HighLoad++ 2024
2024-12-02..2024-12-03
-- online --
[HighLoad++ 2012, HighLoad++ 2013, HighLoad++ 2014, HighLoad++ 2015, HighLoad++ 2016, HighLoad++ 2017, HighLoad++ 2019, HighLoad++ 2022, HighLoad++ 2023, HighLoad++ 2024, HighLoad++ 2025]
| title | author | length |
|---|---|---|
| Report from HighLoad++ 2024 | 02m | |
A report listing major HighLoad++ 2024 events for high-load system developers, with dates and locations for related conferences: Saint Petersburg (June 2026… | ||
| Architecture of VKontakte's storage | Denis Deryugin | 44m |
Description of VKontakte’s storage architecture for trillions of photos, audio and other files across thousands of servers. Covers the evolution of their… | ||
| Real-time A/B: how to get metrics for experiments | Danil Valgushev | 40m |
Describes building a real-time A/B metrics counting system for high load and large data volumes, explaining why real-time experiment metrics are needed, the… | ||
| Program like a hacker, defend like a programmer | Vladimir Kochetkov | 50m |
Describes modeling an application the way an attacker does: maintain a changing mental model of critical execution points, ways to influence them, and expected… | ||
| How to reconcile networking in Kubernetes with legacy systems | Dmitriy Samokhvalov | 49m |
Talk demonstrates using eBPF to integrate legacy network environments with Kubernetes, showing real-world cases and tools that let developers and SREs work… | ||
| How to drive a million kilometers in YTsaurus | Vladislav Dolbilov | 34m |
YTsaurus is presented as a distributed data-processing platform; the talk explains its map-reduce operation and how it runs a large-scale autonomous-vehicle… | ||
| Iterators in Go 1.23: how to use them and how fast are they? | Pavel Agaletskiy | 53m |
The talk explains the new iterator feature added in Go 1.23 (aka ‘range over funcs’), why it was introduced and the language motivation. It presents… | ||
| Code generation tips you didn't ask for | Dmitriy Novikov | 51m |
Practical advice and lessons about code generation for high-load systems, drawn from the presenter’s long-term experience from student projects to current… | ||
| T-Bank video call architecture | Evgeniy Firstov | 49m |
Describes T-Bank’s video-calling architecture used for customer service and secure authentication in sensitive operations. Covers technologies and design… | ||
| Experience running a Service Mesh at Avito | Igor Balyuk | 49m |
Avito describes their operational experience running a Service Mesh to make microservice communication more flexible, stable, and reliable. The talk covers… | ||
| Designing our own in-house Feature Platform | Alina Baymasheva | 43m |
A talk about designing an in-house Feature Platform for machine-learning systems, describing its component architecture, key design considerations, and market… | ||
| Nvidia Triton Inference Server: building production ML without developers | Anton Alekseev | 51m |
A practical case study of building a production ML inference platform using NVIDIA Triton Inference Server and Kubernetes. Covers problems encountered with… | ||
| Continuous Profiling — what, why and how | Aleksandr Lutsenko | 47m |
Continuous profiling for production systems: why traditional profiling (done locally or in limited environments) misses real application behavior, what… | ||
| YTsaurus Dintables — yet another DBMS from Yandex | Ruslan Savchenko | 47m |
Overview of YTsaurus dynamic tables (dyntables): how Yandex uses them as a new DBMS feature for dynamic, high-throughput workloads. Covers YTsaurus… | ||
| Can we go with a database but without a cache layer in 2024? | Aleksey Rybak | 44m |
Benchmarks compare running high-load services with only a database versus using cache layers, using hardware with a 48-core Xeon and databases PostgreSQL and… | ||
| Smart apartment buildings — how realistically secure are they? | Nikolay Fomin | 42m |
The talk examines legal and technical security requirements being introduced for smart multi-apartment buildings in Russia, contrasting single-family… | ||
| Scaling DDoS Protection to Tens of Millions of RPS: Yandex’s Experience | Ivan Tyamgin | 42m |
Yandex describes its Antirobot system for protecting services from DDoS attacks and web scraping, a service that handles tens of millions of requests per… | ||
| The Eternal Pain of Accurate Addresses | Mikhail Kolosov | 40m |
Addresses (legal, factual, postal) are essential for systems at any scale; the talk analyzes how to prepare and handle addresses correctly in software. It… | ||
| Workshop: Containers and Networking — Learn, Inspect, Debug | Aleksandr Popov, Georgiy Melikov | 01h 54m |
Workshop covers low-level Linux networking for container and cloud environments, focusing on iproute2 utilities, Linux network namespaces, virtual network… | ||
| Epic Programming: How We Write Clear, Maintainable Sagas | Aleksey Bakin | 38m |
The talk explains using the Saga pattern to implement coordinated, consistent changes across multiple microservices. It covers practical approaches for writing… | ||
| Secrets of High Performance in Multicore Systems ☠️ | Nikita Galushko | 52m |
Explains when to shard work on atomic values and concurrent maps to improve throughput on multicore systems, and when that is beneficial. Describes algorithms… | ||
| Fast Cashbacks for Fast Payments | Konstantin Guzarov | 40m |
An overview of a loyalty program implementation at NSPK describing the initial MVP, methods for calculating cashback (including near-real-time within 30… | ||
| FinOps: Optimize for K8s | Ilya Kochnev | 46m |
Explains FinOps techniques for Kubernetes to reduce cloud spend when deployments fail due to perceived resource shortages despite low observed utilization… | ||
| From Legacy to Building Our Own Feature Store: Active Steps | Evgeniy Dashchenko | 40m |
Describes migrating from legacy feature handling to building an in-house Feature Store for an organization with many ML models. Covers integrating the new… | ||
| Building an LLM Infrastructure from Scratch: Lessons from X5 Tech | Michil Egorov | 46m |
Describes how X5 Tech built an internal LLM infrastructure from scratch, covering what large language models are, how they can change business automation, and… | ||
| Under the Hood of WordPress — Exploring the Engine and Tools | Olga Glekler | 46m |
Explains how WordPress works internally, covering its architecture and proven design decisions that power about 43% of the web. Reviews new WordPress APIs and… | ||
| Incremental Backups in PostgreSQL | Aleksey Darvin, Sofya Kopikova | 39m |
Presentation covers Ptrack, a tool for incremental backups for PostgreSQL that stores changed pages using Bloom filters, and compares that approach to… | ||
| Reliability at 45M Customers — Tools and Practices of a Digital Bank | Aleksey Merson | 44m |
A presentation about reliability engineering at a digital bank serving 45 million clients, describing tools and practices used to operate tens of products… | ||
| Mailing Adventures with PostgreSQL: Taming 650+ Shards and Surviving | Kirill Grigorev | 39m |
A presentation about a PostgreSQL-based sharding service used for mail at Yandex: it explains why they built the sharding service, how the central metadata… | ||
| Inventory That: Building an Automated DevSecOps for 40K Repositories | Aleksandr Melnikov | 33m |
Describes building an automated DevSecOps inventory pipeline for 40,000 code repositories: what metadata to collect and store in a database, how the inventory… | ||
| Speeding Up LLM Training by Over 45% | Mikhail Khrushchev | 46m |
Describes bottlenecks in training large language models on big clusters and presents practical solutions: the YaFSDP library to reduce communication… | ||
| Declarative Access Management Platform: From Roles to Dynamic Policies | Oleg Kozyrev | 42m |
Describes a declarative access-control platform that moves beyond static role-based access to dynamic policies. Covers why authorization is needed, the… | ||
| Fear and Loathing in Vi.Tech: How to Live Without Microservices | Kirill Kuzin | 50m |
Describes replacing a complex ecosystem of three PHP monoliths, numerous background jobs, and mixed Go/PHP services—deployed with Docker, Kubernetes, Terraform… | ||
| How We Increased Image Search Throughput | Denis Sakhnov, Dmitriy Buzmakov | 47m |
Describes an image search system architecture and current optimization techniques, comparing their pros and cons. Explains a tiered (multi-tier) indexing… | ||
| What I Got Wrong Building a Platform at a Market-Leading e‑Grocery | Oleg Fedotkin | 44m |
A first-hand account of building an internal platform for an e-grocery leader (Cooper / ex-SberMarket), describing mistakes made while designing and… | ||
| From Mistakes to Success: Evolution of the ML Feature Store at Flocktory | Oleg Kabanov | 44m |
Describes Flocktory’s journey building an in-house ML Feature Store, moving from standard backend storage to a custom system optimized for Data Science and… | ||
| From NSX to OVN: Four Years of Prep and a Live Cloud Migration | Vladislav Odintsov | 47m |
Describes migrating a cloud virtual network from VMware NSX to OVN over four years while running production. Covers reasons for migration (end of NSX… | ||
| Using and (slightly) abusing APCu cache in high load concurrent environments | Ioannis Loukeris | 45m |
APCu is used as a local PHP in-memory cache to cut DB reads and achieves roughly ~500 requests/sec fewer queries but forces careful tuning: split huge arrays… | ||
| Go in the WebAssembly World: Beyond the Browser | Maksim Kokryashkin | 36m |
Overview of using WebAssembly (WASM) outside the browser with Go: explains how Go can compile to or interoperate with WASM runtimes to run high-performance… | ||
| Transactional Work with Topics | Aleksey Nikolaevskiy | 38m |
Explains transactional operations on topics: what a topic is, transactional writes and transactional reads. Compares a model sharding task requiring ordering… | ||
| Is It Easy to Give Out Over a Million Discounts in Two Days? | Anna Kustareva | 46m |
A presentation on building a promotions/discounts distribution system that scaled to over a million coupons in two days. It describes the system architecture… | ||
| Security Gate: Building a Homegrown SAST Orchestration Platform | Igor Ignatev | 47m |
Describes building an in-house orchestration platform for SAST and SCA tools after DefectDojo failed under load; the platform runs about 1,500 scans daily… | ||
| Iron FinOPS | Aleksey Uchakin | 45m |
The talk examines cost management for on-premises infrastructure by adapting FinOps practices used in cloud environments. It discusses applying FinOps tools… | ||
| Modern Approaches to Product Matching Using LLMs | Vitaliy Kuliev | 47m |
The talk explains how large language models (LLMs) are combined with traditional machine-learning algorithms to solve product matching (determining whether two… | ||
| Pushing Go TLS to 100 Gbps from a Server ☠️ | Kirill Shvakov | 46m |
A case study of building a CDN in Go focusing on achieving 100 Gbps TLS performance from a single server. It explains why Go was chosen, describes the system… | ||
| In and Out in 20 Minutes: How We Rewrote a Bank's API Gateway | Maksim Denushev | 31m |
A software engineer explains how they rewrote a bank's API gateway, covering how to recognize when the existing system is insufficient, how to choose and… | ||
| Architecture of the National Video Hosting RUTUBE: How We Serve 5 Tbps | Eldar Niyazov | 47m |
Presents the modern architecture of the Russian national video hosting RUTUBE, describing its storage design (explicitly not S3) and the reasons for that… | ||
| Disaster Non-Recovery | Artem Fedorov | 43m |
Describes a real-world disaster-recovery exercise involving data-center shutdowns and a concrete failure where Patroni-managed PostgreSQL clusters recovered… | ||
| Centralized MLOps Platform: From R&D to Inference in a Single Loop | Elizaveta Gavrilova | 47m |
A case study of building an internal MLOps platform used by 20+ teams, covering technology choices driven by business needs and ML developer requirements. It… | ||
| ML Deployment Methodology Illustrated by Route Time Estimation | Dmitriy Klebanov | 50m |
The talk covers applying machine learning to predict express delivery travel times, describing the problem of delivery time estimation and how ML models are… | ||
| Talking to Smart Devices with PHP | Aleksandr Cherednikov | 40m |
Shows how to build IoT device communication using PHP and MQTT: explains MQTT advantages (lightweight pub/sub), how to implement publish/subscribe flows with… | ||
| Automating Incident Management for 10,000 Engineers at T-Bank | Ivan Yurchenko | 49m |
Automation of incident management at T-Bank for a 10,000-engineer organization, covering detection and remediation of incidents, management of monitoring… | ||
| Apache Kafka as the Single Source of Truth: Usage, Problems, Pain and Consequences | Dm. Emelyanov | 45m |
Examines using Apache Kafka not only as a messaging system but as the sole data storage for applications, describing practical usage patterns, problems… | ||
| Code Beauty Through the Eyes of a Neurophysiologist-Programmer | Grigoriy Petrov | 42m |
Overview of scientific findings about how humans perceive and understand source code, discussing cognitive limits (how many lines a programmer can hold in… | ||
| The Endless War in Memory: A Retrospective on Defenses Against Binary Threats | Sergey Ignatov | 47m |
A retrospective on 35+ years of buffer overflow defenses, surveying technologies developed to eliminate memory-based binary exploits. It covers historical and… | ||
| Applying Heterofunctional Graph Technology | Anton Katenev | 49m |
Describes using a heterofunctional graph-based Control Plane developed at SkalR to manage complex IT infrastructures; the implementation draws on experience… | ||
| Operator Development: Inside K8s controller-runtime | Stanislav Ivankevich | 43m |
Explains internal components of the Kubernetes controller-runtime library—Manager, Controller, Reconciler, Client—and how they interact. Covers controller… | ||
| How to Sell Microservices: A Step-by-Step Plan for Developers | Artem Rybin | 43m |
A practical guide to selling a microservices architecture to stakeholders by explaining how to split an application into independent services with clear… | ||
| Not Omnipotent etcd, or Why It Doesn't Handle Heavy Loads | Aleksey Volkov | 47m |
The talk examines etcd’s limitations under heavy load, explaining why horizontal scaling of etcd clusters is problematic and why etcd isn’t suited for very… | ||
| This Is How I Now Do It: How We Migrated News Sites to Drupal 8 | Aleksey Demyanov | 36m |
A presentation about migrating Drupal 8 news sites from a monolithic PHP setup into Kubernetes, addressing autoscaling and operational management of Drupal in… | ||
| From UDF to Broadcast Join and Back: The Story of a Spark Optimization | Pavel Molchanov | 48m |
A practical case study of optimizing a Spark job: starting from Python UDFs, the talk examines execution plans and performance pitfalls, then refactors the job… | ||
| Storage Architecture for Yandex.Direct Advertising Objects | Bulat Gayfullin | 40m |
Describes the architecture of Yandex.Direct’s storage for advertising objects, replacing MySQL by turning an in-memory hash map into a full database and using… | ||
| Fear and Loathing in Crypto | Sergey Zhuk | 47m |
Covers data structures used for working with cryptocurrencies, the lifecycle of a transaction from a user request to the blockchain, and how to interact with… | ||
| Evolution of the Metrics Pipeline: How the Architecture Scaled with Load | Ruslan Boyarskiy | 48m |
Describes how Tinkoff Bank’s observability platform Sage for telemetry evolved as load increased: which components and storage technologies were used to… | ||
| Inside the New Generation of Alisa | Ruslan Akhtariev | 35m |
Presentation describes the engineering of the new-generation Alice voice assistant: using techniques to reduce response latency so replies start in far less… | ||
| Return of Zero-Copy: How We Hooked up kTLS to Apache Kafka | Ivan Dashchinskiy | 41m |
Shows how the team implemented kernel TLS (kTLS) support for Apache Kafka with a focus on Java integration. Explains the technical approach used to enable… | ||
| How the Fuzzing Process Is Built at Yandex | Pavel Cheremushkin | 41m |
Describes how Yandex automates detection of memory-corruption vulnerabilities by fuzzing its services, detailing the fuzzing platform’s internal architecture… | ||
| One streaming to rule them all: Streaming as the foundation of an analytics ecosystem | Dmitriy Shirokov | 47m |
Describes using a streaming-first architecture to unify and accelerate an analytical ecosystem, including real-time analytics and ML, Big Data/DWH, feature… | ||
| How we replicate and localize 100000 repositories almost daily 🕺🏼 | Aleksandr Kaloshin | 40m |
Describes building a system to replicate and localize about 100,000 Git repositories daily for the Russian market: it automates bulk replication of… | ||
| Reconciling the irreconcilable and accelerating the unacceleratable using Go assembler☠️ | Igor Panasyuk | 53m |
Shows how to speed up Go programs by writing low-level Go assembler snippets and using SIMD instructions not exposed by the standard toolchain. Covers concrete… | ||
| 10 years of evolution in the architecture of networked stateful applications | Pavel Ivashchenko | 41m |
A retrospective of a networking application platform’s architecture evolution from using L3fwd (no real stack) to a full software platform built on an… | ||
| Multitenant Kubernetes: squeezing out maximum efficiency | Aleksandr Kachmashev | 51m |
The talk explains how to run Kubernetes in a multi-tenant environment and maximize its capabilities. It covers concrete multi-tenancy challenges in Kubernetes… | ||
| Optimizing Spark applications from simple to complex with examples | Aleksandr Markachev | 43m |
Covers practical Spark optimization techniques: using a custom script plus manual and automated repartitioning to improve data processing, tuning window… | ||
| Distributed SQL engine in Picodata DBMS: how it works and architectural decisions | Emir Vildanov | 46m |
An overview of Picodata’s distributed SQL query engine: it explains how the engine executes SQL queries across large distributed data volumes, describes its… | ||
| Machine learning and parallel GPU computing in PHP | Aleksey Nechaev | 50m |
Explains why training neural network models on GPUs is faster than on CPUs by describing how GPU computation works and the GPU execution process. Reviews… | ||
| PHP-FPM, (g)unicorn, Puma and uWSGI — will no longer be needed | Valentin Bartenev | 51m |
Overview of how common app servers like PHP-FPM, Gunicorn/Unicorn, uWSGI, Puma (and sometimes Apache) are used together with nginx to run PHP, Python, and Ruby… | ||
| Millions of hours: finding duplicates in VK Video | Nikita Kochetkov | 47m |
VK Video receives hundreds of millions of new videos and needs to detect copies to protect creators. The talk explains the system built to find duplicate or… | ||
| WebAssembly: from browsers to high load | Aleksey Komisov | 52m |
Overview of WebAssembly (Wasm) evolution and its important backend advantages, including performance and behavior under load. Compares Wasm to Java, explains… | ||
| Defending CI/CD against attacks | Aleksey Fedulaev, Andrey Moiseev | 45m |
The talk analyzes attack vectors against CI/CD pipelines, showing concrete examples of what those attacks can achieve and their potential impact. It then… | ||
| YTsaurus SPYT: how we got rid of an Apache Spark fork | Aleksandr Tokarev | 39m |
Describes why the team originally forked and patched Apache Spark for the YTsaurus SPYT project, what operational and maintenance inconveniences those changes… | ||
| From assembler to C++: structure and optimization of CGo ☠️ | Vladimir Pustovalov | 51m |
Explains how Go's goroutine scheduler and the CGo bridge between Go and C/C++ work, focusing on how memory allocations are tracked and how data is marshaled… | ||
| Concurrency in Go: from hardware to code ☠️ | Maksim Goroziy | 44m |
Explains how concurrency primitives in Go (mutexes, atomic operations) relate to hardware and memory models, examining how they are implemented under the hood… | ||
| Conquering the Linux network stack: decapsulating packets with eBPF at 6Mpps+ | Al-dr Shishebarov | 46m |
Describes moving customer traffic to VXLAN and the tools used to analyze that traffic, then evaluates options for VXLAN decapsulation and explains why eBPF was… | ||
| Compressing Java resource usage in containers: 80% of results in 20% of time without OOMs | Denis Nikulin | 50m |
Describes techniques to reduce Java resource usage when running in containers to avoid OOMs and get most gains quickly (the talk claims an 80% of result for… | ||
| How to build Data Lineage on Apache Spark logs | Ramil Mamedov | 35m |
Describes building a Data Lineage tool for Apache Spark used in a large national data warehouse. Covers the problem of tracing data flow from sources to… | ||
| Data Quality vs Everyone | Bogdan Volodarskiy | 47m |
Discussion of why engineers should assume data is low quality by default and treat data quality as a core engineering problem. Covers how bad data undermines… | ||
| Twelve‑factor app for PHP applications: how to become cloud native | Pavel Ivanov | 35m |
Covers applying the Twelve-Factor App methodology to PHP applications to make them cloud-native. Walks through each of the twelve principles and shows how to… | ||
| Artifacts of architecture: which, why and how to organize them | Nail Minnakhmetov | 52m |
The presentation surveys which architecture artifacts (diagrams, documents, models, etc.) teams need in a company, categorizes common types and their purposes… | ||
| Fixing, not hunting: developing and deploying an AI assistant for vulnerability remediation | R. Lebed | 44m |
Describes developing and deploying an AI assistant that automatically fixes software vulnerabilities as part of DevSecOps. Covers using AI to prioritize and… | ||
| How to speed up any training on GPU | Sergey Skorodumov | 39m |
Describes practical techniques Yandex uses to speed up machine learning training on large GPU clusters, covering both per-job and multi-job optimization… | ||
| AI in Go: how we build ML applications using the pipeline pattern | Timofey Milovanov | 40m |
How T-Bank builds machine-learning applications in Go, focusing on the backend engineering rather than model internals. It covers using Go for ML services and… | ||
| How to keep client-side data always up to date | Vitaliy Moskovkin | 44m |
The talk describes problems encountered during incremental evolution of a service where stale data on client services slowed development, and explains an… | ||
| When Your Small Cloud Stopped Being Small | Maksim Yakovlev | 39m |
Covers problems of organic growth of cloud networks at very large scale—how L2 domains spanning tens of thousands of machines arise and how physical and… | ||
| When PowerShell Is Better Than Ansible? Recipes for Managing 1,000+ Servers | Mikhail Sokolov | 49m |
Comparison of PowerShell and Ansible for managing large fleets, showing practical DevOps patterns and business cases for using PowerShell in distributed… | ||
| How We Replaced Hundreds of Joins with One Real‑Time Processing Pipeline at 1M RPS | Maksim Statsenko | 45m |
Describes replacing hundreds of JOINs with a single real-time processing pipeline to handle 1 million RPS for ad events. It explains building a system that… | ||
| A Practical Approach to Using LLMs: Peculiarities and Pitfalls | Ivan Krasnikov | 49m |
Practical guidance on using large language models (LLMs) in real systems, focusing on concrete challenges like hallucinations, performance bottlenecks, high… | ||
| Yii3 | Aleksandr Makarov | 01h 01m |
Yii3 is now modular and mostly ready: 105 of 115 packages released with docs and tests, Active Record advancing fastest and already usable in production. The… | ||
| System Design Interview: To Execute or Pardon | Vl. Nevzorov, Al-dr Polomodov, Filipp Delgyado | 01h 45m |
The talk examines the role and realism of system-design interviews that require candidates to design a complete system in about 45 minutes. It discusses what… | ||
| Anomalies Under Load in PostgreSQL 2.0 | Mikhail Zhilin | 50m |
PostgreSQL performance can degrade even when queries are indexed and tuned; this talk examines anomalous behavior under load that isn’t fixed by classic query… | ||
| How We Sought an IT Approach to a Diverse Fleet of Scooters | Andrey Kalinin | 47m |
Describes reworking system architecture to improve stable connectivity with IoT scooters, avoiding downtime in a high-frequency service by changing how… | ||
| Agilifying Threat Modeling in Development Teams | Nikolay Khechumov | 39m |
Describes a practical approach to catch business-logic vulnerabilities early in the development lifecycle by integrating tooling and processes into development… | ||
| How to Debug an Asynchronous Odyssey Without Attracting the Orderlies’ Attention | Roman Khapov | 36m |
Debugging Odyssey (a PostgreSQL connection pooler) using GDB to find and fix bugs in asynchronous applications. The talk explains practical techniques and… | ||
| Taming Player Technical Events in VK Video | Aleksey Shpagin | 46m |
Compares different approaches to collecting technical telemetry from client apps to backend, and shows how VK Video migrated from one method to another on a… | ||
| WebAssembly and High Load | Maksim Kokryashkin | 34m |
Overview of using WebAssembly runtimes outside the browser, focused on integrating a WASM runtime into the Tarantool database/application server. Covers… | ||
| Architecture in Go and What Rust Has to Do with It | Emil Sharifullin, Anton Kulinskiy | 51m |
Compares architecture approaches when rewriting legacy systems in Go and considers relevance of Rust. Covers applying Clean Architecture, Hexagonal… | ||
| Engineering Culture at Scale: How to Grow and Assess Practices | Evgeniy Kharchenko | 52m |
A new platform called “Technical Capabilities” is presented that automates assessment of engineering practices and collects metrics by integrating with tools… | ||
| Grow a B2C Service or Become a SaaS? We Decided Not to Choose | Pavel Podkolzin | 48m |
The talk explains how a company decided to operate both a B2C product and a SaaS (B2B) offering without choosing one over the other. It describes technology… | ||
| How Data Delivery Evolved for Yandex’s Ad Engine | Daniil Gavrilovskiy | 39m |
A Yandex team describes changing the data format and introducing a new index to avoid constant field reshuffling and to enable fast delivery of updates to… | ||
| How Apache Iceberg Really Works | Vladimir Ozerov | 49m |
Explains Apache Iceberg’s architecture and implementation details as a table format for lakehouse platforms, covering how Iceberg organizes data and metadata… | ||
| 9 Steps for Safe Handling of User Files on a Website | Oleg Ekhlakov | 50m |
Covers secure handling of user-uploaded files on websites and apps, naming common vectors like photos, videos and attachments that can threaten applications or… | ||
| Life After Greenplum: Choosing an Open Source Analytics Solution | Petr Gurinov | 40m |
Examines Greenplum’s role in an existing analytics platform, the expected features from the planned Greenplum 7 release, and how the project’s move away from… | ||
| How to Combine Data from Different DBMSs and Do It Efficiently | Vitaliy Isaev | 49m |
Shows how to join data stored in different database systems and file stores (for example, relational databases and CSV files on S3) efficiently. Covers… | ||
| The Challenge of Processing Power Grid Time Series Data | Sergey Yushmanov | 48m |
Examines processing time-series data from power grids: whether off-the-shelf platforms meet all needs or when in-house development is required, data… | ||
| Identity Aware Proxy — Taking a Meaningful Step Toward Zero Trust | Sarkis Nanyan | 35m |
Identity Aware Proxy (IAP) is presented as a way to move toward a Zero Trust security model by replacing or augmenting traditional VPN access. The talk… | ||
| Speeding Up the Index in VK Video Search | Fedor Petryaykin | 39m |
Describes accelerating the candidate-selection index for VK Video search to handle billions of documents. Explains engineering techniques and system-level… | ||
| Valkey 8 — Release of a Redis Fork Focused on Performance | Evgeniy Dyukov | 41m |
Valkey 8 is a fork of Redis focused on performance improvements; the talk describes that Valkey is fully open-source (no separate Enterprise edition) and… | ||
| To Make Code Fast, All You Need Is… ☠️ | Denis Bozhok | 55m |
Talk examines optimizing Go programs by helping the compiler and using classic low-level tricks; it covers compiler-assisted optimizations, Golang runtime… | ||
| Web over gRPC: Which Technology to Choose 🕺🏼 | Edgar Sipki | 46m |
Compares three ways to use gRPC for web development—gRPC-Web, Buf Connect, and gRPC-Gateway—explaining each approach’s main pros and cons. Presents detailed… | ||
| All You Need for Application Network Security on the Platform | Maksim Chudnovskiy | 48m |
A practical overview of configuring network security for applications running on Kubernetes platforms, covering policies, settings and common misconfigurations… | ||
| Fast Open-Loop and Closed-Loop Simulators for Autonomous Vehicles | Nikolay Polivanov | 39m |
Explains why the team simulates a million kilometers per night for autonomous vehicles and compares Open Loop vs Closed Loop simulation use cases. Describes… | ||
| How Not to Degrade an Ad-Serving Service When the World Goes Crazy | Artem Bukin | 50m |
Explains how an online ad-selection service works and which metrics to monitor, then describes tools and techniques to analyze service health, drill from… | ||
| How I Deleted Clickstream and It Was Restored from Oblivion | Dmitriy Nemchin | 31m |
An engineer describes a data-platform incident where a clickstream dataset was accidentally deleted because some storage systems lacked proper data governance… | ||
| What’s Wrong with Asynchronous PHP? | Vadim Zanfir | 39m |
A walk-through of implementing a minimal asynchronous runtime for PHP to expose how async works at the implementation level. The talk compares PHP’s bare‑bones… | ||
| Artificial vs Natural Intelligence in Annotation Tasks | Aleksandr Kapitanov, Evgeniy Rossinskiy, Evgeniy Makarov, Dmitriy Antipov, Roman Kutsev, Dmitriy Pimenov | 53m |
Compares manual annotation workflows with modern AI approaches across vision, language and audio: covers SAM for detection/segmentation, vision-language models… | ||
| On Geo-Distributed Transactions | Vladimir Komarov | 51m |
Explains limits of databases (both monolithic and distributed) that arise from physical-world constraints, and identifies application properties that let… | ||
| Mobile Payments with Mir | Dmitriy Vikulov | 45m |
The talk examines how the Mir Pay mobile payments application and the Mir Plat.Form team handle steadily increasing load. It details the technologies and… | ||
| How We Learned to Manage Millions of Accounts and Their Secrets | Dmitriy Chekhlystov | 40m |
How to manage millions of user accounts and their secrets by tying account/product/component inventory to support and development teams, and mapping… | ||
| Specialized vs Multimodal Models in Face Liveness | Oleg Grinchuk | 38m |
Comparison of specialized versus multimodal models for face liveness detection, describing how each approach works and how they differ from deepfake detection… | ||
| Cost-Based Optimizer in YDB — How, Why and What For? | Pavel Velikhov | 48m |
YDB added a cost-based query optimizer to improve query planning and execution. The talk explains why the optimizer was introduced, what problems it addresses… | ||
| Fuzzing in Go: Fear, Despair, Acceptance and Long-Awaited Success ☠️ | Roman Ognev | 44m |
Describes building a fuzzing toolchain for Go when no ready-made solutions existed, covering the problems encountered and the team's incremental approach from… | ||
| Sber Smart Home: Growing from 0 to Hundreds of Thousands of Online Devices and Surviving | Vadim Tregubov | 41m |
Sber Smart Home platform handles hundreds of thousands of long-lived online connections to user devices to provide 24/7 real-time device control. The talk… | ||
| A Transformation Story: How We Failed at 20,000+ RPS and What We Learned | Lev Nemirovskiy | 47m |
A postmortem-style case study of a system that suddenly faced over 20,000 RPS at a critical moment, describing the operational and technical decisions taken… | ||
| How to Use a Percolator as Intended and More | Vladislav Sikach | 37m |
The talk explains how the percolator feature works in Sphinx, compares it to ElasticSearch’s percolator implementation, and why Sphinx’s version is performant… | ||
| The Road to Stable and Fast Disks in Yandex Cloud | Daniil Tatyanin | 37m |
Yandex Cloud storage team describes challenges of making block and networked disk storage both fast and reliable for cloud customers. They cover technologies… | ||
| Merkle Forest, or How We Reduced Metadata Size and Also Speeded Up Duplicate Search | Sergey Li | 47m |
Describes how the TATLIN.BACKUP storage system reduced metadata size and sped up duplicate detection in long-term backup storage by redesigning metadata… | ||
| How We Built Fault-Tolerant Billing for BelkaCar Carsharing | Aleksey Solodkiy | 49m |
Describes how BelkaCar implemented a fault-tolerant pricing, tariffing and billing system for carsharing by integrating those services so the overall system… | ||
| Microservices on InfiniBand: 800 Gbps in Distributed Training | Aleksey Morozov | 49m |
Describes using InfiniBand to run microservices for distributed training of a ranking model at Yandex. Focuses on handling 1+ PiB datasets requiring… | ||
| Incognito Won’t Help: Understanding Fingerprint Identification | Aleksandr Bobryakov | 48m |
Examines current browser and device fingerprinting algorithms, detailing how they compute fingerprints, their strengths, weaknesses, and limits. Discusses… | ||
| How to Load a DB Without Load Testing During a Migration to PostgreSQL | Yaroslav Kiselev | 46m |
Discusses migrating databases to PostgreSQL when conventional load testing isn’t available, outlining risks from lack of ready load tests and techniques to… | ||
| The Evolution of SSH Authentication: From Passwords to Today | Eldar Zaitov | 38m |
Survey of SSH authentication history explaining how methods evolved from passwords to modern mechanisms like public-key authentication and agent forwarding… | ||
| Challenges Scaling a Service to Evade Anti-Bot Systems | Evgeniy Slaykovskiy | 39m |
Examines the ongoing arms race between anti-bot defenses and services that try to bypass them, describing the current state of that battle. Discusses tensions… | ||
| WASM: Purpose, Architecture, Prospects | Vladimir Khomutov | 41m |
Explains WebAssembly (WASM): its origin, original problems it aimed to solve, and current use cases across environments from browsers to the Linux kernel… | ||
| Workshop: How to Use Temporal to Build an MVP | Evgeniy Konechnyy | 01h 47m |
An instructor-level session on using the Temporal workflow engine to build an MVP, covering concrete use cases like order and payment processing, complex… | ||
| Instant code navigation for any commit ☠️ | Olga Lukyanova | 48m |
Describes a solution for instant code navigation for any commit when reviewing pull requests in web interfaces. Names the problem: web code viewers lack proper… | ||
| Centralized logging and sleeping well at night | Filipp Bocharov, Yuliya Taltskova | 46m |
Describes building a geo-distributed centralized logging system on OpenSearch storing 300+ TB and serving 3,000 users. Covers how the architecture and tech… | ||
| Artificial intelligence in software development: current technologies and prospects | Egor Baranov | 46m |
Overview of current AI technologies used in software development, focusing on architectures and operation illustrated by the product GigaCode. Covers problem… | ||
| Challenges of processing high-volume camera data streams | Dmitriy Barinov | 46m |
Presentation describes the data-processing pipeline for the Spectro robot’s camera scanning system, focusing on obtaining synchronized frames from multiple… | ||
| How to fix policy violations while preserving content quantity and quality | Vladimir Morozov | 41m |
Discusses moderation strategies for user-generated images, comparing strict blocking to hiding violations and arguing that masking violations can improve user… | ||
| Exceptions in PHP: more complicated than it appears | Ilya Lazarev | 51m |
The talk examines PHP exceptions: it surveys the exception class hierarchy and the lack of strict standards, then analyzes subtleties of exception handling and… | ||
| Hybrid search based on OpenSearch and Qdrant | Egor Prokhorenko | 49m |
Describes implementing hybrid search combining OpenSearch full-text search with Qdrant vector search at T-Bank, outlining the engineering problems encountered… | ||
| How we at Avito analyze 5 million traces and conduct architectural oversight | Ivan Neshchadin | 46m |
Avito describes processing five million distributed traces to build a service-interaction map of their microservices using tracing tools and related… | ||
| Picodata: lots of small data | Konstantin Osipov | 42m |
Picodata is a distributed DBMS built on Tarantool as the local storage and replication layer. On top of that they implemented a new distributed SQL engine and… | ||
| Key aspects of securing cloud infrastructure | Nikolay Panchenko, Elizaveta Leonova | 48m |
Covers security considerations for cloud infrastructure in rapidly growing IT companies, focusing on challenges that arise when provisioning new hardware and… | ||
| Integration evolution through a PaaS platform | Aleksandr Knyazev | 48m |
App.Farm is a PaaS platform used at Rosselkhozbank (RSHB) to manage integration flows and the application execution environment. The talk describes how… | ||
| Fintech: the reasons I hate it | Ekaterina Lysenko | 46m |
An exploration of why fintech has become pervasive across industries like e-commerce, e-grocery, and betting rather than being limited to banks, identifying… | ||
| I'll teach you functionalism in Go! | Vladislav Belogrudov | 51m |
Explains how to use pure functions and other functional-programming techniques to improve code safety in Go, surveying which FP elements are already used in Go… | ||
| Building a private cloud platform with 20,000 VMs – Wildberries' experience | Aleksey Churkin | 47m |
Wildberries describes building its own cloud platform to support about 20,000 virtual machines as part of a technological sovereignty strategy. The talk covers… | ||
| Speeding ML model inference without unnecessary cost | Denis Zakharov | 39m |
Describes how DomKlik uses neural networks for voice and text bots and examines when moving models to GPUs is not the best option; discusses practical… | ||
| A cloud-native approach to working with Ceph | Igor Shishkin | 41m |
Describes using Ceph as an easily deployable storage cluster and how to run functional testing on virtual machines. Explains the workflow from developing… | ||
| How we brewed the Gigachat Pretrain data | Ivan Baskov | 48m |
Describes how pretraining datasets for GigaChat are prepared: mapping web-data clusters across Russian and English segments, deduplicating massive volumes… | ||
| How PHP died | Oleg Mifle | 46m |
An exploration of PHP’s perceived decline: it traces how critics declared PHP ‘dead’, examines the technologies and projects that were positioned as its… | ||
| Adding ++ to Prometheus | Vladimir Guryanov, David Megton, Evgeniy Bastrykov | 55m |
Shows how the team reduced Prometheus resource use by tens of times by reimplementing its Time Series Database in C++ and optimizing encoding and storage… | ||
| Extracting a microservice from a 15-year-old monolith: a one-year adventure | Stanislav Kondratev | 48m |
A team describes extracting a microservice from a 15-year-old monolith, covering the full year-long process. They explain moving from an initial modular split… | ||
| How KION builds realtime personalized recommendations and storefronts | Aleksey Zhiryakov | 43m |
KION describes how it routes content into recommendation systems and personalized storefronts, including the development of a TV-specific storefront. The talk… | ||
| Bastion DB: enabling secure developer access to databases | Ilay Kobrin | 39m |
Bastion DB describes a solution for secure developer access to production databases by replacing or augmenting password-based access with a controlled bastion… | ||
| OpenRAN — the gateway to telecom for everyone | Nikolay Morozov | 56m |
OpenRAN is presented as a set of specifications that enable small teams to design and build individual RAN (radio access network) nodes on modest budgets. The… | ||
| How we built recommendations, cut the vendor, and made money | Danila Fedyukin | 40m |
Describes building an in-house recommendation system that replaced a vendor off-the-shelf solution and generated additional revenue. Covers practical… | ||
| High-load CGo: calling nonblocking C/C++ code without surprise ☠️ | Evgeniy Chormonov | 34m |
Explains pitfalls and techniques for calling non-blocking C/C++ code from Go using cgo, focusing on issues that arise with asynchronous OS operations and heavy… | ||
| Go developer grades, or What distinguishes a senior gopher from the rest | Vitaliy Levchenko | 54m |
Looks at how the profession of Go developer is evolving and critiques common interview practices like LeetCode puzzles and rote questions about threads vs… | ||
| What a product team must do BEFORE a full system blackout | Natalya Makarova | 48m |
A CDEK engineering team shares practical pre‑blackout preparations for enterprise systems, describing what must already exist in code and in engineers' minds… | ||
| Spark in K8s for dozens of DS teams | Arsen Gumin | 52m |
Describes how Spark workloads were implemented on a recommendation-system platform, detailing problems encountered with Spark and the reasons for the chosen… | ||
| Scaling under uncertainty | Mariya Nasyrova | 45m |
Covers preparing and scaling a system for launch to a multi-million user audience: architectural choices, selecting deployment cadence and throughput targets… | ||
| State machines for product data from YDB and queues | Vitaliy Moskovkin | 47m |
Describes redesigning a product-data management system built on YDB and queueing to handle over a billion items, explaining why prior evolutionary scaling… | ||
| PHP 8.4: more alive than the living | Valentin Udaltsov | 55m |
Overview of the upcoming PHP 8.4 release and its notable features: a redesigned JIT, new hooks and asymmetric property visibility, new syntax allowing new… | ||
| Speeding up data labeling with neural nets: pipeline, metrics and life hacks | Dmitriy Antipov | 50m |
Describes building a data-labeling pipeline that uses neural networks to speed up complex annotation tasks, detailing the system architecture, the metrics used… | ||
| API Gateway at Avito | Aleksandr Dankovtsev | 51m |
Description of the API Gateway built at Avito: explains the technologies and design decisions used to implement their gateway, the problems it solves for… | ||
| So, you decided to reliably write data to disk | Dmitriy Rodionov | 53m |
Covers why durable disk writes matter as much as query speed for databases, examines common causes of data loss during persistence, and reviews techniques for… | ||
| Digital signature, the evolution | Sergey Prilutskiy | 49m |
Covers the evolution of digital signatures from simple symmetric MACs through threshold and group signatures to post-quantum signature schemes and even… | ||
| Under the hood of Wiremock: unraveling a performance problem | Sergey Gogolev | 28m |
An investigation of WireMock used as both mock server and proxy revealed serious performance failures: WireMock crashed under about 100 RPS, Kubernetes… | ||
| Searching video content with AI | Aleksandr Sokolov | 50m |
Describes extracting features to search for a specific frame across a collection of 40,000 videos using vector search; explains techniques to keep the vector… | ||
| The best mock in the world: examining mock generation tools in Go | Semen Eygin | 50m |
A comparison of mock-generation tools for Go: Mockery, Gomock, and Minimock. The talk evaluates each tool on practical examples, discussing pros and cons, ease… | ||
| Speeding up Go by designing the app with CPU architecture in mind ☠️ | Aleksandr Shakmaev | 47m |
Explains how CPU memory and cache behavior affects Go application performance and shows techniques to design Go programs with CPU architecture in mind. Covers… | ||
| How (not) to survive a data center outage | Nikolay Butenko | 47m |
VK Cloud's postmortem of a data center outage: how they restored services and adjusted the system architecture without client or service downtime. The talk… | ||
| Binary Java: CDS, CRaC and AOT for faster JVM startup and warmup | Rustam Kuramshin | 50m |
Covers Java startup and warmup improvements focusing on CRaC (Coordinated Restore at Checkpoint) and GraalVM/AOT techniques to create pre-initialized or… | ||
| Peculiarities of modern hardware | Maksim Davydov | 38m |
A year-long study in Yandex Cloud investigating performance issues caused by contention for L3 CPU cache and the memory bus. It covers methods for detecting… | ||
| A new turn in building analytics on embedded databases | Aleksandr Borgardt | 46m |
The talk explains how embedded databases can replace heavy DWH and OLAP systems to reduce complexity, resource use, and cost. It covers practical approaches… | ||
| How to work with legacy so neither you nor the project burns out | Aleksey Obyskalov | 52m |
Explains what legacy code is, how legacy systems arise, and when refactoring is genuinely necessary. Describes how to estimate the costs of refactoring and… | ||
| AppHost: how Yandex organizes interaction of hundreds of microservices | Konstantin Vukolov | 50m |
Yandex describes AppHost, an internal platform that coordinates interaction among hundreds of microservices. The talk explains how AppHost traces request… | ||
| The development of 'civil' cryptography. Releasing mobile app updates with FIPS-certified crypto | Anton Meluzov | 41m |
Discusses organizing the development and release process for mobile apps that include certified cryptographic protection modules (СКЗИ), focusing on meeting… | ||
| A testing system for management services of enterprise wireless networks | Aleksandr Loshkarev | 38m |
Describes a test system for services of a software-hardware wireless access controller that manages thousands of Wi‑Fi devices. It explains creating a test… | ||
| Freedom, equality, gophers 🕺🏼 | Mariya Letta | 44m |
A presentation about creating and evolving the Free Gophers Pack for the Go community: how the original Go mascot assets were limited (no facial expressions… | ||
| Optimizing concurrent applications | Maksim Miroshnichenko | 46m |
A clear algorithm for designing concurrent applications is presented, focused on solving performance problems and building efficient high-load systems. The… | ||
| How to stop being afraid and love Open Source. The story of creating one product | Dmitriy Kudyukin | 39m |
Talk describes how a Russian software team built a product based on open source to replace departed Western vendors after 2022 sanctions. It covers the problem… | ||
| How to Run a Metropolis Without SMS or Registration: The Story of an ASUDD Pilot | Arseniy Sazanov | 52m |
An engineer describes a pilot deployment of an urban traffic control system (АСУДД) that operates without SMS notifications or user registration. The talk… | ||
| How We Built a Shared Bus of Dynamic Resources at Yandex | Nikolay Gayvoronskiy | 42m |
Describes Yandex’s design for a shared dynamic-resource bus that allows updating a resource’s implementation inside a container without restarting the process… | ||
| The Importance of Component Testing When Testing Architecturally Complex Systems | A.Isaeva | 11m |
Discusses why component-level testing is essential for architecturally complex, high-load systems. Covers practical testing approaches and practices for… | ||
| Portrait of a Systems Analyst: The Mobile SMB App Team Case | Gleb Dobrov | 10m |
A portrait of a systems analyst working on an SMB mobile app, describing the analyst’s role, responsibilities and skills within that product team. Covers… | ||
| Code as a Form of Art | Denis Rodin | 09m |
Explains how code can be treated as an art form, discussing aesthetic and expressive qualities of programming rather than purely technical aspects. Covers… | ||
| How to Build Reliability Starting from the CJM | Vladimir Perfilev | 10m |
Explains how to build reliability starting from the Customer Journey Map (CJM) by using CJM-driven analysis to identify failure modes and design improvements… | ||
| Open Source Low Code Technologies in Import Substitution | Ivan Ratnikov | 10m |
Survey of open-source low-code platforms used for import substitution in Russia, comparing specific technologies and how they can replace proprietary systems… | ||
| Payment Links: Types and How They Work | Artem Boyarshinov | 11m |
An overview of different kinds of payment links, how they are implemented and organized, and how they work in payment systems. The talk discusses technical… | ||
| Performance of Financial Systems: What You Must Not Forget | Aleksandr Lantsov | 11m |
Covers performance considerations specific to financial systems, focusing on what must not be forgotten when designing and operating them. Discusses practical… | ||
| Building an Architecture Documentation Framework from Scratch | Dmitriy Zharikhin | 11m |
A presentation describes building a documentation framework for system architecture from scratch, covering design and implementation of tools to capture and… | ||
| Opening | Oleg Bunin, Nastya Demidova | 30m |
Announcement and logistics for the HighLoad++ professional conference for developers of high-load systems, including dates and venues: Saint Petersburg (June… | ||
| Closing | Nastya Demidova, Martin Shtern | 20m |
A closing note for the HighLoad++ conferences announces dates and locations for upcoming editions: Saint Petersburg (June 2026 at DESIGN DISTRICT DAA) and… | ||
| nothing matches | ||