коммерсант

Few engineers outside Russia have peeked under the hood of Коммерсант's digital ecosystem-here's what makes its mobile news delivery so resilient under load.

Коммерсант isn't just the oldest surviving business daily in Russia; it is a case study in how a print-born institution can evolve into a high-traffic, digitally native platform. Senior engineers who have worked on content-heavy mobile applications will immediately recognize familiar patterns: a React front-end, a Next js-powered server-side rendering layer, and caching strategies borrowed from the playbooks of Vox Media and The Atlantic. But the engineering decisions that keep Коммерсант's iOS and Android apps fast, its push notifications timely. And its article pipelines accurate under both DDoS and traffic spikes from breaking macroeconomic news are uniquely grounded in a Russian internet ecosystem that operates under a different set of constraints-from latency patterns across 11 time zones to strict data localization laws.

I've spent the last decade designing mobile news apps for high-readership outlets. And when I started reverse-engineering the Коммерсант mobile clients last year, I was surprised by the deliberate simplicity of their caching layer and the advanced Russian-language natural language processing (NLP) that powers their "Сюжеты" (storylines) feature. In this article, I'll dissect the platform's architecture layer by layer, share concrete production metrics where available. And extract five engineering heuristics that any team building a content-heavy mobile product can apply today.

The Digital Platform That Powers Russia's Business News

Коммерсант's website is a statically generated affair that looks deceptively simple. But a quick glance at the network waterfall reveals a Nextjs incremental static regeneration pattern combined with a Varnish cache sitting behind an Nginx reverse proxy. This stack-which I first saw documented in a now-archived SpbLinux community talk by a former Kommersant developer-is designed to absorb enormous spikes when the Russian Central Bank announces a rate hike or when energy market shocks hit. On a regular day, Коммерсант's mobile web performance hovers around a Lighthouse score of 89; during the February 2022 currency volatility, the team's patch notes revealed they were pushing edge cache TTL adjustments within minutes.

The content management system (CMS) is no off-the-shelf WordPress installation. Multiple former employees have referenced a proprietary editorial tool called "Edition" that interfaces with a Postgres-backed metadata layer. Articles are composed in a custom XML schema that allows journalists to embed financial tables, stock tickers. And interactive widgets-all of which must render consistently on feature phones still common in Russia's regional centers. That constraint drove an interesting engineering choice: the team built a transpiler that converts the XML into a lightweight JSON structure at publish time, stripping out any non-essential tags before the content reaches the CDN.

Мощный Стек: From Print Legacy to Cloud-Native Infrastructure

The transition from a print-oriented workflow to cloud-native is rarely smooth. And Коммерсант's engineering leadership navigated this over a decade. Around 2015, the paper began containerizing its backend services using Docker and later migrated orchestration to Kubernetes on a hybrid infrastructure that spans on-premise servers in Moscow and cloud instances in a Russia-based provider, likely Yandex. Cloud. One ex-SRE described a CI/CD pipeline built on GitLab that automatically rolls out canary deployments for the mobile API during Moscow business hours, with automated rollbacks triggered if 99th-percentile latency exceeds 300 milliseconds for the /v2/feed endpoint.

What makes this stack "мощный" (powerful) isn't the individual components but the resilience engineering baked into the data layer. Коммерсант runs a multi-master PostgreSQL cluster with logical replication to a read-only replica in a different availability zone, ensuring that even if the main editorial datacenter in Moscow goes dark, the mobile apps can still pull article content. This design is codified in their disaster recovery runbook, parts of which were shared at a 2019 HighLoad++ conference. The team also employs a custom endpoint `/health/full` that performs a deep check-verifying database connectivity, CDN purge queue depth, and push notification service availability-allowing SREs to make informed decisions during incidents.

Mobile App Engineering: Push Notifications at Breaking-News Scale

Коммерсант's native mobile apps-available on iOS, Android. And Huawei's AppGallery-handle one of the most demanding notification workloads in Russian media. When A Major commodity price swing occurs, the backend pushes breaking alerts to over 2 million devices within seconds. The engineering team documented their notification pipeline at a 2021 AppsConf Moscow talk: they use Firebase Cloud Messaging (FCM) for Android and Apple Push Notification service (APNs) for iOS, but they built an in-house priority queue in Golang that categorizes news into three tiers-critical, market-moving, and editorial-and throttles the lower tiers if the upstream rate approaches FCM's per-second limits.

The most instructive part for mobile engineers is how Коммерсант handles notification deep-linking. Instead of a simple URL scheme, they implemented Android App Links and iOS Universal Links with a fallback to a custom URL pattern that carries an encrypted payload. This payload includes an article UUID, a client-specific render token, and a timestamp that allows the app to pre-fetch content before the user even taps the notification, reducing perceived latency. Their production metrics show that pre-fetching improves the cold-start article-open time from 2. 8 seconds to under 800 milliseconds, as measured by the app's custom RUM SDK built on top of OpenTelemetry.

Real-Time Data Pipelines for Financial Market Feeds

One of Коммерсант's differentiators is its live financial data integration-exchange rates, bond yields, and indexes that update directly inside articles. This isn't a simple iframe embed. The engineering team built a stream processing pipeline using Apache Kafka to ingest data from Московская Биржа (MOEX) and other sources at sub-second latencies. A set of microservices written in Java and Kotlin then enriches the raw tick data with metadata (currency pair descriptions, percent change calculations) and publishes to a Redis Pub/Sub channel. The mobile clients subscribe via a WebSocket connection managed by a lightweight Node js gateway, which multiplexes thousands of connections with minimal overhead.

To avoid state synchronization nightmares, Коммерсант's team adopted the JSON Patch format (RFC 6902) for sending incremental Updates to the mobile UI. Instead of retransmitting the entire financial table on every tick, the server sends a small patch document like `{"op":"replace","path":"/rates/0/last","value":89. 45}`. This conserves both bandwidth and device battery life-a critical consideration for users on metered connections in Russia's regions. An engineer later told me they saw a 40% reduction in data transfer volume after switching from full-polling to a patch-based push. Which directly lowered the app's crash rate on low-memory devices.

CDN and Edge Delivery: Serving Коммерсант to a Global Audience

While Коммерсант's primary readership is inside Russia, a significant diaspora and business audience reads from Europe, Central Asia. And the Middle East. The engineering team leverages a multi-CDN strategy: static assets and article HTML are served through Qrator Labs (a Russian DDoS protection and CDN provider) for domestic traffic, with fallback to CloudFront for international requests. This geo-aware routing is implemented by a custom DNS gateway that resolves A records based on the client's source IP and real-time RTT measurements from RIPE Atlas probes.

The CDN configuration itself reflects hard-earned lessons about content freshness vs, and caching efficiencyКоммерсант's homepage is cached for just 30 seconds at the edge. But individual article pages are immutable once published-invalidated only on correction or update-so they carry one-year `Cache-Control` headers. The invalidation mechanism uses a Kafka-powered CDN purge queue that propagates a content hash to all edge nodes within 200 milliseconds, ensuring no reader sees a stale version of a corrected financial figure. This design is strikingly similar to The New York Times's "Purge by Hash" system described in their 2019 engineering blog.

Cybersecurity and DDoS Mitigation in a High-Stakes Media Environment

Russian independent media outlets frequently face DDoS attacks. And Коммерсант is no exception. The security posture, however, is built around more than volumetric mitigation, and the platform uses Qrator's always-on filtering,But internally the SRE team deploys a custom eBPF-based rate limiter that operates at the Kubernetes ingress level, blocking traffic patterns that match known bot signatures long before they reach application pods. This approach was detailed in a talk at eBPF Summit 2022 where an engineer from the media sector (not explicitly naming Коммерсант) described using XDP programs to drop packets at the NIC level.

Beyond DDoS, Коммерсант's security team focuses on content integrity and journalistic account protection. They enforce hardware security keys (FIDO2) for editorial staff accessing the CMS from outside the office network, and they run a machine learning anomaly detection system that flags unusual article modifications-for instance, a sudden change to a stock price figure by a non-superuser. This system, built on an ML model trained with the scikit-learn Isolation Forest algorithm, sends alerts to a dedicated Telegram bot used by the security operations center, a pattern I've seen increasingly adopted by eastern European media.

Observability, SRE. And the Metrics That Keep the Newsroom Online

The reliability of a news app directly impacts reader trust; when Коммерсант is down during a market panic, users switch to competitors. So the SRE team at Коммерсант built an observability stack centered on Prometheus, Grafana. And Loki, with custom exporters for their Kafka and Redis clusters. They define service-level objectives (SLOs) aggressively: 99. 95% availability for the article delivery API, measured over a 30-day rolling window, with a quarterly error