Saint HighLoad++ 2022
2022-09-22..2022-09-23
-- online --
[Saint HighLoad++ 2022, Saint HighLoad++ 2023, Saint HighLoad++ 2025]
| title | author | length |
|---|---|---|
| Video Report on | 03m | |
Video report from the Saint HighLoad++ 2026 conference in Saint Petersburg, which is the largest professional event for developers of high-load systems. The… | ||
| Architecture: History and Future Illustrated by VKontakte | Aleksandr Tobol | 01h 05m |
Covers the evolution of architecture for high-load systems, tracing principles, implementation patterns and tooling choices. Uses VKontakte as a concrete case… | ||
| Reliability of High‑Load C++ Applications in Yandex.Market | Kirill Gorelov | 42m |
The talk describes how Yandex.Market ensures reliability of high-load C++ applications that serve millions of daily shoppers while development pace and feature… | ||
| Wi‑Fi/GSM Geolocation in Large Cities Using ML for 30 Million Users | Artem Kravchuk | 31m |
Uses machine learning to provide Wi‑Fi/GSM geolocation for 30 million users in large cities by building and using millions of 2D signal-strength maps of Wi‑Fi… | ||
| Deduplication of 5 Million Events per Second on YDB in AppMetrica | Artem Ismagilov | 37m |
Describes AppMetrica’s event-processing pipeline built to deduplicate 5 million events per second using YDB. Covers the problem of massive-event deduplication… | ||
| Face Pay | Mikhail Volkov, Guseyn Rimikhanov | 32m |
Face Pay runs a contactless fare system using VisionLabs face recognition; users upload a photo, a server extracts a 512-byte descriptor from ~40 facial… | ||
| Programming Languages: Past, Present and Future | Dmitriy Zavalishin | 52m |
Overview of programming languages' past, present, and future with a focus on how to evaluate a language’s quality and value beyond immediate programmer-centric… | ||
| How We Protect Millions of Customers During Overloads | Anton Kolesov | 49m |
Describes protecting millions of users from cascading failures caused by overloads in high-load systems by isolating and limiting cross-component impacts… | ||
| Speech recognition for subtitles in VK Video | Vitaliy Shutov | 42m |
Describes VKontakte’s speech recognition system used to produce subtitles for videos, explaining differences between recognizing short voice messages and long… | ||
| From 0 to 200,000,000 players — on backend evolution in 40 minutes | Andrey Mikheev | 36m |
War Robots is a multiplayer 3D shooter whose backend architecture evolved over eight years from a simple single-cloud service handling player profiles to a… | ||
| Distributed high-load BI engine for Google Data Studio and Microsoft Power BI | A.Serbul | 47m |
A distributed, high-throughput cloud BI engine built to serve Google Data Studio and Microsoft Power BI is described, covering design, implementation and… | ||
| Postgres from the dawn of time to the present day | Oleg Bartunov, Ivan Panchenko | 50m |
A historical overview tracing database systems from early precursors to modern PostgreSQL, explaining how the concept of DBMSs emerged and how PostgreSQL… | ||
| Can data drive data center hardware configuration? | Pavel Lavrenko, Anton Katenev | 43m |
Discusses how to use data to drive selection of hardware configurations and storage/management strategies for high-performance and high-load data centers… | ||
| Applying machine learning to analysis of scientific data | Aleksandr Proshunin | 26m |
Describes applying machine learning and AI methods to analyze scientific data, including image recognition for processing large image datasets and signal… | ||
| Component graph: how to display 100 components and 500 connections | Anastasiya Nekrashevich | 41m |
Describes visualizing a system of about 100 components and 500 connections in a microservice architecture: the problem is how to represent many services and… | ||
| Preparing Yandex search data — the library and processes we built | V.Agapitov, K.Khamitov | 21m |
Yandex Search data-prep team describes building a processing pipeline that handles 5 Gb/s of streaming traffic. The talk explains why they chose a hybrid… | ||
| On the history and future of search | Andrey Aksenov | 49m |
A 30-minute overview of the 60-year history and future of search engines, surveying the major theories and techniques that have been tried. Covers which search… | ||
| Containers are dead. Long live virtual machines! | Viktor Popov | 45m |
Explains what container runtimes are and the different types available, then introduces MicroVMs (micro virtual machines), how they relate to containers, and… | ||
| How to deploy a recommendation service with a BERT-like model to high-load production | Mark Panenko | 47m |
Describes deploying a high-load production recommendation service built with a BERT-like model at Rabota.ru to improve job search user experience. Covers… | ||
| Hybrid architecture: a monolith split into microservices | Stanislav Sidristyy | 44m |
A talk about designing a hybrid architecture that starts from a monolith and incrementally splits into microservices to meet 24/7 high-load requirements. It… | ||
| Centralized self-service ETL | Andrey Goncharov | 51m |
Describes building a centralized self-service ETL platform to let product and business analysts automate batch transfers and transformations between systems… | ||
| Superspeed. Yandex unified express delivery platform | Sergey Khoroshenkikh | 51m |
Description of Yandex’s unified express-delivery platform that handles orders requiring immediate delivery (examples: Yandex.Lavka and Yandex.Eda). Covers the… | ||
| Pipeline evolution and lifecycles | Aleksandr Oblivalnyy | 41m |
Discusses challenges of introducing and evolving CI/CD pipelines in teams, noting that resistance often comes from focusing on tool selection rather than… | ||
| Authors on Zen and how we find their audience | Anastasiya Pavlovskaya | 33m |
Describes how Yandex.Zen helps content creators find their audience using recommendation systems. Covers the recommendation technologies and mechanisms used to… | ||
| No A/B — result XZ, or how we built an A/B testing platform at Ozon | Evgeniy Pak | 48m |
Describes how Ozon built an internal A/B testing platform that handles business needs (hypothesis validation), statistical concerns (significance), and… | ||
| Dashboard as code, or the path from UI edits to grafonnet | Anton Bystrov | 25m |
Shows how to treat dashboards as code by moving from manual UI edits to programmatic definitions using tools like grafonnet. Covers techniques and tooling to… | ||
| One platform to rule them all | Mikhail Kabishchev | 52m |
Ozon replaced several large monoliths with thousands of polyglot microservices and faced duplicated engineering effort solving the same problems. The talk… | ||
| Blockchain in enterprise architecture — fashion or necessity? | Denis Vasin | 46m |
Examines whether blockchain belongs in corporate IT architecture by weighing its pros and cons and asking if it is hype or a necessity. Discusses blockchain… | ||
| Building a cloud AutoML platform for creating voice robots | Artem Bondar | 47m |
Describes building a cloud AutoML platform for creating voice robots using NLU technologies. Covers using out-of-the-box systems (DialogFlow), open-source… | ||
| All about smart homes and apartments through the eyes of an IT architect | Andrey Loginov | 47m |
Overview of smart home systems from small buildings to skyscrapers, covering the history of technologies and paradigms used to build them. Describes what a… | ||
| Expert systems for metallurgy: a five-year plan in one year | Sergey Okatov | 53m |
Describes building expert systems for the steel industry, focusing on how metallurgical projects differ from typical business automation because budgets and… | ||
| How we build our own PaaS at 1C | Basel Darvish | 34m |
Describes how the 1C company built an internal PaaS to host thousands of business applications and tens of thousands of users on a single infrastructure… | ||
| High-load blockchain in a mission-critical system | Ivan Kosarev | 35m |
An engineering team describes building a high-load federal electronic voting system using blockchain. The talk covers system requirements for a… | ||
| 10ms Response with Transactions, Big Data, Flexible Logic and OpenSource | Vladimir Bogdanovskiy | 49m |
Describes connecting two open-source storage systems via queues inside a service platform to build a strictly-consistent transactional system that supports… | ||
| Channel. Product. Platform or The Evolution of Approaches to Developing Tinkoff Mobile Bank | A.Polomodov | 58m |
This talk describes the evolution of Tinkoff Mobile Bank: changing business requirements, team management and application architecture as the project scaled… | ||
| Analytics for S7 Aircraft: Pets vs Cattle | Lyubov Rozhkova | 44m |
An S7 Airlines engineering talk about aviation analytics contrasts treating aircraft as "pets" versus "cattle": it describes building operational analytics for… | ||
| Global Move to Unicode: Preparing Mail Systems for Russian Addressing | V.Mikhaylov | 44m |
Covers the global shift to Unicode for email addressing, explains the history of internationalizing internet identifiers and current work by standards bodies… | ||
| To Peer or Not to Peer or Transaction Ordering in Public Blockchains | Sergey Prilutskiy | 41m |
Examines ordering of transactions in public blockchains, focusing on ranking, chain throughput, and transaction delivery over P2P networks. Describes new… | ||
| Microservices Through Pain and Perseverance | Filipp Delgyado | 49m |
This talk examines the real costs and challenges of adopting microservices, comparing expected benefits to practical problems encountered in projects. It… | ||
| Service Mesh on Steroids, Part 2: Zero Deployment Downtime in Enterprise Applications | A. Efimov | 50m |
Describes using Service Mesh in enterprise applications to achieve zero deployment downtime by applying multi-level, managed Service Mesh patterns for… | ||
| How We Brought Bioreactors and ML Together | Vladimir Merkulov, Vasiliy Vologdin | 38m |
Describes integrating bioreactors with machine learning at pharmaceutical company BIOCAD: using ML to monitor and control bioreactor processes in drug… | ||
| How to Forget About Performance Problems? Tarantool as the DBMS | Aleksandr Kharchenko | 38m |
Describes using Tarantool as the database to handle Citilink’s product catalog where caching is not possible due to dynamic pricing. Covers deploying and… | ||
| Which Architectural Decisions in Yandex Go Enable Running Dozens of Product Experiments | O.Ermakov | 42m |
Yandex Go’s product team describes moving from a small monolith to a 50+‑person organization running a microservices architecture to support dozens of parallel… | ||
| Seamless Integration of Security Practices into the DevOps Pipeline | Andrey Ivanov | 50m |
Examines how to integrate security practices seamlessly into a DevOps pipeline by explaining when an organization should adopt DevSecOps and how to prepare for… | ||
| A Pipeline to Transcribe Speech in Millions of Videos per Day | Filipp Malkovskiy | 45m |
Description of a pipeline used at VK to transcribe speech from millions of videos per day, covering how the team scaled speech recognition from a voice-message… | ||
| Multi-tenant Kubernetes | Mikhail Sidorov | 51m |
Three practical multitenancy patterns are compared—separate clusters, namespace-based (Capsule), and virtual control planes—along with concrete primitives… | ||
| How to Save on Scaling by Moving from Cassandra to Scylla DB | Mikhail Malyshkin | 35m |
Describes migrating a transactional ML inference service from a 12-node Cassandra cluster (10 cores, 32 GB RAM per node) to ScyllaDB to reduce scaling costs… | ||
| How to Survive Under Load with 100TB in an Unsharded MongoDB | Andrey Komyagin | 41m |
Describes running a non-sharded MongoDB database holding about 100 TB and keeping it reliable under load. Covers practical lessons and operational techniques… | ||
| Digitizing Business with Jira: From Sales to Accounting | Darya Petrova | 50m |
Shows how to digitalize business workflows using Jira by building custom processes for every team — from sales to accounting — and explains the configuration… | ||
| Operating Voice in Marusya: Problems We Faced and Why GPUs Matter | Pavel Alekseev | 42m |
Discussion of running large-scale voice processing in Marusya using GPUs, focusing on hardware and driver reliability problems (cards failing, disappearing, or… | ||
| Frequency and Bayesian Approaches to Estimating TPR with Partial Labels | Alan Savushkin | 31m |
Compares frequency-based and Bayesian methods for estimating true positive rate (TPR/Recall) when labeled data is incomplete; discusses practical evaluation of… | ||
| Optimizing Data Storage Costs in Object Storage | Vitaliy Isaev | 49m |
Describes Dispersed Object Store (DOS), an object storage developed inside the Mailion corporate mail system, and how deduplication of whole emails and their… | ||
| OK S3: Building the System Ourselves | Vadym Tsesko | 54m |
Overview of Odnoklassniki’s in-house distributed, fault-tolerant object storage systems (one-blob-storage and one-cold-storage) used to store and process… | ||
| How Not to Fail at Import Substitution | Evgeniy Lukin | 48m |
Examines how large enterprises replacing foreign vendors (Oracle, SAP, IBM) attempt to implement import substitution for high-throughput data buses. Describes… | ||
| The full chain load test in AliExpress Russia | Hu Junyuan | 22m |
AliExpress Russia runs a full-chain load test that models peak QPS using a formula with recent campaign peaks and a campaign adjustment coefficient, splits… | ||
| The full chain load test in AliExpress Russia | Hu Junyuan | 22m |
AliExpress Russia runs a full-chain load test that models QPS from daily peaks plus a campaign adjustment coefficient and separates read/write chains by DAU… | ||
| Vector Search in ClickHouse | Artur Filatenkov | 37m |
Describes using ClickHouse for vector search to accelerate multidimensional queries such as those arising from text and image embeddings. Covers performance… | ||
| The General Development Flow at Ozon: How to Make Developers' Lives Easier | Daniil Razumov | 47m |
Describes how Ozon redesigned its development workflow to make engineers’ lives easier, explaining the problems developers faced previously and the motivations… | ||
| The Necessity and Pain of Moving from IBM MQ + RH Fuse to Apache Kafka + Apache Camel | Aleksandr Krylov | 45m |
Compares migrating from IBM MQ with Red Hat Fuse to an Apache-based stack using Apache Kafka and Apache Camel, examining trade-offs between enterprise and… | ||
| How We Hosted 200+ Data Scientists in a K8S Cluster | Aleksey Kuznetsov, Mikhail Lepeshkin | 39m |
Describes deploying a scalable JupyterHub environment on Kubernetes to host over 200 data scientists, focusing on easy scaling, moderate fault tolerance, and… | ||
| Our Own Distributed S3 on MinIO — Practical Experience Stepping on Rakes | Aleksey Pletnev | 39m |
A practical report about building a self-hosted S3-compatible storage using MinIO, describing why projects need reliable, fault-tolerant, and performant object… | ||
| Optimizing ClickHouse query performance | Maksim Kita | 42m |
Practical techniques for improving ClickHouse query performance, including how a CI/CD infrastructure is used to test for regressions, detect slow queries, and… | ||
| OpenSource as an attack vector: how dangerous is it and how to fix it? | Aleksey Morozov | 52m |
Examines vulnerabilities originating from open-source components and libraries that become attack vectors, using real-world examples of notable attacks through… | ||
| ERP project for tens of thousands of users: scared but doing it — how? | Aleksey Kislov | 51m |
An ERP implementation for tens of thousands of users is described, covering project start, team selection, and long-term support. The talk details problems… | ||
| RePlay — a library for building recommender systems | Aleksey Vasilev | 39m |
RePlay is an open-source framework for building recommender systems that emphasizes a convenient API and scalability. It shows how to quickly construct… | ||
| Technical governance of an IDP (Internal Developer Platform) for 7,000 developers | Dmitriy Gaevskiy | 48m |
A company faced rapid engineering growth and decided to build an internal developer platform (IDP) from scratch to support 7,000 developers. The talk explains… | ||
| How to know a project is in trouble if you are an engineer | Yuliya Belozerova | 48m |
A talk about how an engineer can detect early warning signs that a project is in trouble, illustrated with a cloud migration project. It details concrete… | ||
| The history of online video | Maksim Lapshin | 52m |
A historical overview of online video usage on the internet, showing that entertainment streaming dominates user consumption (up to ~4 hours/day) while video… | ||
| Introducing OpenSource solutions into industrial control systems (ICS) | Vladislav Maslov | 47m |
Describes using open-source software in industrial control systems (АСУТП) for manufacturing sectors like oil & gas, metallurgy and pharmaceuticals. Covers… | ||
| Automatic parameter tuning for Spark applications | Valeriya Dymbitskaya | 44m |
Describes an automated system that chooses optimal Hadoop/Spark resource configurations for hundreds of daily and thousands of hourly Spark jobs with different… | ||
| Second space age: how to overcome the gravity of an internal development platform | Anastasiya Abrashitova | 47m |
An internal developer platform accelerates project launches, integration and scaling by providing shared infrastructure and expert support. The talk examines… | ||
| Under the red flag: how an engineer can tell something is wrong in a project | D.Podolskiy | 40m |
Describes how an engineer can detect signs that a software project is going wrong due to poor technical decisions by management or the team. Covers practical… | ||
| The evolution of distributed attacks on the Internet: 1994 to the present | Georgiy Tarasov | 50m |
A historical survey of distributed attacks on the Internet from 1994 to the present, tracing major technical and public milestones such as Panix, Sony… | ||
| AI maturity index — why and how AI maturity is measured | Andrey Anisimov | 50m |
An overview of the AI maturity index as an international framework for systematizing how Data Science and AI are adopted in business processes. It explains why… | ||
| Load testing speech synthesis and recognition at SberDevices | Andrey Chernopyatov | 37m |
Describes how SberDevices performs load testing for its speech synthesis and recognition pipelines to handle billions of seconds of audio with guaranteed… | ||
| The ML wedding between millions of products, or how to withstand load in a stream | Van Khachatryan | 52m |
A real-time matching algorithm for millions of e-commerce products is described, including how it was built to handle high throughput and scale on a very large… | ||
| How we migrated our database to the cloud | Oleg Blokhin | 34m |
A postmortem of migrating a production MySQL database to the cloud after an unexpected minor MySQL upgrade caused failures during a company offsite. It… | ||
| I still don't know how to network | Evgeniy Idzikovskiy | 45m |
Practical tips and exercises for developers on becoming better at professional networking: how to overcome the fear of starting conversations and how to pitch… | ||
| Networking workshop | Evgeniy Idzikovskiy | 49m |
A networking workshop presented at the Saint HighLoad++ conference focused on practical network engineering for high-load systems. It covers hands-on… | ||
| TechTalk #1: Big Data challenges in the turbulent year 2022 | Sergey Mikhalev | 12m |
Discusses Big Data challenges in 2022 faced by large-data companies, focusing on hardware limitations and shortage of skilled engineers. Covers practical… | ||
| TechTalk #2: From manual tester to automation engineer | Dmitriy Gorelov | 09m |
A talk about transitioning from manual testers to automation engineers, discussing whether to hire experienced automation professionals or train existing… | ||
| TechTalk #3: Building a banking system with DBI properties | Evgeniy Sultanov | 06m |
CFT aimed to make its 20+ year banking system DBI (database-independent) so it can run unchanged on any RDBMS; they focused on adapting the system to Oracle… | ||
| TechTalk #4: How search works on VKontakte | Bogdan Garkushin | 12m |
Overview of how VKontakte implements search by combining multiple specialized search systems rather than a single solution. Describes which off-the-shelf… | ||
| TechTalk #5: Users came here to 'finish off': a simple product is no reason to stop growing | V. Kolontaev | 10m |
Practical discussion of organizational methods to prevent team burnout when working on a simple, low-change product. Covers management practices and approaches… | ||
| Conference opening | 15m | |
An opening announcement for the Saint HighLoad++ conference for high-load systems developers, presenting the event details: Saint Petersburg, June 2026 at… | ||
| Conference closing | 08m | |
A closing note for the Saint HighLoad++ 2022 conference in Saint Petersburg, summarizing the event details: it was a major professional conference for… | ||
| nothing matches | ||