Open-Source Datadog Alternatives: HyperDX, SigNoz and Grafana Compared on Self-Hosting Cost

HyperDX, SigNoz and the Grafana stack can all replace Datadog without a licence fee, but only the first two give you logs, metrics and traces in one product out of the box. On Datadog's list page (as of 13 October 2026), infrastructure monitoring starts at $15 per host per month billed annually, and APM adds $31 per host. Self-hosting swaps that bill for servers and engineer time, and the swap is only worth it at the right volume.

Last checked: 13 October 2026. Prices and licences change, so confirm on the linked pages.

What are the best open-source Datadog alternatives for self-hosting?

The three most credible are HyperDX, SigNoz and the Grafana stack (Grafana, Loki, Tempo, Mimir). HyperDX and SigNoz both store telemetry in ClickHouse and ingest OpenTelemetry. Grafana's projects use their own storage engines and are assembled piece by piece.

Star counts are a popularity signal, not a quality measure.

What is HyperDX and who maintains it now?

HyperDX is an open-source observability platform built on ClickHouse and OpenTelemetry. It launched in this form as a Show HN on 12 November 2024 titled "Open-source Kibana alternative for logs and traces in ClickHouse", posted by one of its founders, who pitched it as a developer-friendly UI for teams already moving observability data to ClickHouse for cost or privacy reasons.

ClickHouse announced it had acquired HyperDX (page dated 13 March 2025) and says the open-source project "remains actively maintained and developed" and HyperDX Cloud keeps serving customers. The repository still sits under the hyperdxio account and describes HyperDX as "a core component of ClickStack".

The ClickHouse post claims its storage delivers "10x savings over traditional solutions". That is the company's own claim, made in an acquisition announcement, not an independent benchmark.

What does it take to self-host each option?

HyperDX has the lightest start, SigNoz needs a few more moving parts, and Grafana's logging stack has the widest range depending on how you deploy it.

• HyperDX: one all-in-one Docker image exposes the UI on 8080 and OpenTelemetry ports 4317 and 4318. The README says it deploys as part of ClickStack, which includes ClickHouse, HyperDX, an OpenTelemetry Collector and MongoDB, and recommends "at least 4GB of RAM and 2 cores for testing". That is a testing figure, not a production sizing.
• SigNoz: the Docker docs require Docker Engine 20.10+ with the Compose v2 plugin and "at least 4GB of memory allocated to Docker". The sample deployment runs ClickHouse, ClickHouse Keeper, Postgres 16, a SigNoz collector and the SigNoz app. The older install.sh script is deprecated as of v0.130.0 in favour of the Foundry tool, so older tutorials may not match.
• Grafana stack: Loki does not index log contents, only labels, and stores compressed chunks in object storage such as S3 or GCS. It offers three deployment configurations, and Simple Scalable Deployment mode is deprecated and set for removal in Loki 4.0. Mimir and Tempo are separate services to run.

For scale, Loki's sizing guide only documents microservices mode. Its smallest tier, for under 100TB of logs per month, lists component requests that add up to roughly 38 CPU and 59 Gi of memory when its table is summed. Most teams will not need that tier, but it shows that "free" Loki at scale is a real cluster.

How much does Datadog charge, and where does the bill grow?

Datadog bills in several units at once, which is why the host price understates the total. As of 13 October 2026, the list page shows these annual-billing rates:

The main pricing page says Pro includes 100 custom metrics per host and Enterprise 200. Ingestion and indexing are separate charges, so a team that sends 100 GB of logs pays $10 to ingest them, plus whatever it indexes.

How much do the vendors' own clouds cost?

If your goal is lower cost rather than owning servers, the open-source vendors sell hosted versions. These are company-published prices as of 13 October 2026.

Sources: SigNoz pricing and Grafana pricing. SigNoz's page shows a free trial and the Community Edition, but no free cloud tier. Grafana's free tier is limited to 3 active users.

What would a small team actually pay?

A rough illustration with an assumed workload: 10 hosts with infrastructure and APM, plus 100 GB of logs per month. Datadog's annual list rates give (15 + 31) x 10 = $460 per month for hosts and APM, plus $10 for log ingestion, before indexed logs, custom metrics and indexed spans. The workload is our assumption; your bill will differ.

On the same 100 GB of logs, SigNoz Cloud's $0.3/GB comes to $30, which is under the $49 monthly minimum, so $49 would be the floor. Grafana Cloud's starting log rates sum to $0.55/GB (process, write and retain), though the $19 Pro fee includes 50 GB, so the exact figure depends on how those allowances apply.

Self-hosting has no such per-GB meter, but a server running ClickHouse, backups and an on-call engineer have costs the vendors' pages do not publish. None of the three projects publishes a first-party figure for self-hosted total cost, so treat any single number you see online as an estimate.

Where does Datadog still win?

Datadog sells breadth and support as one contract. Its Infrastructure plans, APM, log management and security products share one agent and one bill, and an operations team gets a vendor to call. The open-source options ask you to run ClickHouse (or Loki, Mimir and Tempo), upgrade it, back it up and watch it.

Two facts to weigh. SigNoz's README says a native AI teammate is available only in SigNoz Cloud, and its pricing page lists anomaly detection, SAML and SOC2/HIPAA compliance under Teams. So the free Community Edition is not feature-identical to the paid tier. If you also self-host AI agents and want a comparison of that tradeoff, see our look at open-source agents you can self-host, and for how tool pricing shifts by tier, Cursor vs Windsurf vs Claude Code.

What's still unclear?

• Real self-hosting cost. None of the three projects publishes a total-cost figure. HyperDX's "4GB and 2 cores" is for testing, and Loki's sizing guide covers large installations only.
• Licence nuances. AGPL-3.0-only applies to Grafana, Loki and Mimir, but each repo points to a LICENSING.md with Apache-2.0 exceptions. SigNoz's enterprise directories are covered by a separate ee/LICENSE file.
• HyperDX's future. ClickHouse says the open-source project stays maintained. In the original Show HN thread, one commenter worried that contributor agreements could allow relicensing, and a founder replied that the project would stay MIT. The current repo is MIT, but a future change cannot be ruled out from public statements.
• Datadog price variants. Datadog's main page and list page show different on-demand figures for some APM lines, so this post uses the list page and annual rates.
• No testing. We did not run these tools. Everything here comes from vendor pages, repositories and documentation.

Which one should you pick?

• Already running ClickHouse, want a fast UI: HyperDX. MIT licence, one container to try, and session replay included.
• Want the closest all-in-one Datadog replacement, self-hosted or cloud: SigNoz. Check which features you need are in the Community Edition.
• Already use Grafana dashboards and Prometheus-style metrics: the Grafana stack, with Grafana Cloud's free tier as a low-risk trial.
• Small team without an infra engineer: price the vendor clouds first. SigNoz's $49 floor and Grafana's free tier may beat the cost of a server and on-call time.