SigNoz vs Datadog in 2026: OTel, ClickHouse & APM TCO
Engineering and 36-month TCO teardown of SigNoz vs Datadog and Grafana Cloud in 2026: OpenTelemetry Collector tail-sampling, unified ClickHouse telemetry, p99 host high-water-mark billing, PG17 FORCE RLS alert planes, and 150-host math.

On this page

Somewhere past fifty Kubernetes nodes, most platform teams start asking the SigNoz vs Datadog question, usually with Grafana Cloud pricing, OpenTelemetry tail-sampling and ClickHouse BYOC observability close behind. Datadog is easy on a dozen nodes. The bill is what changes. Stacked per-host SKUs, $5.00 per 100 custom metric series, split log ingestion and indexing fees, and p99 high-water-mark billing on autoscaling bursts compound, and by year three the invoice can rival the compute it watches.
Keep stack layers distinct: ClickHouse vs TimescaleDB benchmarks customer-facing OLAP engines, and Langfuse vs LangSmith evaluates gen_ai.* prompt and trajectory evals for agentic AI in production. This guide covers the layer underneath: infrastructure metrics, distributed APM and structured logs, the telemetry that keeps microservices, queues and databases inside their SLAs.
At Idealogic, our custom software development, SaaS development, software development consulting, and dedicated development team squads architect telemetry backbones across fintech, healthtech, and logistics and supply chain. This 2026 teardown compares SigNoz, Datadog, Grafana Cloud and an owned OpenTelemetry + ClickHouse stack, walks through an OTel Collector Gateway (tail_sampling plus PCI DSS v4.0.1 PII scrubbing) with a PostgreSQL 17 or 18 FORCE RLS SLO plane, and models 36-month TCO at 150 Kubernetes hosts, 6,000 GB/mo of logs, 3,000 GB/mo of traces and 50,000 custom metric series.
What are SigNoz and Datadog? SigNoz is the open-source Datadog alternative: an OpenTelemetry-native observability tool with an MIT-licensed core (enterprise modules sit under a separate licence) that stores traces, logs and metrics in ClickHouse. Datadog is a proprietary SaaS observability suite billed per host, per GB and per custom metric. Grafana Cloud sits in between as the managed version of the AGPLv3 Grafana, Loki, Tempo and Mimir stack.
The short version
- Datadog leads on turnkey breadth and loses on compound SKU and
p99host math: Datadog pricing stacks Infrastructure Pro ($15/host/mo) or Enterprise ($23/host/mo) and APM Enterprise ($40/host/mo), all billed annually (list prices checked November 2026), plus Custom Metrics ($5.00 / 100series) and split Log Ingestion ($0.10/GB) + 15-day Indexing ($1.70/Mevents), amplified by hourlyp99high-water-mark host metering. - SigNoz, the open-source alternative to Datadog, wins on OpenTelemetry portability, unified ClickHouse queries and no host or seat fees. It's built on OpenTelemetry (MIT core plus Enterprise modules) and ClickHouse
MergeTree, and SigNoz pricing charges zero per-host and zero per-seat fees ($49/moTeams base credit,$0.30/GBlogs and traces,$0.10/Mmetric samples, plus Enterprise/BYOC tiers). - Grafana Cloud is strongest for Prometheus mixins across three split backends. It pairs industry-standard dashboards with Mimir, Loki, and Tempo, requiring SREs to juggle
PromQL,LogQL, andTraceQLplus per-user Enterprise IRM/SAML packages. - At
150K8s hosts (9 TB/mologs+traces,50kseries,45engineers), Owned OTel + BYOC wins TCO: Over 36 months, Owned OTel Gateway + Self-Hosted SigNoz / ClickHouse EKS ($144,900CEE) saves$455,400vs Datadog Enterprise ($600,300),$198,000vs Grafana Cloud Pro ($342,900), and$63,000vs SigNoz Cloud Enterprise ($207,900).
2026 observability internals & billing mechanics: SigNoz vs Datadog vs Grafana Cloud
Evaluating signoz vs datadog (or datadog vs signoz) comes down to four structural boundaries: agent instrumentation lock-in, host high-water-mark billing, storage correlation, and metric cardinality math.
1. Datadog: Proprietary agent polish vs five compounding billing meters
With 1,000+ built-in integrations, eBPF Universal Service Monitoring, Continuous Profiler and Cloud SIEM, Datadog covers more ground out of the box than either rival. The catch shows up in observability vs monitoring at Kubernetes scale, where five billing meters stack on top of each other:
- Stacked Host SKUs (
$15/$23Infra +$40APM Ent.): Pairing Infrastructure Enterprise ($23/host/mo,200custom metrics,10containers; vs Pro at$15/host/mo,100custom metrics,5containers) with APM Enterprise ($40/host/mo) sets your baseline at$63/host/mobefore logs or metric overages. - Container Pools & Hourly
p99High-Water-Mark Taxes: Under Datadog container billing, extra containers above your host allotment (5on Pro,10on Ent.) add per-container fees, while host counts are metered hourly at a monthlyp99high-water mark: Datadog excludes the highest 1% of hourly readings (about 7 hours in a 720-hour month) and bills the maximum of the remaining 99%. If Karpenter bursts your 150-node cluster to about 177 nodes for two hours each weekday (40 hours/mo) during checkout or data pipeline peaks, that burst outlasts the 7-hour allowance, so all 27 extra nodes are billed for the month across Infra and APM. - Split Log Ingestion + Retention Indexing: Datadog charges
$0.10/GBto ingest logs, plus$1.70 per million eventsindexed at15-dayretention, forcing fragile exclusion filters that drop debug context right before outages. - Custom Metric Cardinality (
$5.00 per 100series): Every100custom series above your host pool (100/host Pro,200/host Ent.) costs$5.00/mo. Addingtenant_idorendpointlabels to histograms multiplies series overnight. - Agent and DogStatsD gravity. Datadog accepts OTLP directly and ships its own collector distribution, but live processes, USM and profiler flamegraphs work best with
dd-traceand proprietary monitors, and that's where vendor lock-in builds up.
2. SigNoz: the open-source Datadog alternative built on OpenTelemetry and ClickHouse
SigNoz (SigNoz/signoz) is engineered natively around OpenTelemetry semantic conventions and ClickHouse columnar storage:
- 100% Vendor-Neutral OTel: Services instrument once with upstream OpenTelemetry SDKs and Kubernetes collectors, exporting
OTLP/gRPC(:4317) orOTLP/HTTP(:4318) into the OpenTelemetry Collector. - Unified ClickHouse Correlation: Instead of running Elasticsearch for logs, Cassandra for traces and Prometheus TSDB for metrics, SigNoz writes all three signals into ClickHouse
MergeTreetables.ZSTDcompression often reaches 5x to 10x or more on repetitive logs, though you should measure it on your own data. Materialized hot columns (service.name,http.response.status_codeor the legacyhttp.status_codefrom older SDKs,tenant.id) let you jump from a slowp99Flamegraph span to its container logs in milliseconds. - Usage-Only Pricing (
$0Host,$0Seat,$0Series Tax): SigNoz Cloud Teams ($49/mobase credit) bills$0.30/GBfor logs and traces (15-day search included, no separate indexing fee) and$0.10/Mmetric samples regardless of series cardinality, with 1-month metrics retention and no host or seat fees. Enterprise starts at $4,000/mo with usage included and adds SAML, Dedicated and BYOC options.
3. Grafana Cloud Pro / Advanced (LGTM) trade-offs
Grafana Cloud (Loki, Grafana, Tempo, Mimir) shines for teams invested in Prometheus mixins, trading single-engine simplicity for three separate query engines (PromQL, label-only LogQL chunk scans, and TraceQL), active-series metering and per-active-user Enterprise SAML/IRM pricing. Grafana bills on 95th-percentile active series and monthly active users, so the Grafana figures below are our assumed list-price model; check Grafana Cloud billing against your own usage.
Production blueprint: OpenTelemetry tail-sampling gateway, PCI DSS scrubbing and a multi-tenant SLO plane
Teams that move multi-tenant SaaS architecture and how to build a SaaS platforms off Datadog without a gateway tend to hit two failures. The first is sampling: head sampling misses rare downstream errors, while keeping 100% of traces floods storage, and unredacted card numbers or JWTs in spans put you in breach of PCI DSS v4.0.1. The second is noisy-neighbor alert storms that leak one tenant's metadata to another.
We fix the first with a two-layer OpenTelemetry Collector topology. Lightweight node collectors tag spans with Kubernetes metadata and pass them to a stateful gateway tier, which redacts PII with the transform processor and then applies the tail sampling processor. In our model this cuts network egress and ClickHouse trace storage by about 65% and still keeps every error, slow and VIP trace the policies match. Your ratio will depend on error rate and traffic mix. The sampling policy waits 10 seconds for a trace to finish, then keeps every trace with an error, every trace slower than 500 ms, every trace on the VIP payout, checkout and settlement routes, and a 5% baseline of everything else.
The tail sampling processor is still beta for traces and holds spans in memory for the full decision window. Size gateway replicas for that wait and test pod restarts before you rely on it.
For the second failure we put Google SRE multi-window SLO burn-rate alerting behind a PostgreSQL 17 or 18 control plane with Row-Level Security, so each tenant's SLO policies and incidents stay isolated across multi-tenant SaaS boundaries. The fast-burn rule is 14.4x over both one hour and five minutes, which spends 2% of a 30-day error budget in an hour. These are the design decisions that carry the blueprint:
Auditors for SOC 2 Type II and PCI DSS v4.0.1 ask about exactly this kind of tenant-isolation control. The design helps, but it won't pass an audit for you.
SigNoz vs Datadog TCO at 150 Kubernetes hosts: the 36-month model
The short answer: at 150 Kubernetes hosts and 9 TB/mo of logs and traces, our 36-month model puts Datadog Enterprise at $600,300, SigNoz Cloud at $207,900 and an owned OTel plus ClickHouse stack at $144,900.
Model a regulated B2B SaaS cluster over 36 months: 150 Kubernetes hosts (<= 10 containers/host, weekday autoscaling bursts), 6,000 GB/mo logs (1,500M indexed events/mo at 15-day retention, assuming roughly 4 KB per event), 3,000 GB/mo trace spans, 50,000 active custom metric series (2,500M samples/mo), 45 engineers, Okta SAML SSO, and SOC 2 Type II + PCI DSS v4.0.1.
Labor compares Idealogic's $75/hr Senior CEE vs $165/hr US agency rates (2.2x). The $165 figure is a typical US agency rate in our experience, not an official statistic; for labour-cost context see the Eurostat hourly labour cost table (lc_lci_lev) and O*NET Software Developers (15-1252.00). Infrastructure uses us-east-1 list prices for Amazon EKS ($0.10/hr), AWS EBS gp3 ($0.08/GB-mo), and Amazon S3 ($0.023 Standard, $0.0125/GB-mo Standard-IA) pricing:
- Option A: Datadog Enterprise (
150Hosts + APM Ent. +15-DayIndexed Logs +p99Burst):- Setup & FinOps (
120hsetup =$9,000CEE /$19,800US +15h/mo x 36 = 540htag/exclusion governance =$40,500CEE /$89,100US ->660h=$49,500CEE /$108,900US). - 36-mo Datadog SaaS (
$15,300/mo x 36 = $550,800): Infra Ent. (150 x $23 = $3,450/mo,30kpooled custom series) + APM Ent. (150 x $40 = $6,000/mo, annual list price) + Custom Metrics20,000overage (200 x $5 = $1,000/mo) + Log Ingestion (6,000 x $0.10 = $600/mo) + 15-day Indexed Logs (1,500M x $1.70/M = $2,550/mo) =$13,600/mobaseline +p99autoscaling burst (27extra nodes x$63= about$1,700/mo) =$15,300/mo->$600,300CEE ($659,700US).
- Setup & FinOps (
- Option B: Grafana Cloud Pro / Advanced (
Loki+Tempo+Mimir+45-User Enterprise IRM/SAML):- Setup & PromQL/LogQL/Tempo Ops (
180hsetup =$13,500CEE /$29,700US +20h/mo x 36 = 720hops =$54,000CEE /$118,800US ->900h=$67,500CEE /$148,500US). - 36-mo Grafana Cloud (
$7,650/mo x 36 = $275,400, assumed list-price model):$3,000/mologs +$1,500/motraces +$400/mo50kPrometheus series +$2,750/moEnterprise IRM/SAML for45monthly active users (our assumption) ->$342,900CEE ($423,900US).
- Setup & PromQL/LogQL/Tempo Ops (
- Option C: SigNoz Cloud Enterprise (
$0Host Tax,$0Seat Tax, Unified Managed ClickHouse):- Setup & OTel Collector Ops (
120hsetup =$9,000CEE /$19,800US +15h/mo x 36 = 540hops =$40,500CEE /$89,100US ->660h=$49,500CEE /$108,900US). - 36-mo SigNoz Cloud (
$4,400/mo x 36 = $158,400): Logs (6,000 GB x $0.30 = $1,800/mo) + Traces (3,000 GB x $0.30 = $900/mo) + Metrics (2,500M x $0.10/M = $250/mo) + Enterprise plan uplift ($1,450/mo, our assumption: SigNoz lists Enterprise from$4,000/mowith usage included, so we model$4,400/moall-in;$0host fees,$0seat fees) ->$207,900CEE ($267,300US).
- Setup & OTel Collector Ops (
- Option D: Owned OpenTelemetry Gateway + Self-Hosted SigNoz / ClickHouse EKS BYOC (
MITcore):- Turnkey Build (
600h@$75/hrCEE =$45,000;$165/hrUS =$99,000): HA OTeltail_sampling+ PII redaction gateway (65%trace reduction assumed), self-hosted SigNoz (MITcore) + Altinity ClickHouse Operator on EKS (NVMehot +S3-IAcold), andPG17 FORCE RLSSLO plane. - 36-mo AWS EKS + ClickHouse + S3-IA (
$1,650/mo x 36 = $59,400): Right-sized after65%trace tail-sampling reduction (1,050 GB/mopersisted spans +ZSTD-compressed logs (about10xassumed) onr8gdNVMe/gp3+$0.0125/GB-moS3-IA+ RDSPG17) +15h/mo(540h) SRE ($40,500CEE /$89,100US) ->$144,900CEE ($247,500US).
- Turnkey Build (
Assumptions behind these numbers: list prices checked in November 2026 with annual billing where the vendor offers it, Datadog log events of roughly 4 KB, a burst of 27 extra nodes for 40 hours a month, a 65% tail-sampling reduction and about 10x ZSTD compression on logs. The Grafana Cloud and SigNoz Enterprise lines are modelled, not quoted. Swap in your own invoice before you act on any of it.
Sensitivity check: Sub-30 host SaaS cutoff & CapEx payback
- Under
30hosts (<1.5 TB/mologs+traces): Skip self-hosting; run SigNoz Cloud Teams ($49/mobase,$0.30/GB,$0host/seat fees) behind an OTel PII scrubber (~$350 to $600/mo). Adding65%OTeltail_samplingin front of SigNoz Cloud Ent. also trims Option C by another$585/mo($21,060over 36 months). - Payback at
150hosts ($75/hrCEE): Option D OpEx is$2,775/mo($1,650AWS +$1,125SRE) vs$16,425/mo(Datadog),$9,150/mo(Grafana), and$5,525/mo(SigNoz Cloud). Extra$36,000CapEx pays back in2.64 monthsvs Datadog ($36,000 / $13,650/mo) and13.09 monthsvs SigNoz Cloud ($36,000 / $2,750/mo).
SigNoz vs Datadog migration costs: senior CEE vs US rates
Our custom software development, SaaS development, and dedicated development team squads deliver Datadog-to-OTel migrations at $75/hr senior CEE vs $165/hr US agency rates (2.2x):
Inside the worked 600-hour ($45,000 CEE vs $99,000 US) owned OpenTelemetry stack
- 1. OTel Tail-Sampling (
25%/150h/$11,250CEE vs$24,750US): Node DaemonSets,traceID-routed stateful OTel Gateway,tail_sampling(100%error/>500ms/VIP,5%baseline), andPCI DSS v4.0.1PII scrubbing. - 2. ClickHouse & S3-IA (
25%/150h/$11,250CEE vs$24,750US): Altinity ClickHouse Operator on EKSr8gdNVMe nodes,15-dayhot to90-dayS3-IATTL tiering,ZSTDcodecs, and S3 Gateway Endpoints. - 3.
PG17Alert Plane (20%/120h/$9,000CEE vs$19,800US):PG17 FORCE RLSmulti-tenant SLO burn-rate tables (14.4xmulti-window), cooldown deduping, and encrypted PagerDuty/Slack routing. - 4. RED Maps & Hooks (
20%/120h/$9,000CEE vs$19,800US):dd-traceto OpenTelemetry migration, materializedREDservice maps, Okta SAML SSO, and GitOps dashboards. - 5. Replay & Cutover (
10%/60h/$4,500CEE vs$9,900US): 14-day dual-export shadow parity against Datadog,k6backpressure chaos QA, and clean Datadog agent teardown.
Teams rarely overspend on observability because they log too many real incidents; they overspend because Datadog bills high-water-mark host taxes on autoscaling bursts while indexing billions of healthy spans. Putting an OpenTelemetry tail-sampling and PII-redaction gateway in front of ClickHouse keeps one hundred percent of broken traces while cutting three-year spend by seventy-five percent.
Shipped Idealogic proof and a 5-step SigNoz vs Datadog decision checklist
Two shipped projects where we built this kind of telemetry:
- Exchange & banking SLO telemetry (Bitholder & Fese): Matching-engine and ledger trace correlation, with PAN redaction designed for PCI DSS v4.0.1 scope.
- Logistics SLA observability (HaulBreeze): Carrier webhook and cold-chain burn-rate alerting across logistics and supply chain fleets.
When is Datadog still the right call? If you run dozens of legacy integrations across clouds, use Cloud SIEM, RUM and eBPF Universal Service Monitoring every day, and the bill is a small slice of your cloud spend, the migration labor may never pay back. Paying for breadth is a fair trade when nobody on your team wants to run collectors.
For everyone else, five checks settle SigNoz vs Datadog vs Grafana Cloud in 2026:
- Under
30hosts (<1.5 TB/mologs+traces)? Run SigNoz Cloud Teams ($0.30/GB,$0host/seat fees) with an OTel PII scrubber. - Legacy cloud sprawl needing turnkey Cloud SIEM & eBPF USM? Keep Datadog behind an upstream OTel Gateway trimming noisy logs and metrics before ingest.
- Deep Prometheus/Grafana culture
<5 TB/mo? Use Grafana Cloud Pro with Alloy metric relabel drops. 50 to 150hosts wanting managed ClickHouse APM without K8s storage ops? Pick SigNoz Cloud Enterprise ($207.9k3-yr TCO vs$600.3kDatadog).100+autoscaling K8s hosts,5 TB+/motelemetry, or strict VPC residency? Deploy Owned OTel Gateway +MIT-core SigNoz / ClickHouse on EKS (NVMe+S3-IA+PG17SLO Plane) ($144.9k3-yr TCO). Contact our team for a 48-hour Datadog-to-OTel TCO audit.

CTO and Co-founder at Idealogic. Directing engineering, systems architecture, security, and full-stack delivery across web, mobile, and blockchain.
Frequently asked questions
Choose SigNoz when your microservices standardize on vendor-neutral OpenTelemetry instrumentation, your Kubernetes autoscaling triggers expensive p99 host high-water-mark surcharges on Datadog, or your security team wants an MIT-licensed open-source core and BYOC ClickHouse data residency with no per-host or per-seat fees. Choose Datadog when your organization runs hundreds of legacy heterogeneous cloud integrations, relies on turnkey eBPF Universal Service Monitoring, and prioritizes out-of-the-box Cloud SIEM and RUM breadth over telemetry margin.
Datadog bills Infrastructure Pro ($15 per host per month), Infrastructure Enterprise ($23) and APM Enterprise ($40) at annual list prices, and it counts active hosts every hour. It drops the highest 1% of hourly readings, roughly seven hours in a 720-hour month, and bills the maximum of what is left. So if Horizontal Pod Autoscalers or Karpenter keep extra nodes running for more than about seven hours a month, during traffic peaks or batch jobs, those nodes are billed for the whole month on every attached host SKU.
Datadog allocates 100 custom metrics per Infrastructure Pro host and 200 per Enterprise host, charging $5.00 per 100 additional custom metric time series per month. Adding a single high-cardinality Kubernetes label such as pod_uid, tenant_id, or endpoint across microservice histograms easily spawns tens of thousands of distinct combinations. SigNoz eliminates series-cardinality surcharges on its cloud and self-hosted plans, metering $0.10 per million ingested metric samples regardless of label combinations.
Grafana Cloud runs three separate specialized engines underneath one UI: Mimir for Prometheus metrics, Loki for label-indexed object logs, and Tempo for trace blocks, requiring engineers to pivot across PromQL, LogQL, and TraceQL. SigNoz stores traces, structured logs, and metrics inside a single unified ClickHouse columnar cluster with materialized high-cardinality columns, allowing engineers to correlate a slow trace span directly to raw container logs and host metrics in sub-second queries.
Head-based sampling decides whether to keep or drop a distributed trace at the ingress root span before downstream services execute, inevitably discarding rare downstream database timeouts or payment failures unless you keep high percentages of traffic. Running an OpenTelemetry Collector Gateway tier with the tail_sampling processor buffers spans by trace_id in memory for several seconds and retains one hundred percent of error traces, one hundred percent of slow traces above 500 milliseconds, and one hundred percent of critical checkout routes while sampling healthy baseline requests at five percent.
Never forward raw HTTP headers, request bodies, or SQL bind parameters directly from application pods to external SaaS observability endpoints. Deploy an OpenTelemetry Collector processor pipeline inside your VPC using the transform processor (OTTL) to redact primary account numbers, Bearer JWTs, emails, and tax identifiers in flight before exporting sanitized telemetry over mTLS to ClickHouse or SigNoz.
Below roughly 1,500 gigabytes of monthly logs and traces, SigNoz Cloud Teams (starting at $49 per month base credit plus $0.30 per gigabyte for logs and traces and $0.10 per million metric samples) is far more economical than operating your own storage cluster. Once sustained telemetry crosses five to ten terabytes per month under strict PCI DSS v4.0.1 or HIPAA VPC residency rules, running an owned OpenTelemetry tail-sampling gateway into ClickHouse on EKS with NVMe hot disks and S3 Standard-IA cold tiers cuts three-year TCO by thirty percent versus SigNoz Cloud and over seventy-five percent versus Datadog.
Partnering with Idealogic's senior Central and Eastern European platform engineers at $75 per hour (versus typical US agency rates of about $165 per hour), a two to three week OpenTelemetry Collector Audit and Tail-Sampling Sprint costs $11,250 to $18,000, a managed SigNoz Cloud Enterprise migration with Terraform alert parity runs $18,000 to $27,000, and a turnkey 600-hour owned OpenTelemetry, ClickHouse EKS, S3-IA, and PostgreSQL 17 or 18 SLO control-plane build costs $45,000 ($99,000 at US rates), saving $455,400 against Datadog Enterprise over thirty-six months.
More from the journal

PlanetScale vs Neon in 2026: Metal NVMe, CoW & PG17 TCO
Engineering and 36-month TCO teardown of PlanetScale for Postgres vs Neon (Databricks) and AWS Aurora I/O-Optimized in 2026: local NVMe Metal unlimited IOPS vs disaggregated Pageservers, CoW branching, Neki sharding preview, FORCE RLS on PG17, and 1.2 TB math.

Modal vs Baseten in 2026: Serverless GPU, vLLM & H100 TCO
Engineering and 36-month TCO teardown of Modal vs Baseten, Replicate (Cloudflare), and Together AI in 2026: per-second Python H100/B200 containers and GPU memory snapshots vs Truss TensorRT-LLM, KV-cache prefix routing, and 3,200 H100-hr/mo math.

GrowthBook vs LaunchDarkly in 2026: Flags, CUPED & TCO
Engineering and 36-month TCO benchmark of GrowthBook vs LaunchDarkly in 2026 alongside Statsig (OpenAI / Amplitude) and CNCF OpenFeature: warehouse-native CUPED SQL, client MAU vs $40/seat billing, PG17 SSE providers, and 500k-MAU math.