Parseable pitches open-source observability datalake it says handles 100M time-series per minute

What happened

Parseable reached the Hacker News front page on October 6 with a Show HN post presenting an open-source observability datalake that, according to the post, handles 100 million time-series per minute. The company describes the project as a single binary that unifies logs, metrics and distributed traces, stores them in an open columnar format on object storage, and keeps full-fidelity data queryable. As with any self-reported benchmark, the throughput figure is the vendor's own claim and has not been independently verified.

Architecture: columnar, stateless, object-store native

According to Parseable's site, the system is built to address high cardinality — the explosion of unique label combinations that traditionally degrades time-series performance — through its columnar layout, and it advertises above-average compression as a result. The specific compression percentage was truncated in the material reviewed, so only the direction of that claim can be reported.

Deployment is pitched as deliberately simple: run the binary on a public or private cloud, point it at an object store, and configure your telemetry agents. The design is stateless, with the object store acting as the system of record, and the platform is OpenTelemetry-native and composable. The feature set bundled into the single binary includes alerts, dashboards, log and metric views, distributed traces, service maps, an error page, AI-driven analysis and root-cause analysis. Data can be queried with SQL or natural language, and forecasting is listed among the platform's capabilities.

Deployment tiers and the AI toolbox

Parseable offers three deployment options: self-hosted open source, a managed cloud, and an enterprise tier with bring-your-own-cloud. The enterprise edition adds PromQL support, distributed queries, an AI-native interface, advanced access control and governance, and priority support for critical deployments.

The company also ships a set of companion tools. An MCP server connects AI agents to live telemetry through what Parseable calls safe, structured tools. A Slack bot brings investigation and operational answers directly into chat. PAI automatically instruments Kubernetes workloads to collect signals. PB CLI handles querying from the terminal, and OTex is described as an AI-powered observability engineer that plans, instruments and reviews telemetry. Beyond these in-house tools, the site lists integrations across telemetry agents, data sources, visualization tools, authentication providers and LLMs.

Data ownership as a pitch

A recurring theme in the company's messaging is control. Parseable emphasizes that telemetry lives in open formats on object storage that the customer owns, paired with enterprise-grade controls aimed at security-conscious teams. That positioning targets one of the most common complaints about commercial observability platforms: data held in proprietary formats, with migration and egress costs attached to leaving.

Why it matters

Observability costs scale poorly at high cardinality, and systems that store telemetry in open columnar formats on inexpensive object storage attack that economics directly. If the throughput claim holds up under third-party testing, the combination of a single binary, an OTel-native ingestion path and full-fidelity data on object storage would position Parseable as a credible open-source alternative to assembled stacks of separate logging, metrics and tracing tools. The AI layer — MCP integration, natural-language querying and the OTex agent — also reflects where the broader observability market is heading: assistants that investigate incidents alongside engineers. For now, the headline number remains a claim on a landing page rather than a benchmarked result, so teams evaluating it should treat the performance figures as a starting point for their own load testing rather than settled fact.