world is when you just sit and say: “I would like everything to be like in Moscow, but the houses were like in London, the air temperature is always 30 degrees, it would rain for two days ... ”That is, if one could tell the script - and all this would appear. To date, this does not exist.

Robots.Robots are a combination of hardware and software, but software is of primary importance. In man, the main thing is also software. If you turn off software in a person, it begins to be ... a corpse. It is necessary to develop platforms for various robotic systems. There will be a lot of them. The simplest example is automatic cars: soon all or many cars will be without drivers. This also includes all sorts of tracking systems, control systems. It is necessary that it is possible to conveniently program the "mind" of these systems, their behavior. Now there are many different systems, and they are quite unsuitable for real use. Therefore, developers use systems that are not written for this at all - platform, storage systems, OS, and so on. One of the problems of robots is real-time communication among themselves. Now their communication systems are very slow, inefficient. If they are connected to each other, then everything is not bad. And if they are under water, where there is no Wi-Fi and the data transfer system only draws out 1 Kb / s? If you have a group of robots, it will be very difficult to ensure that they behave like a school of fish. Fishes see each other, and the interface between them is not 1 Kb / s, but, say, 1 Mb / s.

Internet of things. Simply put, the internet is everything. A huge number of things are now connected to the Internet. Usually, when people talk about the Internet of things, they think about some kind of refrigerators connected to the Network. But Internet of things is not necessarily about that. This also includes alarm systems, data management, temperature control systems, as well as cars and robots that must communicate with each other through some kind of network.

Acronis company’s blog on Habr.