They were the first

MLIT is ready for any challenges

Starts at No. 46 from 12/7/2023

In the 1980s and 1990s, many processors based on the Unix platform appeared. CERN, DESY and other nuclear physics centres were equipped with processors from Sun, HP, DEC, IBM, SGI and others. 7-8 groups worked at CERN that adapted the CERNLIB library for each platform. And we adapted Cernov programmes for Windows and Linux that were installed on our personal computers. This was our initiative. At some point, I went to CERN and offered to sign an agreement so that we could do this officially. And they were surprised: why is this necessary? Who needs a personal computer as a tool for analyzing scientific data? Nevertheless, they agreed to give us a new version of CERNLIB every three months so that we could adapt it and make it freely available. These two platforms are included in the Cernov library, but all questions and advice on use should be addressed to us. We did this for several years, but life showed that after 10 years all the platforms supported at CERN began to become history and the Windows and Linux platforms became the main ones at CERN and they told us: "Thank you! Then we are on our own." We can say that the Cernov software and the culture of working in Linux in general were purely our doing. We did this at our own peril and risk, but for 25 years this software has been the main thing not only at CERN, but also in world high-energy physics and in science, in general.

In the late 1990s, the level of communications increased, speeds became gigabit. At this time, they began to think about a computing model for the Large Hadron Collider that would be launched in 2005. It was believed that software for the LHC could not be developed in the same way as it had been developed before - by physicists, mathematicians, that is, by non-professionals. You need to take the best software from large professional companies and rely on it. Rene Bran, just like N.N.Govorun in his time, opposed everyone. He said that no large companies will make software for the LHC; they live off the replication of their products. CERN refused the services of Rene, so he developed the ROOT package that became the main one for the LHC and for all high-energy physics with a small team that included our employee. He was again invited to work at CERN, but before that, for five years at the European Centre they had tried unsuccessfully to integrate various acquired professional software. They hadn't succeeded.

Grid technologies have appeared. Ian Foster and Carl Kesselman at Argonne National Laboratory developed the first Globus Toolkit that allows the resources of geographically distributed clusters to be combined into a single environment in order to jointly meet common issues. It was decided that grid technologies would become the main concept of computing at the LHC and both Dubna and Russian centres, headed by CERN, began to work on this. And then there were many European projects in which both we and the United States participated - thus, a geographically distributed computer infrastructure was developed, in which we currently work. In 2003, a Tier2 level centre was established at MLIT; it became a very important element of computing at the LHC; 10 years later, a decision was made to establish a Tier1 centre on the basis of the National Research Center "Kurchatov Institute" and JINR. This is also a significant event - establishment of two powerful centres at JINR, in Russia, not just for processing, but also for participating in storing data with one hundred percent reliability. We are responsible for a very large section of physics at the LHC, for almost all the data from the CMS experiment goes to us for storage, more than 100 centres download this data for processing and analysis and we also store the simulation results. Our Tier1 is already a servant of not two, but four masters. We perform four asynchronous tasks: everything that is given to us from CERN, we should obtain and store; provide data to all centres that request it, including our Tier2; save simulation results; ensure processing of tasks on Tier1 in 24x7x365 mode. Accomplishing these tasks is very important and we have acquired a culture of work at all levels. Therefore, when we began to implement our large NICA project, we already more or less understood what we would do. Without this experience, I think we would not know how to organize all the stages: data accumulation, storage, processing, analysis.

Today, the developments that were made at the LHC are currently used and a lot has been done there with our participation, especially in the field of monitoring. Our team participated in projects to develop data transfer monitoring, service management, grid integration, cloud technologies, to create an environment for using supercomputer resources and other tasks in the development of the Worldwide LHC Computing Grid (WLCG) architecture and experiments at the LHC. Unfortunately, there are very few such specialists in the world; in most countries there are none at all.

No one trains specialists in geographically distributed scientific computing. Universities train specialists in areas where ready-made solutions are available. Here all the solutions, models, software are very specific. Of course, there are always enthusiasts in some universities thst are ready to establish departments, but this is rather the exception. At Dubna University, we are the same exception, we prepare our students and invite students from other universities, they learn to work in this environment, but if they go to another field, unique skills are lost. They acquire a completely different breadth of views, because in addition to commercial solutions from Google and Microsoft, there are solutions from "amateurs" in this matter that may be inferior in quality, but in architecture and scope are incomparable with commercial ones.

No one will undertake to combine grid clouds, clusters and supercomputers into a single environment, because this is an extremely difficult task. But we know how to do this and we believe that this is the prospect of development, including for Russian science. For all megascience projects in Russia, it is necessary to develop a similar computer infrastructure, in which good storage, security and data transmission systems should be organized. And everything should be well integrated so that users from different scientific groups can communicate with each other while working on common projects. This task is extremely difficult, but without meeting it, that will be difficult for Russian science to develop. Our joint efforts should be aimed at creating a distributed environment for large scientific projects and involving the majority of research centres and universities in it. All new trends: big data analytics, quantum technologies, photonic technologies, artificial intelligence and many others will largely depend on the solution to this global problem.

When this Seven-Year Plan was drawn up, we discussed the future, but I think there are hardly many specialists that can say what will happen in seven years in the IT field. For two or three years it is more or less possible to predict, for five years it is more difficult. In the second and third years, in the two past seven-year plans, we did something that was not included in them: in the first seven-year plan, we established a Tier1 centre that was not planned, in the second, we introduced the Govorun supercomputer that was also not planned either. I can't yet say what will happen in the new seven-year plan. Of course, universal computers will remain and specialized ones will be added to them, this is obvious. Today, there are ordinary computers and some with graphics processors that are good at meeting a certain class of tasks. Quantum computers will appear; at present, they are not commercially available and do not meet all the requirements, but in seven or eight years the technology will change. Photonic computers are currently developed; they will be coprocessors for universal ones, as well as biologically based computers. I think different areas and technologies will develop, artificial intelligence has eclipsed everything else, maybe this is a temporary phenomenon, but it is clear that machines will be more and more universal. They already say that artificial intelligence should be limited; some kind of ethical standards are needed in any case. As someone that stood at the origins of the Internet in Russia, I was once asked whether it brought benefit or harm to the country? When we were doing this, we thought only about the benefits, not understanding what it could turn into. Certain ethical standards should be developed, but this is too difficult, just as the number of people that want to use the Internet for bad reasons is too large. It's especially scary when very young children get stuck on their phones, stop communicating face-to-face and are drawn into the virtual world without knowing the real one. This seems scarier to me than hacker attacks and Internet scammers. Who could have predicted all this 20-30 years ago? I think common sense and ethical standards will force the majority to find reasonable forms and methods of work.

We would like to develop not only a Russian infrastructure for our megascience projects, but also to involve most countries in it, primarily the JINR Member States. This is an extremely important task for us. This year alone we opened two cloud computing centres - in Tashkent and Almaty and a centre was opened in Cairo. When we opened the centre in Almaty, I noted that not only the resources of the scientific centres of Kazakhstan for our tasks are important to us, but it is important for us that a community is developed around the infrastructure of the republic and that the tasks needed by Kazakhstan are met. If the cloud resources are not enough, there is somewhere to transfer tasks, where to store data, etc. This allows the JINR Member States to have large resources in Tashkent, Cairo, Almaty. Of course, we understand that in addition to megascience projects currently implemented in Russia, there are projects currently implemented in other countries. First of all, this is China with one of the world's largest neutrino projects JUNO, the participants of which turned to us to establish a Tier1 centre for data storage in Dubna. An agreement has already been signed and work has begun. I think we will attract other states, our potential partners, to this distributed infrastructure. We do not abandon the infrastructure for the LHC. I think everything will make sense soon. I was just at CERN and the WLCG project manager believes that our cooperation will continue no matter what.

Meanwhile, we currently expand our tape library from 40 petabytes to 90. Together with disk, our storage will be approximately 150 petabytes at the beginning of next year. This is more than Yandex, Mail.ru, State Services and Sberbank all together. This wealth should be used primarily for megascience projects, but not only. In principle, we can provide storage resources for a lot of projects. Data cannot be lost; the culture of working with it is the most important thing today. And at present, few people know how to organize such work. Even at CERN, despite their enormous work experience, there were problems. The results of experiments on which millions, even billions of dollars have been spent, are priceless. I know that we have nowhere to store the data that is obtained by sensing the Earth from satellites. So, at the recent conference at the Russian Academy of Sciences, I said that all unique data should be preserved at all costs. And we are ready to participate in it.

Of course, to unite geographically distributed centres, high-speed communication channels are required and their development should be a priority in developing a national scientific infrastructure in order to eliminate the digital divide between the country's regions. This is very important for the development of communications with the JINR Member States that we widely work on.

I think we will always be in trend, we are ready for any challenges.

Olga TARANTINA,
photo from the MLIT archive