Open broadcast of the main hall HighLoad ++

Great news - like last year we will be broadcasting the main hall of HighLoad ++ with the most interesting reports for free!

Broadcast HighLoad ++

Do not forget to click on " Remind " and also subscribe to our channel . We constantly post videos of our reports in it.
')
Only the main hall will be broadcast in open access; if you want to watch the other 9 halls , you can also purchase private access .

The full schedule is available on the website ( PDF , HTML ), and we will tell you about the most interesting reports in the roll-up.

Life after sharding

November 8, 14:00

Many people think that in order to create a good highload project, it is necessary to load data in a proper way and sleep well, delivering the server as needed. In fact, it is only after sharding that the most interesting begins, since after physical separation the data does not cease to be logically related.

Mikhail Kurmaev from Badoo will talk about how they solved problems of this class with the MySQL database. We will discuss both the typical problems outside of sharding, and the process of its construction itself: the processing of real-time events, queues in geographically separated data centers and much more.

As I was a team leader, and now - the head of the direction

November 8, 11:00

In the report of Vitaly Sharovatov from Badoo, he will talk about two topics: how the growth from the team leader looks to the head of the department, and how to choose the team leader to replace him, to help him become effective.

We will talk a little about psychology, about the atmosphere in the team, mentoring, of course, about hiring and much more. From Vitaly's speech you will learn how:

• correctly choose a team leader : based on what criteria, how to check how a person can approach, how to sell the idea of ​​a career change and deal with classical fears;
• to help the chosen person to confidently enter into the course of business and become as efficient as possible, to fight;
• monitor the situation in the teams and keep your hand “on the pulse”;
• stop leading the team yourself , and how lead management differs from developer management.

2017 headliner Mikhail Sychev from Google with a report

YouTube.com Reconstruction - rewriting the second most popular website in the world

November 7, 13:00

I 'll tell you why we decided (and only now) to switch to a client-side-application, how we evaluate the performance of YouTube (and why "Time To Glass" and "Above the Fold" is only the beginning of the story), as front-end and back -end work together to achieve maximum performance when business logic is designed for dozens of clients.

We will discuss why we chose Polymer as a framework for the site (no, not just because Google made it), a new development platform based on it, A / B testing and “experiments” to evaluate changes and control user reactions, and which metrics we pay attention to ...

Quality control of highly loaded systems

November 8, 12:00

Agree that building high-load systems blindly is unrealistic? How to build quality control in the development of complex services? What exactly to look for when testing performance? The topic of testing loaded systems has been covered more than once at many conferences; there are a lot of articles, blog entries, instructions from vendors. A huge pile of information.

Without training in this abundance of tips and solutions, you can easily get confused and waste precious time on testing and wasting quality control. And you can listen to Andrei Drozdov with a story about how the quality control of loaded systems in Avito is arranged.

In his story, Andrew will tell you how to build a system for testing and controlling the quality of services from scratch. Together with you, we will hear about fault tolerance, automation, see what tools Avito uses and what they do. And we will also see how correctly selected metrics and accurate accounting.

Effective use of ClickHouse

November 8, 15:00

ClickHouse just over a year is available in open-source. During this time, gained experience in its use in Russian and foreign companies. In some of them, data volumes and loads exceed those with which we deal in Yandex. Since ClickHouse is a specialized system, it is important to take into account the peculiarities of its architecture.

In his report, Alexey Milovidov will talk about examples of typical mistakes when using ClickHouse, which can lead to inefficient work. The examples from practice will show how the choice of a particular data processing scheme can change the performance by orders of magnitude.

How to rewrite the VK private messages database from scratch and migrate to it without downtime

November 7, 6 pm

For many years, private messages in the social network VKontakte worked on the same samopritny database. Over time, it is morally obsolete, making it difficult to implement a new functional, and the growing loads have led to the fact that a huge number of servers are required for the functioning of the messenger.

Dmitry Egorov will talk about how, within 10 months, a new message storage system was developed from scratch and removed many architectural constraints, as well as having much higher productivity, which allowed the company to save millions of dollars! The report will discuss the architecture of both personal message engines, many optimizations, and the technically most difficult part — migration without downtime — is described in detail.

Interesting?

Everything, we go to Skolkovo for construction, and we are waiting for you on the online broadcast :)

See you!

PS: By the way, NGENIX delivers and accelerates the broadcast, for which she thanks a lot! And the technical director of the company, Dmitry Krikov, together with Alexander Petrovsky from Mellanox, will read on HighLoad ++ a report entitled " Disaggregation and openness on Ethernet networks ".