48 thousand participants and more than 1 million broadcast views allow you to feel the scope of this event. The technology conference seems to have become a real inspiration for the IT community.
If you are interested in reading about how it was, we recommend the article “From techies – for techies: how I looked into the future at the SmartDev 2023 conference.”
In this article, we have prepared a brief summary of the DevOps track for those who missed the conference: key theses and the hottest questions that, in our opinion, still remain open and which can be continued to be discussed in the comments. We invite you under cat.
The DevOps track consisted of a panel discussion “DevOps in Russia: state and prospects” and five reports:
Creation of a unified environment for DevOps processes of any complexity in conditions of constant scaling and growing number of projects.
Clickhouse as a backend for Prometheus.
What should a Russian web service be like for hosting IT projects?
Kubernetes platforms: doing good and doing good.
How Russian Java is developed.
And although the titles of the reports give the impression that there is practically nothing in common between them, and about some you can even ask: “What does DevOps have to do with it?”, in fact, they are all united by a single topic: “What are the global trends and challenges in the field?” IT companies in Russia are now facing DevOps and what are we going to do about it?” This topic was asked at the beginning of the conference in a panel discussion.
Panel discussion: “DevOps in Russia: state and prospects”
Specialists from Axenix, Hilbert Team, Sber, and SberTech took part in the discussion. The moderator was Sergey Khrapov, Managing Director of SberWorks and leader of the DevOps Community (SberAcceleration).
Managing Director of SberWorks, leader of the SberAcceleration DevOps Community, Sber
First topic of discussion: trend towards import/vendor substitution
According to all participants, import/vendor substitution affected every IT company in Russia. But what to do about it, and will competitive products for development and DevOps appear in Russia in the next one or two years?
Most of the participants are optimistic about the future, because there is a good open-source database, adding security to it, you can get cool products, and besides, Russia has cool developers and an equally cool design school.
A period of one or two years seemed unlikely, but we had to start somewhere. And now there are good opportunities to start making cool products not only for the small market of Russia and the CIS, but also for the global one.
According to the results of the launched survey, the audience supported the optimistic mood of the discussion participants by a small margin.
The situation is quite complicated, in Russia the market for local products has just begun to appear, it has begun to fill up, and the problem is that there are no full-fledged products yet, but they will definitely appear in my opinion.
I believe in people, from my work experience, and I especially believe in our developers and our engineers – they will carry anything! The Russian developer is a brand!
Let’s look at the facts: there is a good Open Source base in the world, on which, adding security, adding a little reliability, scalability, etc., you can build one or another solution on this basis. The only question is how competitive it will be and where?
We need to focus on the world market, not just the Russian one. In the sense that we have really cool developers, managers, designers and products. But in Russia the market is not that big, that is, we have 140 million people there + CIS, however, this is not a billion people. And therefore, focusing only on global markets is a more correct goal setting.
It seems to me that conferences like this one can become a platform where we at least learn that we have competitive products.
The skepticism that is present in the voting results coincides with my initial one, which I stated. But, making bridges to the next reports, I would still like us to understand that stop waiting, we need to become competitive anywhere and it’s time to get it done!
The second topic of discussion: “AI Copilot – will AI really change the world, or is it just another hype?”
The theme of AI this year runs through all significant initiatives both in Russia and in the world. And the point is not so much that this is hype, but that it changes the balance of power and approaches to building DevOps in organizations, makes you think about how not to stay on the sidelines, preaching old approaches.
Perhaps this was the most lively part of the discussion, during which it was not possible to form polar opinions. All participants agreed that changes are inevitable, but how long it will take is a big question.
If we talk about the world as a whole, then, probably, something will change there, but if we talk about Russia and our industry, then these technologies cannot be particularly applied. Because standard solutions are more likely to meet the needs. We need to come up with some kind of high or low level architecture to suit the changing requirements. And here, it seems to me, all these AI will not help in any way.
However, AI can satisfy many Junior requests. For example, if we are talking about DevOps, then he will write out instructions for you on how to raise Kubernetes. But the question arises: who will check these instructions? And here we need Senior and Middle.
GitHub already sets its own benchmark, which says that overall development speed increases by 55% thanks to code generation, code completion, etc. At Sberbank we use GigaCode, and thanks to code completion, the speed of work of any developer increases. He does more features in the same time.
It seems to me that these tectonic changes, if you don’t jump on this train now, it may turn out that you will remain at the station.
With these systemic shifts, the transformation of DevOps not into a set of pipelines, but into some kind of platform-as-a-service, only senior guys can solve such a problem. At the same time, Juniors will be washed out, because we now have Co-Pilot, GigaCode, etc. The question is – when will we run out of Senior engineers?
Despite the fact that we have a lot of engineers with awesome education and great experience, for some reason we believe in “silver bullets”. “If AI comes, it will do everything!” But no! The question I would rather ask is, “Is AI 100% usable?” I think this will happen in 5-10 years. In a maximum of 10, when the AI acquires some subjectivity. But then he will take away simple tasks from Junior. And it turns out that the people who will be able to upgrade themselves – the Junior generation – will have big problems.
It turns out that you speed up the process of developing and testing your hypotheses. You will need to deal with more systematic and complex, non-standard tasks. And there will definitely be growth, but the question still remains open: “How can Junior grow?”
I will answer where to get Junior and what to do with them. There will be few live combat missions for them, but you will have a non-sleeping AI nearby who will present you with these tasks. It will need to be fine-tuned so you can grow. It seems to me that this is the same personal training that we have been told about for 20 years now.
Below is the result of an audience poll on the topic.
Third topic: “Data and monitoring”
After a heated discussion of the previous topic, it became clear that no AI will happen if there is no data on which to train the necessary models. And bad, low-quality data will only hinder the implementation of AI. At the same time, the amount of data is growing.
The more releases, the more data this whole enterprise generates: logs, metrics, whatever. But even though I have terabytes of data, what should I do with it? How can we make a sample from this that can be used to teach AI?
For predictive response, in addition to business, product and application metrics, we need infrastructure metrics. We must configure monitoring correctly and saturate our AI system with this data.
Previously, there was Oracle, Terradata, and now all these things need to be built by DevOps engineers, and developers need to write on top of it all. This is a challenge for the entire Russian industry, including the DevOps community and developers. Because data is constantly needed in fintech, retail, and industrial production. I heard that a Boeing plane generates 1 TB of telemetry in one flight.
When the term BigData appeared, there was such hype. But everyone forgot about the nuance that the data was called large because it was inaccurate. And when you feed this big inaccurate data to some model, it does a great job of it. There are no bad or good data, there are bad questions that you ask.
At the end of the discussion, the participants answered several questions asked by the audience. Here is the most interesting one, which immediately puzzled them:
Give a clear answer: who is DevOps?
I believe that DevOps is a super engineer who understands something in support and development, and most importantly: he feels responsibility at the intersection of Dev and Ops. Because often the developer wrote the code, threw it over the fence, it’s good if they caught him, shook him off and put him on his feet. Then the support team always asks: “Who wrote this, who does this?”, and sends it back in the form of a Change Request. As a result, DevOps comes, which is both ours and yours. So for me DevOps is a super engineer.
Full version of the panel discussion can be viewed via the link.
Creating a unified environment for DevOps processes of any complexity in the context of constant scaling and growing number of projects
Or how we dug into building DevOps processes so as not to dig into hiring DevOps engineers.
Product owner Release conveyor PV Works, SberTech
At the beginning of the report, Sergey is nostalgic for the old days, when there were independent developers who could independently create the entire product from start to finish. The times of John McCarthy and Alexei Pajitnov.
Today, for the most part, we are in a situation where the growth of business needs and the growth of infrastructure complexity, a huge number of different requests leads to the fact that the developer can no longer do it all as an indie. He is forced to focus on a narrow part of his production process.
And it turns out that we have a developer and this complexity, especially in Enterprise, which leads to the fact that over time the developer begins to pinch off certain parts of the process. When you were alone, you controlled everything yourself. Now you can’t do that simply because you don’t have the competence. We have to wait for the staging, the designs, the architecture… wait… wait… wait…
We have a dream: we want to go back to the past. I want it to be like before, when an indie developer pulled everything into one place. After all, it was cool! Then there was a feeling that you were a real creator of what you created. There was a complete association of me with the product that I make.
Now it turns out that I, as a developer, only write code in a small part of the production process, and some other people have to pick it up and do something further with it.
It is clear that we cannot go back to the past, because modern technologies, clouds, microservices and AI have appeared. We want to create a Productivity Platform – a productivity platform that will allow the developer to feel like an indie and return to that very same time.
Most of our work consists of routine. These same type of tasks do not allow us to develop. Productivity Platform OrchestraR should focus the attention of DevOps engineers on non-standard, complex and interesting tasks that will help improve their skills. We want it to be simple, for the tool to be clear, intuitive, to “walk us by the hand” during the setup process, to predict some of our actions.
What does the user want?
Something simple and clear
Set it and forget it
They don’t bother me
What do we offer:
Something simple and clear
Release Automation Tool: automation, orchestration of management of assembly and deployment of releases in all network segments
All requirements for the software production process are met out of the box, the team does not waste time implementing them
One hour to set up and launch your Release pipeline and DevOps CI/CD pipeline.
The developer will be able to create a pipeline on the platform in the OrchestraR tool using a wizard without involving DevOps.
Over time, the DevOps engineer may become a centralized function and will be engaged in the development of the master system, and not in the development of the next pipeline.
If you are interested in this topic, we recommend view the entire reportthere are many details and nuances about which you can write a separate article.
Clickhouse as a backend for Prometheus
At the panel discussion, we came to the conclusion that everyone needs data, including from monitoring systems, in order to use it in various AI and BI systems. How exactly is this data stored and transmitted?
The most common monitoring system today is Prometheus – it is an industry standard. With the increase in the number of infrastructure and metrics sources, we are faced with the need to scale Prometheus into several instances, and for metrics visualization systems, for example Grafana, in order not to glue metrics from several Prometheus instances, we need to connect Long Term Metrics Storage – long-term storage.
The report examines in detail several such products: Thanos, Grafana Mimir, Victoria Metrics and others.
Moreover, when storing metrics you need:
Minimize dependence on geopolitical risks
Be able to export to AI systems
Maintain high productivity
Be able to further scale horizontally
Be able to use Managed Data Storage
You can record your report view here.
GitVerse is a new web service for collaborative development and hosting of IT projects
CPO Platform V, SberTech
Technical expert, SberTech
The entire global IT market is moving towards open-source, and it is important that Russian developers have all the necessary tools to create advanced technological solutions.
GitVerse is a web service from Sber for collaborative code development and hosting that will allow developers to create open and closed source projects and attract new participants to the development of their solutions.
This is a completely domestic solution, independent of foreign vendors and technologies. With GitVerse, developers can build a project, test it, deploy it, and test it. The service will also help automate development, store code, create and delete repositories, assign tasks, and add comments.
And through GitVerse, it will be possible to connect an AI assistant for the developer – GigaCode, which speeds up writing and working with code.
Benefits of GitVerse:
The service was created and hosted in Russia – the risks of developments and code being unavailable to Russian users are eliminated
Free quotas up to 2 GB for using service resources
Plugins for all popular integrated development environments (IDEs), simplifying and speeding up development
Information portal where news and materials on open-source topics are published
In the future, GitVerse will be complemented by various tools for developing open-source projects, which early testers will have access to first.
For now, GitVerse can be used by Sber employees, but at the beginning of 2024 it will become available to everyone. Developers can already apply for participation in early testing — and be the first to gain access.
We will provide:
Git repository and code review functionality
Plugins for IDE GigaCode
Access to open-source projects
Participation in the development of the GitVerse project
And what’s under the hood, what it’s made of and why the solution should be reliable and safe was best told by Arseny Golushkov, plus there are answers to tricky questions at the end. Who hasn’t seen it yet? see the recording.
Kubernetes platforms: doing good and doing good
A report about Kubernetes platforms and why they are needed at all. It will be most interesting to those who already know what Kubernetes is, and to those who doubt that they need Kuber.
In the world and in Russia, Kubernetes is almost equally widespread. Its advantages:
Large market share
Created a new revenue-generating customer experience
Profitability has increased
Infrastructure management has become more efficient
Developers are more productive
Helps IT management demonstrate IT as a source of revenue rather than just a cost.
If any of these questions resonate with you:
Which OS should I choose for cluster nodes?
How to update the OS?
Which cluster architecture to choose?
Etcd on separate nodes?
Which CNI to choose?
Cilium or Calico?
Containerd or Docker?
Which ingress controller should I choose?
Nginx ingress or envoy?
Multicloud or hybrid cloud?
Do I need a Service Mesh?
Istio or something else?
… we highly recommend it view the report. There are many other useful things there.
How Russian Java is developed
A report for everyone who loves Java, about the Russian Axiom JDK, its advantages and features. Oleg talks in detail about how Java is still the main development platform, a powerful and comprehensive tool on the basis of which you can create solutions for any problem.
Below is a short excerpt from the report.
Java is now one of the most popular languages. In the “Most Favorite Programming Language” category, it is only ahead of Python. Why is Java so popular? There is a lot to list here, but it’s worth mentioning:
Long term support
Clear and predictable releases of new versions
Ability to run on different platforms and processor architectures, which is very important for Russian manufacturers
Ease of migration between platforms
Clouds have become very popular lately, but the downside is that they are not free. And modern basic images live in RAM and occupy a lot of it. What can a developer do about this? Nothing special. But it turns out that you can take one of the most popular Alpine images and teach it to work faster with Java.
Previously, everyone tried to rewrite everything on open-source. But there may be many vulnerabilities and specially created bookmarks. There are many examples. In Axiom JDK, development is carried out within the framework of the Secure Development Lifecycle (SDL).
Dependency Track – checks dependencies for open vulnerabilities;
SpotBug – static code analyzer;
Svace is a cool static analyzer from ISP RAS.
Axiom JDK is certified by FSTEC SZI UD4, this allows it to be used for the development of government information systems and critical infrastructure systems up to the first class inclusive. For example, the Faster Payment System (FPS) uses the Axiom JDK, and the number of transactions at the beginning of 2022 reached 5 million per day.
If the topic interests you, then everyone else details in the report recording.
Vendor substitution: We have really great developers, managers, designers and products, so everything is possible, we must “pull it in”, and we must start yesterday.
AI systems, assistants, chats, etc..: AI systems will definitely help with routine tasks, Senior specialists will be able to focus on more non-standard and complex tasks of developing DevOps platforms, which in the future may turn into new stages in the development of AI systems. Junior specialists will be able to develop while interacting with AI systems.
Monitoring and data: everyone needs a lot of them in order to subsequently develop AI systems for forecasting and predictive response to infrastructure and other problems and incidents.
Productivity Platform and DevOps-as-a-Platform: an inevitable evolutionary path in the context of all other aspects of technology development. Give back the opportunity to the developer to create their products independently from start to finish. Let DevOps specialists develop on interesting and non-standard solutions, rather than getting bogged down in the routine of servicing development teams.
GitVerse: a web service from Sber for joint development and hosting of code, which will allow developers to create open and closed source projects and attract new participants to the development of their solutions while minimizing the risk of vendors leaving the Russian market.
Kubernetes platforms and Platform Engineering: Flant already has vast experience and its own solutions to avoid “cycling”.
Russian Java: everything is fine with it, because there is Axiom JDK.
We thank all the speakers and colleagues who took part in preparing the DevOps track, it seems we did a great job and wish us all even more cool joint events and great conferences in the future.
Following the discussion…
The panel discussion issues obviously remain open for discussion. And it seems to us that some kind of separate platform is needed to discuss such problems and issues. The site is far beyond the boundaries of Sberbank or another organization. More like some cross-organizational DevOps community. What do you think about this and what value can we get? Share your opinion, take a short survey.