---
type: "post"
id: "ytsaurus-flow-open-source-streaming-platform-analysis-f4cc7d"
slug: "ytsaurus-flow-open-source-streaming-platform-analysis-f4cc7d"
title: "Ytsaurus flow open source streaming platform"
description: "YTsaurus Flow’s release makes a production stream-processing framework available beyond Yandex."
retrieval_nugget: "YTsaurus Flow’s release makes a production stream-processing framework available beyond Yandex."
published_at: "2026-10-06"
updated_at: "2026-10-06"
record_date: "2026-10-06"
date_kind: "published_at"
topics: []
source_urls: ["https://habr.com/ru/companies/yandex/articles/1087470"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:ytsaurus-flow-open-source-streaming-platform-analysis-f4cc7d"
status: "published"
visuals: []
editorial_provenance: {"schema_version":"newruntime-editorial-copy-v1","content_status":"source_grounded_final","final_copy_sha256":"sha256:782ad0b1a3bf45acbe80b338c73cdff673d5238bbf94a4402776e6426b0f2d7d","reviewed_at":"2026-10-06T20:10:41.795Z","source_evidence_count":1,"verified_claim_count":3,"site_analysis_schema_version":"newruntime-site-analysis-v1","site_object_kind":"field_note","observed_fact_count":3,"implication_count":1,"watch_condition_count":1,"related_record_count":0}
analysis: {"schema_version":"newruntime-site-analysis-v1","object_kind":"field_note","thesis":"The notable change is access to a stateful streaming framework built around YTsaurus’s shared transaction system, rather than a general claim that streaming itself is new.","observed_facts":[{"text":"The source says YTsaurus Flow is used by Yandex Advertising, Market and anti-fraud systems.","source_urls":["https://habr.com/ru/companies/yandex/articles/1087470"]},{"text":"Yandex reports that Advertising processes more than 100 GB of data per second through Flow.","source_urls":["https://habr.com/ru/companies/yandex/articles/1087470"]},{"text":"Yandex says preparation for retraining some models fell from more than ten hours to about two.","source_urls":["https://habr.com/ru/companies/yandex/articles/1087470"]}],"mechanism":"The article says Flow atomically changes pipeline state and queue metadata such as offsets in a shared transaction, while relying on YTsaurus storage and queues for recovery. This design aims to prevent loss or duplicate accounting across failures.","why_now":"An Apache 2.0 release allows external teams to inspect and test a system Yandex says is already operating in production.","implications":["For teams already using YTsaurus, the framework may offer a way to build low-latency stateful pipelines without assembling separate mechanisms for state, offsets and recovery."],"evidence_boundary":"Performance and adoption figures come from Yandex’s own account; the supplied source does not establish independent verification or comparable performance on other platforms.","watch_conditions":["Whether external users can reproduce the reported reliability and latency outside Yandex’s YTsaurus environment will determine how broadly the release matters."],"related_records":[],"new_branch_reason":"No prior New Runtime record is supplied, so this opens a distinct analytical record focused on the public release of YTsaurus Flow and the testability of its production claims."}
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/ytsaurus-flow-open-source-streaming-platform-analysis-f4cc7d/","markdown":"https://newruntime.com/posts/ytsaurus-flow-open-source-streaming-platform-analysis-f4cc7d.md","json":"https://newruntime.com/posts/ytsaurus-flow-open-source-streaming-platform-analysis-f4cc7d.json"}
---

# Ytsaurus flow open source streaming platform

## Retrieval answer

YTsaurus Flow’s release makes a production stream-processing framework available beyond Yandex.

YTsaurus Flow’s release makes a production stream-processing framework available beyond Yandex. Thesis: The notable change is access to a stateful streaming framework built around YTsaurus’s shared transaction system, rather than a general claim that streaming itself is new. The company says it has opened the code under Apache 2.0 and describes exactly-once processing as the default.

Observed facts: the source says Flow is used by Yandex Advertising, Market and anti-fraud systems. It reports that Advertising processes more than 100 GB of data per second through Flow, and that data preparation for retraining some models went from more than ten hours to about two. Mechanism: The article says Flow atomically changes pipeline state and queue metadata such as offsets in a shared transaction, while relying on YTsaurus storage and queues for recovery. This design aims to prevent loss or duplicate accounting across failures. Why now: An Apache 2.0 release allows external teams to inspect and test a system Yandex says is already operating in production.

Implications: For teams already using YTsaurus, the framework may offer a way to build low-latency stateful pipelines without assembling separate mechanisms for state, offsets and recovery. Evidence boundary: Performance and adoption figures come from Yandex’s own account; the supplied source does not establish independent verification or comparable performance on other platforms. Watch condition: Whether external users can reproduce the reported reliability and latency outside Yandex’s YTsaurus environment will determine how broadly the release matters.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)