Saint HighLoad++ 2023
2023-06-26..2023-06-27
Saint Petersburg, Russian Federation
[Saint HighLoad++ 2022, Saint HighLoad++ 2023, Saint HighLoad++ 2025]
| title | author | length |
|---|---|---|
| Video Report from Saint HighLoad++ 2023 | 02m | |
A video report about the Saint HighLoad++ 2023 professional conference for developers of high-load systems, held 26–27 June 2023 in Saint Petersburg at DESIGN… | ||
| Everything Everywhere at Once: Tracing under High Load with 10k Backends | Aleksandr Kirsanov | 54m |
Describes expanding distributed tracing to meet VKontakte’s needs in a very high-load environment with around 10,000 backends. Covers collecting cross-service… | ||
| Delivering Telemetry from a Warehouse Robot and What to Do with the Data | Nikolay Dolganov | 43m |
Delivery of telemetry from warehouse robots developed at Yandex Market: the talk describes the telemetry stack for robots that have many compute nodes… | ||
| Group Chats on Odnoklassniki | Andrey Timofeev | 40m |
Describes how Odnoklassniki/TamTam handles extremely high messaging load (about 50M MAU and over 10 billion events per day) and the engineering challenges of… | ||
| CI/CD Security | Aleksey Fedulaev, Andrey Moiseev | 48m |
Examines security concerns in automated CI/CD pipelines for high-load applications, describing risks introduced by continuous integration and deployment and… | ||
| How to Build Search for an Online Store | Stepan Rodionov | 47m |
Explains the main stages of building a search system for an online store, using a large hypermarket as a concrete example. Covers the key technological and… | ||
| 5 GB of Traces per Second, or How Ozon's Tracing System Works | Vladimir Balun | 42m |
Description of Ozon’s request-tracing system that handles 5 GB/s of traces, covering the system architecture, the trace processing pipeline, and storage… | ||
| Combining Stream and Batch Processes in Camunda BPM | Maksim Pavlov | 40m |
Explains combining streaming and batch processing within Camunda BPM, giving practical examples of process diagrams, techniques, and patterns. Describes how… | ||
| You Need a CDN! ...or Maybe Not? | Aleksey Uchakin | 51m |
Explains what a CDN is, how different CDN providers and architectures actually differ, and how to choose the right one. Covers CDN internals—server deployment… | ||
| Managing Access in a Distributed Linux Infrastructure | Anton Zhabolenko, Pavel Parkhomets | 51m |
Explains privileged access management (PAM) for distributed Linux infrastructures: what a PAM system is, why it’s needed, and the requirements it must meet… | ||
| Searching for Products by Photo | Ivan Krasnikov | 41m |
A practical case study of integrating machine learning into production to implement visual product search for online stores, using embeddings and vector search… | ||
| Microservices as the Worst Architectural Choice for a Startup | Daniil Podolskiy | 52m |
Argues that adopting microservice architecture is often the worst architectural choice for startups. Explains problems encountered with microservices in… | ||
| Warden — Why We Need Our Own Service Mesh? | Ilyas Nezhibitskiy | 48m |
Describes why Ozon built their own service mesh called Warden instead of using existing solutions like Nginx or Istio; it focuses on problems of service… | ||
| Hit 350 Milliseconds: The Evolution of Mail.ru Mail Anti-Spam Architecture | Konstantin Khitsko | 52m |
Describes Mail.ru Mail’s anti-spam system that processes over 400 million emails per day and the architectural evolution to meet a 350 ms processing target… | ||
| Feature store | Anastasiya Tushkanova | 44m |
Search handles ~150 million products with ~10k RPS typical and 35k peaks, pulling up to 2000 candidates per query into ranking. Redis hotspotting with up to 18… | ||
| Replacing Thales — Hardware Encryption Modules | Dmitriy Gordievskiy | 34m |
Discussion of replacing Thales hardware security modules (HSMs) for card data encryption and transaction crypto-security after Thales exited the Russian market… | ||
| Modern Databases: How to Choose an RDBMS in 2023? | Oleg Bondar | 52m |
Overview of the current DBMS landscape in 2023, comparing well-known database systems and offering an objective view of their strengths and trade-offs. Also… | ||
| Use actors, Luke | Andrey Paramonov | 48m |
Company runs payments in ~20 countries and builds a distributed payments coordinator using virtual actors (Orleans/.NET) that host a tiny in-memory virtual… | ||
| Lessons from a Project with a 2001 Commit | Vasiliy Romanov | 49m |
A retrospective on a long-running project started with a 2001 commit, focusing on engineering practices that stay valuable over a ten-year span. Covers… | ||
| Generative Diffusion Models: Development, Training and Releasing Kandinsky 2.1 | Andrey Kuznetsov | 49m |
Covers diffusion-based image generation from text, explaining the diffusion process, types of diffusion models, and how they are used to generate images… | ||
| The Five Stages of Accepting Cybersecurity | Anton Bochkarev | 40m |
A talk maps companies’ journeys toward adopting cybersecurity onto five stages—denial, anger, bargaining, depression, and acceptance—and describes the… | ||
| How We Survive Split-Brain and Keep Writing Data via the S3 Protocol | S. Bogatyrev | 52m |
An object storage system that uses the S3 protocol is deployed across multiple geographically distributed sites and must continue accepting writes even during… | ||
| Using Images in Recommender Systems | Ruslan Vergunov | 25m |
The talk surveys how images are used inside recommender systems, covering techniques for incorporating visual information to influence click decisions and… | ||
| Making Sberbank Online More and More Reliable: How We Prepare for Peak Loads | Artem Aryutkin | 52m |
Describes how Sberbank Online improves reliability and prepares for peak loads by building processes and resilience for a high‑load financial service. Covers… | ||
| Efficiently Updating Database State from Stream Processing Services | Yuriy Pechatnov | 41m |
Describes how Yandex Ads stores and updates complex entities (like advertising banners) in databases for streaming-processing services. Explains problems of… | ||
| Why You Should Write Your Backend in Elixir | Roman Smirnov | 44m |
Explains why Elixir is a strong choice for backend development by comparing it to many other programming languages and highlighting its practical benefits… | ||
| YTsaurus: Operational Experience with a Storage Cluster of 180K Disks | Pavel Sushin | 54m |
Describes operation of YTsaurus, Yandex’s big-data storage and processing platform used for ad-hoc analytics, ETL and batch jobs. Presents the challenges and… | ||
| Rethinking Recommendations in Zen and Implementing an Item-to-Item Scheme | Dmitriy Shishov | 41m |
Describes how Yandex Zen reworked its recommendation system and implemented an item-to-item recommendation scheme to improve personalized feeds of articles… | ||
| OpenAPI Platform: What Kind of 'Beast' Is It and How Do You 'Eat' It? | Ilya Volynkin | 43m |
An exposition of how an OpenAPI-based platform is built and operates, using the TelecomAPI platform Exolve as a concrete example. It covers building… | ||
| Legacy PHP-FPM on Kubernetes: The Ops Team's Thorny Road | Igor Titov | 35m |
A practical migration story about moving legacy PHP-FPM applications into Kubernetes. Covers techniques for horizontal scaling of PHP-FPM and strategies for… | ||
| The Evolution of Replication in Tarantool | Sergey Petrenko | 44m |
Covers the history and design changes of Tarantool replication: it traces the system from original master–replica without transactions through addition of… | ||
| SBOL: Architecture of the Storage Layer | Ilya Kayshauri | 53m |
Describes the storage-layer architecture built for Sberbank Online, including the path the development team followed to create a reliable, fault-tolerant… | ||
| Anti-Fraud in Reverse and the Use of ML Methods in It | Aleksandra Bazhenova | 45m |
Discusses turning traditional anti-fraud systems inside out by using machine learning to detect legitimate human buying behavior rather than only blocking… | ||
| Roasting Architecture Case Studies | Vladimir Perepelitsa, Sergey Mikhalev, Evgeniy Kuzovlev, Evgeniy Rossinskiy, Aleksey Losev, Mona Arkhipova | 01h 42m |
Open expert review sessions analyzing participants' system architecture cases live at the conference, focusing on real-world high-load system designs. Experts… | ||
| Moving to Platform as a Service in Yandex Verticals | Konstantin Kasev, Mariya Vasileva | 47m |
Yandex Verticals describes moving from manual, server-focused deployments (deb packages, file-based logs, slow deployments, ops-managed monitoring and alerts)… | ||
| YTsaurus SPYT: Helping the Apache Spark Scheduler Be Even More Efficient | Aleksey Shishkin | 40m |
YTsaurus SPYT describes using the YTsaurus storage and its SPYT integration to reduce or avoid expensive Apache Spark shuffle stages during GROUP BY and JOIN… | ||
| Internal Cloud at Yandex — From Prototype to Platform | Dmitriy Lipin | 48m |
Description of Yandex’s internal cloud platform: why the company built an internal cloud to provide developers with delivery, deployment, monitoring, and… | ||
| MirAccept Protocol 2.0: What Awaits Us Next | Aleksey Kasyakin | 46m |
Explains the MirAccept 2.0 payment protocol evolution for card and online payments, describing why PIN and SMS confirmations are becoming less common and what… | ||
| How to Migrate 4K People to an In-House Incident and Alert Management System in a Month | V. Vitkovskiy | 41m |
Describes replacing Opsgenie (a SaaS incident-and-alert-management product) with an in-house incident and alert management system at Ozon within one month… | ||
| Universal Pushes for Android: FCM, HMS and RuStore in One Place | Artem Kovardin | 41m |
Covers implementing unified push notifications on Android by integrating Firebase Cloud Messaging (FCM), Huawei Mobile Services (HMS) and regional stores like… | ||
| Distributed Graph DBMSs — The Future for Analytics on Big Data? | Pavel Velikhov | 54m |
The talk examines distributed graph databases for big-data analytics, contrasting them with traditional single-node systems like Neo4j and addressing concerns… | ||
| How to Profile When perf Doesn't See Everything? | Maksim Kokryashkin | 40m |
Shows how to get full-performance visibility for platforms that embed interpreters (VMs) when traditional tools fall short. Describes combining VM-specific… | ||
| Reliability of Data Delivery in Stream Processing | Evgeniy Ravnushkin | 41m |
A presentation about ensuring reliable data delivery in streaming processing for Yandex.Market's offers storage, which stores billions of records and processes… | ||
| Peeking into Linux Infrastructure to Detect Attacks | Anton Zhabolenko, Liana Ostapchuk | 39m |
Explains how to build information-security monitoring for high‑load, distributed Linux infrastructure, contrasting security monitoring tools with traditional… | ||
| From Algorithm to Production: How to Approach Verification of Distributed Systems | Nikita Galushko | 53m |
Describes how VKontakte’s infrastructure team verifies distributed systems that support vk.com, focusing on techniques and practices used to catch high-cost… | ||
| Why Did You Fork NGINX, My Dear?! (Na^WDo You Even Need To?) | Ivan Poluyanov | 41m |
A half-year development report on the Angie web server: what changes were made, what problems were solved, and what further work is needed. The talk also… | ||
| Performer Product Platform in Yandex Go | Sergey Priymenko | 41m |
Yandex Pro is the unified app for workers across Yandex’s O2O services (Taxi, Scooters, Delivery, Food, Shops, Market). The talk explains the architectural… | ||
| CTF as a Service | Lev Khakimov | 38m |
Claim: running CTFs on Kubernetes fixes scale and stability issues the usual platforms hit. Built a ctfd operator that deploys isolated ctfd instances per… | ||
| Processing Terabytes of Data in Credit Scoring | Evgeniy Smirnov | 48m |
Describes how a bank moved from traditional interpretable scoring methods to neural networks that make credit decisions using terabytes of data. Covers the… | ||
| How to Reach Consensus with Paper, Pencil and the Raft Algorithm | Yaroslav Dynnikov | 01h 32m |
A practical exploration of the Raft consensus algorithm, explaining how Raft achieves agreement in distributed systems using only simple tools (paper and… | ||
| VK Calls: Objectively Assessing Video Quality | Aleksey Shpagin | 45m |
Describes objective video quality measurement for VK Calls: identifies causes of frozen, pixelated or slide-show-like video in calls, compares which artifacts… | ||
| We Cool with Air: Yandex's Data Center — What Is It Like? | Egor Gordovskiy | 52m |
Overview of how Yandex designs, builds and operates its air-cooled data centers, including the company’s in-house hardware and software developments that… | ||
| SAST'madness and Courage | Aleksey Morozov | 50m |
A survey of static application security testing (SAST) tools: types of static analyzers, differences between them, and a comparison. Practical guidance on… | ||
| 7 Petabytes of Logs in Elastic. How We Did It | Roman Nikolaev | 54m |
Tinkoff describes scaling their in-house monitoring platform Sage, which uses Elasticsearch to collect logs and metrics from thousands of services. The talk… | ||
| Feed and Recommendation Architecture at VKontakte | Andrey Yakushev | 46m |
Architecture and engineering of VKontakte’s feed and recommendation systems, showing how the feed serves as the primary interaction point for 100 million users… | ||
| GPT3+ Models in Cybersecurity Tasks | Artem Bachevskiy | 48m |
Overview of how GPT-3‑style and newer large language models are already being applied in cybersecurity by defenders, developers, and administrators. Covers… | ||
| Avito Auto Upload: How to Download Millions of Photos per Day | Evgeniy Tolmachev | 46m |
Avito AutoUpload handles tens of millions of listings per hour and downloads millions of photos per day from third‑party hosts of widely varying quality and… | ||
| MinIO — What Changed Over the Year | Aleksey Pletnev | 40m |
A recap of changes in MinIO over the past year, focusing on deploying a geo-distributed, highly available S3-compatible storage built on MinIO. Covers… | ||
| YTsaurus — The Future of DWH, Already Here in Yandex Market | Filipp Kozmin | 48m |
YTsaurus is presented as a replacement for traditional DWH stacks: the team needed a 30 PB storage with 10 PB/year growth and strict ingestion SLAs, found… | ||
| How to Make Stable Diffusion Generate Not Just Anime and Cats, but You | Kirill Semin | 46m |
Overview of methods for personalizing Stable Diffusion-style text-to-image models so they can generate specific people rather than just common subjects like… | ||
| How We Manage Infrastructure Across 1,000+ Servers Using ansible | A.Kuznetsov, B.Murashin | 48m |
Describes managing infrastructure for over 1,000 servers using Ansible to keep Git-stored configuration synchronized with live servers. Explains the scaling… | ||
| Implementing YDB CDC: The Yandex Monitoring Case | Egor Litvinenko | 26m |
Describes deploying YDB Change Data Capture (CDC) in Yandex Monitoring to stream and replicate metric writes. Covers using YDB CDC to handle very high write… | ||
| A Quest to Synchronize Analytical and Operational Stores in Real Time | Andrey Komyagin | 54m |
Describes a practical walkthrough of synchronizing analytical (data warehouse) and operational (OLTP/real-time) stores for high‑load systems. Covers concrete… | ||
| Endogenous Threats to the Physical Integrity of DBMSs | Andrey Borodin | 51m |
Examines internal (endogenous) faults inside DBMS software that can bypass redundancy and backup protections, causing dramatic effects. Describes specific… | ||
| Oh, How I Want to Return, to Burst Back into the Monolith! | Pavel Lakosnikov | 50m |
Discusses the trend of adopting microservices and the hype around them (alongside buzzwords like Bluetooth and blockchain), questioning whether microservice… | ||
| Building Your Own Smart Home on a Shoestring | Andrey Loginov | 01h 15m |
Build a simple, low-cost DIY smart home across four platforms by assembling devices live and showing how each platform supports common smart-home features. The… | ||
| Reinventing the File System: OpenZFS | Georgiy Melikov | 44m |
Presentation explains OpenZFS as a modern file system solution for storage challenges (hardware faults, caching, guarantees, performance, recovery speed, and… | ||
| Alisa, Don’t Lag! | Arslan Urtashev | 53m |
Describes efforts to make Yandex’s voice assistant Alice respond faster by improving its response latency. Covers Alice’s system architecture and how that… | ||
| CPU Isolation by Memory Bandwidth | Lev Pliner | 53m |
Describes methods for isolating CPU performance based on memory bandwidth in multi-tenant systems. Covers how systems track resource usage (quotas, hard/soft… | ||
| A Revolution in Data Management — The Dawn of Graph Databases | Denis Pchelintsev | 01h 05m |
Explains why graph databases are becoming essential for data management by comparing relational models to graph models, and shows business use cases, goals… | ||
| The Dark Gods of Enterprise Architecture. To Warp and Back | Andrey Zhukov | 37m |
Examines why different approaches to corporate software architecture—formal design documents, ad-hoc team-driven designs, and sprint-based changes—nonetheless… | ||
| HighLoad for the "Small" | Oleg Romanenko | 41m |
MediaSniper's journey from an empty repo to a leading Russian ad platform is described, focusing on engineering choices for a small startup without unlimited… | ||
| ORC and Parquet: Formats and Their Use on HDFS | Aleksandr Markachev | 39m |
Compares the ORC and Parquet columnar file formats for analytics on HDFS, explaining their features, trade-offs, and usage patterns. Covers practical… | ||
| Trunk-Based Continuous Deployment in a Bank | Maksim Morev | 09m |
Describes implementing trunk-based continuous deployment in a bank to enable releases from push to production in under 10 minutes. Covers changing the whole… | ||
| ML Research in Zen: How to Live Without KPIs? | Andrey Zimovnov | 11m |
Describes how the ML teams behind Yandex Zen handle research-driven recommendation development without relying on fixed KPI targets. Covers the platform’s… | ||
| Making Security Requirements Understandable and Achievable for All Company Products | N.Andreev | 11m |
Beeline’s approach to making information-security requirements clear and actionable across many products: replace the traditional security-as-a-gatekeeper… | ||
| The MIR Payment System and Its (Non)Personal Data | Mikhail Arkhipov | 10m |
Overview of the data available and analyzable within the MIR payment system used by banks: beyond just transaction records, the talk identifies other types of… | ||
| How We Integrated Load Testing into Unit Tests | Kirill Borisov | 10m |
Shows how to embed load testing into unit tests to support fast delivery in independently operating teams. Describes tools and practices used to run… | ||
| How We Write Code at Zen: move fast and break things | Dmitriy Kondrashkin | 11m |
A talk about how the engineering team at Zen builds and ships code for a large content platform that serves 80.7 million monthly users. It describes the… | ||
| NT as a Service | Anton Dikin | 08m |
Describes integrating non-functional testing—especially load testing—into production workflows to stabilize modern IT systems. Covers benefits such as… | ||
| Dispute Procedures in the Mir Payment System | Roman Nazarov | 11m |
Overview of dispute-handling procedures used in the Russian MIR payment system: describes the small fraction of card transactions that require dispute… | ||
| Opening | Oleg Bunin | 21m |
Announcement and basic information about the Saint HighLoad++ 2023 professional conference for developers of high-load systems, including dates and location… | ||
| Closing | Oleg Bunin | 11m |
A short closing segment for the Saint HighLoad++ 2023 professional conference in Saint Petersburg, held at DESIGN DISTRICT DAA and sponsored by Garage Eight… | ||
| nothing matches | ||