Inside SigNoz: How an Indian Open-Source Engine Challenges Datadog
An architectural, commercial and operational teardown of an OpenTelemetry-native observability platform—with current published pricing, corrected product claims, a reproducible 100 GB/day model and explicit source notes.

High-density telemetry dashboards and open-source query execution. Photo: Markus Spiske via Unsplash (Free License).
Key Analytical Takeaways
- •Architecture: ClickHouse can be efficient for analytical telemetry queries, but the cited SigNoz-versus-ELK storage result is a vendor-run 2023 benchmark—not a current independent comparison.
- •Portability: OpenTelemetry reduces dependence on proprietary agents and SDKs. It does not automatically migrate dashboards, alert logic, vendor-specific attributes or workflows.
- •Economics: In the defined 100 GB/day model, SigNoz Cloud is about 72% cheaper than a FinOps-tuned Datadog scenario and 83% cheaper than a fully indexed scenario. Self-hosting depends heavily on infrastructure and SRE labor.
- •Governance: Hosting telemetry in India can support data-custody objectives, but it is not standalone DPDP compliance.
01 The observability tax—and the Indian contender
Modern software teams emit logs, metrics and traces from services, containers, databases and infrastructure. The challenge is retaining enough detail to diagnose incidents without allowing ingestion, indexing, retention and host-based charges to become unpredictable.
Datadog's commercial model illustrates why a single "price per GB" comparison is misleading. Log ingestion is billed by volume, indexed events by event count and retention tier, while infrastructure, APM, custom metrics and other products add separate meters. Effective cost depends on event size, indexing share, host count, retention and enabled products. [1][2]
SigNoz offers an open-source Community Edition that teams can self-host and a managed cloud service with usage-based charges for logs, traces and metric samples. Its architecture is built around OpenTelemetry and ClickHouse. SigNoz announced $6.5 million in funding led by SignalFire, including a previously unannounced $1.1 million round involving Y Combinator and angels. [3]
Funding and GitHub adoption are not evidence of production suitability. Buyers should evaluate ingestion reliability, query latency, retention behavior, alert parity, operational burden, support and migration costs.
The right question is not "Is SigNoz cheaper?" It is: "For this telemetry mix and operating model, what is the total cost of reliable observability—and what capabilities must we operate ourselves?"
| Statement | Evidence status | Editorial treatment |
|---|---|---|
| SigNoz raised $6.5M | Company announcement | Attributed to SigNoz; not independent funding verification. |
| Up to 80% savings | Vendor marketing claim | Tested against a defined scenario, not generalized. |
| 2023 benchmark | Vendor-run benchmark | Historical evidence only; limitations disclosed. |
| 100 GB/day comparison | Our modeled estimate | Reproducible arithmetic using stated assumptions. |
02 Under the hood: ClickHouse versus inverted indexes
Traditional Elasticsearch deployments use Lucene-based inverted indexes, designed for flexible search across document fields and terms. ClickHouse is column-oriented: it stores values by column and can read only the columns a query needs. For telemetry dominated by filtering, grouping and aggregation over repeated fields, that design can be efficient.
Compression is workload-dependent. Repeated service names, status codes and low-cardinality attributes can compress well; highly variable payloads and large text bodies may not. Partitioning, sorting keys, codecs, secondary indexes, replication and query patterns all affect disk and CPU use. No fixed compression ratio can be promised for arbitrary production logs.
The frequently cited SigNoz benchmark compared a 500 GB synthetic log workload. Its repository identifies Elasticsearch 8.4.3, Logstash 8.4.3, Loki 2.6.1, Promtail 2.6.1, Grafana 9.2.1, ClickHouse 22.4.5 and SigNoz OTel Collector 0.55.3. SigNoz reported roughly half the storage of ELK and faster aggregate queries in that setup; the repository notes Elasticsearch performed better for simple COUNT queries and Loki's high-cardinality configuration was constrained. [4]
This is useful historical architecture evidence, not a current 2026 head-to-head test. It is vendor-produced, synthetic, version-specific and sensitive to pipeline tuning. Do not use it to size a production cluster or forecast savings without a workload-specific proof of concept.
What the architecture can—and cannot—tell you
Can support: column pruning, vectorized execution and compression can make analytical scans efficient for suitable schemas and queries.
Cannot guarantee: lower total cost, lower latency or lower storage for every payload, retention policy, replication factor or workload.
Operational consequence: merges, replication, partitions, Keeper health, disk capacity and upgrades become platform-team responsibilities when self-hosted.
Run representative production payloads through candidate systems. Include normal and burst traffic, high-cardinality attributes, JSON parsing, retention expiry, concurrent dashboards and incident-time queries. Record ingestion lag, dropped data, disk amplification, CPU, memory, p95/p99 query latency and operator time.
03 The benchmark has aged: compare current products
The original benchmark used software versions from 2022–2023. Product capabilities have changed since then. A 2023 result does not settle a 2026 procurement decision.
| Product | Current evidence checked | Editorial treatment |
|---|---|---|
| Elasticsearch | Elastic documents logsdb mode as default for new matching log data streams in Elastic Stack 9.0+. Elastic reports up to 60% lower storage in its own benchmarks, with a 10–20% indexing-performance impact; results vary by dataset and version. [5] | Replace the previous 65% figure with Elastic's published 60% claim and attribute it to Elastic. |
| Grafana Loki | Grafana's Loki 3.7 release notes document ongoing 3.7.x changes and maintenance. [6] | Do not claim blanket performance parity or universal 3.7.x advantage without comparable tests. |
| ClickHouse | Storage, execution and JSON capabilities evolve by release; results depend on version and table design. | Use a release-specific test rather than projecting generic improvements into an old benchmark. |
Elastic's 60% statement is a vendor benchmark, not an independently guaranteed result. Elastic also states a 10–20% indexing-performance impact. The fair comparison is not "ClickHouse uses half the storage" versus "Elasticsearch uses 60% less"; both systems must be tested with identical data, retention and query requirements.
Loki's architecture is not interchangeable with Elasticsearch or ClickHouse. Its index and chunk model, label-cardinality guidance, query language and object-storage orientation produce different trade-offs. A comparison that collapses all three into one storage number is analytically weak.
Treat the SigNoz 2023 benchmark as historical context. Run a 48-hour to two-week proof of concept using current stable releases and real telemetry. Publish the test harness, versions, hardware, dataset profile, configuration and query suite alongside performance claims.
04 OpenTelemetry: less lock-in, not zero migration cost
OpenTelemetry (OTel) is a vendor-neutral collection of APIs, SDKs, semantic conventions and collector components for generating, processing and exporting telemetry. SigNoz's OTel orientation lets teams instrument services with OTel and route telemetry to an OTel-compatible backend rather than depend exclusively on a proprietary agent.
This reduces switching friction, especially when teams standardize instrumentation before selecting a backend. A collector pipeline can route, batch, filter, enrich and export telemetry to multiple destinations during migration, making dual-shipping and gradual cutover practical.
However, OTel does not make migration automatic. Teams may still need to replace proprietary SDKs, vendor-specific attributes, custom span processors, log pipelines, dashboards, monitors, SLO definitions, alert routing, synthetics and incident runbooks. Query languages and product semantics differ even when telemetry is standardized.
The strategic value is portability of instrumentation—not portability of every observability workflow.
A staged migration sequence
Inventory: list SDKs, agents, exporters, custom attributes, dashboards, alerts, retention tiers and integrations.
Standardize: move instrumentation toward OTel semantic conventions and collector-based routing where feasible.
Shadow: mirror one high-volume service to SigNoz while retaining the existing backend as the operational source of truth.
Validate: compare data completeness, alert parity, p95/p99 query performance, trace continuity and incident workflows.
Cut over by signal: migrate high-volume logs first only if the pilot confirms savings and query behavior; then traces and metrics.
A migration should proceed only when on-call engineers can detect, investigate and resolve the same classes of incidents with equivalent or better reliability.
05 Pricing: normalize the meters before comparing
The current public pricing pages reviewed for this report show SigNoz Cloud Teams at $0.30/GB of logs, $0.30/GB of traces and $0.10 per million metric samples, with 15-day log/trace retention and one-month metrics retention in the displayed baseline. The plan has a $49 monthly minimum including $49 of usage. SigNoz Enterprise Cloud is listed from $4,000/month, including monthly ingestion usage up to $4,000; other enterprise deployment options are custom-priced. [7][8]
Datadog's public annual list pricing shows Infrastructure Pro from $15/host/month and APM from $31/host/month when Infrastructure Monitoring is attached. Standard log ingestion is $0.10/GB and 15-day indexed log retention is $1.70 per million indexed events. Flex Logs storage is $0.05 per million stored events. [1][2]
Datadog Pro infrastructure includes 100 custom metrics per host. For 50 hosts, that is 5,000 included custom metrics. If the workload has 10,000 custom metrics, the model treats 5,000 as billable at $5 per 100 custom metrics/month. Actual billing depends on Datadog's metric counting rules and account configuration. [9]
| Dimension | Datadog annual list | SigNoz Cloud baseline | Caveat |
|---|---|---|---|
| Infrastructure | $15/host/month | No host fee | Product bundle and host count must match. |
| APM | $31/host/month | Usage-based traces | Not a like-for-like feature bundle. |
| Logs | $0.10/GB ingestion + $1.70/million indexed events | $0.30/GB ingested | Indexing share and event size drive Datadog cost. |
| Flex storage | $0.05/million stored events | Included in log GB price | Heavy rehydration/query may add costs. |
| Metrics | 100 custom metrics/host included | $0.10/million samples | Custom metrics and samples are different units. |
06 Worked cost model: 100 GB/day
This reproducible model uses a 30-day month and 50 production hosts: 70 GB/day of logs, 30 GB/day of traces and 20,000 active metric series (10,000 standard and 10,000 custom), scraped every 30 seconds.
| Signal | Assumption | Monthly volume / count |
|---|---|---|
| Logs | 70 GB/day × 30; 1.0 KB average event | 2,100 GB; ~2.1 billion events |
| Traces | 30 GB/day × 30; 1.5 KB average span | 900 GB; ~600 million spans |
| Indexed spans | 15% of ingested spans | 90 million indexed spans |
| Metrics | 20,000 series × 2 samples/min × 60 × 24 × 30 | 1,728 million samples |
Datadog assumptions
50 hosts × ($15 Infrastructure Pro + $31 APM) = $2,300/month.
Logs: 2,100 GB × $0.10 = $210 ingestion. At 1.0 KB/event, 2.1 billion events indexed at 100% cost $3,570. Total logs = $3,780.
FinOps: 25% indexed = 525 million × $1.70 = $892.50; 75% Flex = 1,575 million × $0.05 = $78.75; ingestion = $210. Total logs = $1,181.25.
Traces: 90 million indexed spans minus 50 million included (1 million per APM host × 50) = 40 million excess × $1.70 = $68.
Custom metrics: 10,000 less 5,000 included (100/host × 50) = 5,000 billable; at $5/100 = $250.
The Flex scenario assumes the 75% excluded from standard indexing is stored in Flex Logs and does not add heavy rehydration/query compute. It is a modeled configuration, not a claim that all excluded logs are free to query.
Interactive TCO Calculator: Datadog vs. ClickHouse
Simulate annual observability cloud infrastructure spend based on real workload telemetry.
07 Results, sensitivity and what the model does not prove
| Monthly component | Datadog 100% indexed | Datadog 25% indexed / 75% Flex | SigNoz Cloud |
|---|---|---|---|
| Infrastructure + APM hosts | $2,300 | $2,300 | $0 host fee |
| Logs | $3,780 | $1,181.25 | $630 |
| Traces | $68 | $68 | $270 |
| Metrics | $250 | $250 | $172.80 |
| TOTAL | $6,398 | $3,799.25 | $1,072.80 |
| SigNoz modeled saving | 83.2% | 71.8% | — |
Savings = (Datadog modeled total − SigNoz Cloud modeled total) ÷ Datadog modeled total. The model compares public list rates and holds the telemetry workload constant; it is not a universal customer outcome.
Self-hosted scenario: illustrative, not a verified quote
The original draft used $1,233/month for AWS infrastructure in Mumbai and imputed SRE labor of $800/month in India or $3,000/month in the US. Retaining those assumptions yields $2,033/month (India labor) or $4,233/month (US labor). These are not a validated AWS Pricing Calculator export and must be treated as scenario inputs—not sourced market prices. The estimate simplifies or omits transfer, backups, monitoring, support, capacity headroom, taxes and operational coverage.
| Scenario | Modeled monthly total | Interpretation |
|---|---|---|
| Self-hosted + India SRE labor | $2,033 | Cheaper than modeled Datadog baseline; costlier than SigNoz Cloud. |
| Self-hosted + US SRE labor | $4,233 | About 11.4% above the FinOps-tuned Datadog scenario. |
The model supports a narrow conclusion: SigNoz Cloud can be materially cheaper for this workload and price snapshot. It does not prove that self-hosting is always cheaper, that every Datadog account pays list price, or that feature parity exists. Rebuild the spreadsheet with your own event size, indexing share, host count, retention, metric allocation and labor rates.
08 Indian data governance: residency is not compliance
Keeping telemetry in an Indian cloud region can support internal data-custody policies and avoid some cross-border transfer paths. But server geography alone does not establish compliance with the Digital Personal Data Protection Act, 2023 or the Digital Personal Data Protection Rules, 2025.
The final DPDP Rules were published in the Gazette on 13 November 2025. Their commencement is phased: Rules 1, 2 and 17–21 commenced on publication; Rule 4 is scheduled one year after publication; Rules 3, 5–16, 22 and 23 are scheduled eighteen months after publication. The Act's commencement is also phased. The article must not imply that every operational obligation is already in force on 19 November 2026. [10][11]
Rule 7 requires a Data Fiduciary, on becoming aware of a personal data breach, to notify affected Data Principals without delay. It also requires notification to the Board without delay and updated, detailed information within 72 hours, subject to the rule and any longer period the Board may allow on written request. This is not a blanket rule that every affected individual must receive a full report within 72 hours. [10]
Practical controls for observability platforms
Data minimisation: avoid phone numbers, email addresses, payment identifiers, tokens and unnecessary personal data in logs and span attributes.
Redaction: use OTel Collector processors and application-side controls before data reaches storage.
Access: least privilege, role separation, strong authentication and auditable administrative access.
Retention: define purpose-bound TTLs and deletion workflows; do not retain raw telemetry indefinitely.
Security: encryption in transit and at rest, key management, backups and incident response.
Processor governance: document vendor roles, processing instructions, subprocessors and contractual safeguards.
This is an editorial technology analysis, not legal advice. Organisations should obtain advice on applicability, commencement, sector-specific RBI requirements, contracts and their actual personal-data flows.
09 Honest trade-offs: where SigNoz can fall short
1. Day-two ClickHouse operations
A self-hosted deployment requires operational ownership of storage capacity, merges, replication, partitions, backups, upgrades, Keeper quorum, query performance and incident response. Ingestion bursts can create merge pressure and disk growth. Adding replicas or shards is not a substitute for capacity planning and tested recovery procedures.
2. Feature breadth and integration depth
Datadog sells a broad observability and security platform, not only a logs-metrics-traces interface. Depending on the selected plan, buyers may rely on continuous profiling, real-user monitoring, session replay, synthetics, cloud security, security analytics, extensive integrations and vendor-supported workflows. Compare the exact capabilities in use; do not assume feature parity.
3. Enterprise governance and support
SigNoz's current pricing page lists Enterprise Cloud starting at $4,000/month and describes dedicated cloud, BYOC and self-hosted support options. The published page distinguishes features by offering; not every enterprise capability should be assumed in Community Edition. Confirm SSO/SAML, RBAC, audit logs, tenancy, retention controls, support SLA and deployment responsibilities in writing. [7][8]
4. Alternative architectures
Teams already invested in Prometheus and Grafana may prefer the Grafana LGTM stack. Object-storage-oriented systems such as OpenObserve target a different operational trade-off. Evaluate query language, data model, object-storage costs, ingestion durability, retention, cardinality, operational complexity and existing engineering skills.
5. Shared storage and replication claims
ClickHouse documentation describes SharedMergeTree as the engine used by ClickHouse Cloud. Avoid turning this into a blanket claim that self-managed ClickHouse cannot use object storage or tiered storage: self-managed deployments have their own storage policies and replication designs. Tie architecture statements to the relevant version and documentation. [12]
10 Decision matrix and migration playbook
| Choose | When it fits | Main caution |
|---|---|---|
| SigNoz Cloud | You want OTel-native logs, metrics and traces with usage-based pricing and low operational overhead; the model indicates a material saving. | Confirm retention, region, enterprise requirements, ingestion limits and feature coverage. |
| SigNoz self-hosted | You need infrastructure control, can operate ClickHouse reliably and have platform/SRE capacity. | Price SRE labor, HA, storage replication, backups, upgrades and on-call—not just the licence. |
| Datadog | You rely on broad integrations, mature enterprise workflows, profiling, RUM/security or have limited SRE bandwidth. | Apply FinOps controls and validate exclusions, Flex storage and annual commitments. |
A low-risk migration sequence
Step 1 — decouple instrumentation: Standardise services on OTel where feasible and route through collectors while continuing to send telemetry to the existing platform.
Step 2 — shadow pilot: Mirror one representative high-volume service for 48 hours to two weeks using real payloads and realistic peaks.
Step 3 — test operational parity: Rebuild priority dashboards and alerts; test trace-to-log correlation, access, retention, redaction and runbooks.
Step 4 — compare full economics: Include compute, storage, backups, transfer, support, engineering labor, training and migration costs.
Step 5 — cut over gradually: Move one signal or service at a time. Keep rollback paths until on-call teams complete incident exercises.
Write the go/no-go threshold before the pilot: acceptable data loss, ingestion lag, p99 query latency, alert parity, storage growth, monthly cost and operator hours. This prevents selecting a winner based only on the most favourable metric.
11 Conclusion and source register
SigNoz does not make Datadog obsolete, and open source does not make observability free. Its strongest proposition is the combination of OpenTelemetry-based instrumentation, ClickHouse-backed analytical storage and a choice between managed cloud and self-managed deployment. That can improve portability and change the cost curve for suitable workloads.
In this report's 100 GB/day reference model, SigNoz Cloud is approximately 72% cheaper than a FinOps-tuned Datadog configuration and 83% cheaper than a fully indexed configuration at the published annual list rates checked on 19 November 2026. Self-hosted economics are sensitive to infrastructure sizing and engineering labor.
The procurement decision is empirical: standardise instrumentation, run a workload-specific pilot, compare incident outcomes and calculate total cost with the same retention and governance requirements. Treat historical vendor benchmarks as hypotheses to test—not guarantees.
Methodology & Limitations
Public product pages and official documentation were checked on 19 November 2026. The cost model uses stated assumptions, public US-dollar annual list prices and arithmetic shown in the report. It excludes taxes, negotiated rates, some add-ons and contract-specific entitlements. The self-hosted infrastructure estimate is retained only as an illustrative assumption from the draft, not an independently verified cloud quote. This report is editorial analysis, not legal, financial or procurement advice.
Referenced Architectural Specifications in this Report
OpenTelemetry (OTel)
A Cloud Native Computing Foundation (CNCF) open-source observability framework providing a vendor-agnostic specification, APIs, SDKs, and collector pipelines for generating, processing, and exporting traces, metrics, and logs.
ClickHouse Columnar DBMS
An open-source, column-oriented database management system engineered for real-time analytical query processing (OLAP) at petabyte scale using vectorized SIMD execution and columnar data compression.
Editorial Verification & Primary Evidence Register
VerifiedStrictly primary research: cross-referenced against official filings, regulatory circulars, and reproducible data sets.
- [1]
Official published rates for log ingestion ($0.10/GB) and 15-day indexed log events ($1.70/million events).
https://www.datadoghq.com/pricing/ - [2]
Annual list rates for indexed logs, Flex Logs storage ($0.05/million), Infrastructure Pro ($15/host), APM ($31/host), and custom metrics.
https://www.datadoghq.com/pricing/list/ - [3]
Official company announcement of funding led by SignalFire with participation from Y Combinator.
https://signoz.io/blog/signoz-funding/ - [4]
Historical 500 GB synthetic log benchmark configuration, software versions, and comparative findings.
https://github.com/SigNoz/logs-benchmark - [5]
Logsdb mode documentation and published claims of up to 60% storage savings with trade-offs.
https://www.elastic.co/docs/manage-data/data-store/data-streams/logs-data-stream - [6]
Release-specific architecture changes, index features, and maintenance history.
https://grafana.com/docs/loki/latest/release-notes/v3-7/ - [7]
Teams usage rates ($0.30/GB logs & traces, $0.10/M samples), $49 monthly minimum, and Enterprise Cloud baseline.
https://signoz.io/pricing/ - [8]
Enterprise deployment options (Dedicated Cloud, BYOC, Self-Hosted) and published $4,000/month starting tier.
https://signoz.io/enterprise/ - [9]
Custom metrics counting rules, 100 metrics/host allotment, and overage billing mechanics ($5 per 100/mo).
https://docs.datadoghq.com/account_management/billing/custom_metrics/ - [10]
Official Gazette notification of Digital Personal Data Protection Rules, phased commencement schedule, and Rule 7 breach notification rules.
https://egazette.gov.in/WriteReadData/2025/267650.pdf - [11]
Official MeitY regulatory portal detailing enforcement timelines and statutory compliance under the DPDP Act 2023.
https://www.meity.gov.in/ - [12]
ClickHouse Cloud shared object-storage engine architecture and separation of compute and storage.
https://clickhouse.com/docs/cloud/reference/shared-merge-tree
Found this analysis valuable?
Bookmark it to your reading list or share it with your network.
This report was compiled and verified according to The Story Desk's editorial standards. We do not accept payment to influence analytical conclusions. Notice a factual error? Submit an editorial review request through our transparent Corrections Desk.

The Story Desk
Editorial DeskThe Story Desk newsroom provides independent, verifiable reporting and analytical deep-dives across technology, finance, policy, startups, and macroeconomics.
Comments & Reviews
0Share your feedback, perspectives, or questions on Inside SigNoz: How an Indian Open-Source Engine Challenges Datadog.
No comments yet
Be the first to share your perspective, question, or review on this article above.
More From Technology & AI
View All Desk Stories →Inside NPCI: The Distributed Systems Architecture Behind 16.5 Billion Monthly UPI Transactions at Fractional-Paise Settlement Cost
A systems-level engineering teardown of India's real-time retail payments network: two-tier ISO 8583 switch topologies, core banking system (CBS) circuit breaking, HSM cryptographic latency budgets, and the zero-interchange operational paradox.

India's Sovereign Compute Bet: Inside the 10,000-GPU National Supercluster
As hyperscalers corner proprietary AI chips, New Delhi's ₹10,372 crore IndiaAI Mission targets 10,000+ enterprise GPUs under tender to subsidize domestic foundation models. Here is the operational architecture, capacity status, and voucher economics.

BharatGen Goes Live: India's 22-Language Generative AI Foundation Opens for Public Beta
Spearheaded by IIT Bombay and MeitY, BharatGen unveils multimodal foundation models trained on diverse Indian linguistic corpora to support governance, education, and healthcare automation.

