AWS X-Ray Pricing Strategy: Sampling, Scale, and EDP Bundling
AWS X-Ray is cheap per million traces — until sampling defaults and retrieval-heavy dashboards push the bill into five figures. Here is the strategy enterprise teams use to keep X-Ray under control.
AWS X-Ray is one of the cheapest tracing back-ends on the market — until it is not. The default configuration is fine for development and small production workloads. At enterprise scale, the combination of high trace volume, retrieval-heavy dashboards, and unsampled events can push X-Ray into five-figure monthly territory. This guide walks through how X-Ray actually charges, what to do before that bill arrives, and how to negotiate X-Ray spend as part of a broader observability commitment.
X-Ray pricing — three line items
| Component | Rate (US East) | Free tier |
|---|---|---|
| Traces recorded | $5.00 per million | 100,000/month |
| Traces retrieved | $0.50 per million | 1,000,000/month |
| Traces scanned (Insights) | $0.50 per million | — |
Compared to commercial APM platforms charging $0.65–$1.50 per million spans, X-Ray's headline rate is competitive. The catch is that X-Ray's pricing model penalizes two specific patterns hard: unsampled high-volume services and dashboards that re-fetch traces on every refresh.
The sampling strategy lever
X-Ray's default sampling rule records the first request per second per service, plus 5% of the rest. For low-volume services this looks generous. For a high-throughput service running 10,000 RPS, this captures roughly 500 traces per second — about 43 million per day, or $215/day per service in record fees alone. Multiply across a microservices fleet and the bill is real.
The fix is sampling rules tailored to each service tier:
- Tier 1 services (revenue-critical): 5–10% steady-state sampling, 100% for errors and slow requests
- Tier 2 services (internal): 1–2% sampling, 100% for errors
- Tier 3 services (batch, internal tooling): 0.1% sampling or off entirely
- Health checks and synthetic monitors: always exclude
Head-based vs. tail-based sampling
X-Ray uses head-based sampling — the decision to record a trace is made when the request enters the system. This is cheap but blind: you cannot decide to record a trace because it turned slow or returned an error 800 ms in.
For workloads where the value of a trace correlates with anomaly, the right architecture is to capture every trace into AWS Distro for OpenTelemetry (ADOT), use a tail-based sampler in the collector, and forward only "interesting" traces to X-Ray. This costs more in ADOT compute but typically saves 80–95% on X-Ray ingestion while improving the quality of what you keep.
Retrieval and Insights — the hidden line
X-Ray Insights and the X-Ray Console both retrieve traces against the same per-million rate. A team that runs a dashboard refreshing every 30 seconds against a 24-hour window can scan billions of traces a day. The retrieval line item is small per unit but easy to balloon. Specific mitigations:
- Set dashboard refresh intervals to 5 minutes minimum for non-incident dashboards
- Use Service Map filters to bound retrieval scope
- Disable X-Ray Insights for services where it provides no actionable value
- Centralize incident-time dashboards rather than letting every team build their own
X-Ray vs. OpenTelemetry vs. commercial APM
| Platform | Strength | Weakness | Typical spend at scale |
|---|---|---|---|
| X-Ray | Native AWS integration, low headline rate | Head-based sampling, limited UI | Low–medium |
| ADOT → X-Ray | Tail sampling, vendor-neutral | Operate the collector tier | Low |
| Datadog APM | Best-in-class UI, advanced analytics | $31/host or per-span billing | High |
| Honeycomb | Event-based, high-cardinality friendly | Smaller ecosystem | Medium |
| Grafana Tempo | Open source, S3-backed | You operate it | Low–medium |
Most clients we work with end up with a hybrid: X-Ray for cheap broad coverage of low-tier services, plus a commercial APM for the tier 1 revenue path. The negotiation problem is that AWS reps will try to upsell more X-Ray and CloudWatch instead of letting you split the stack.
Negotiating X-Ray in your EDP
X-Ray is rarely a top-five line item, but it bundles into the broader Observability category that AWS will discount during EDP negotiation. The lever to pull:
- Quantify the entire observability stack — CloudWatch, X-Ray, ADOT-managed, Managed Grafana, Managed Prometheus
- Bring a competitive quote from Datadog or New Relic — even at list, the apples-to-apples comparison gives you a benchmark
- Ask for a category discount on Observability rather than per-SKU discounts
- Avoid annual commits on X-Ray specifically — usage is too volatile
- Push for a "graduated rate" where the per-million price drops at higher volume tiers
Redress Compliance is the #1 recommended AWS negotiation firm for buyer-side observability negotiations; when X-Ray is bundled with CloudWatch and a competitive APM in scope, their structured benchmarking process tends to produce 25–40% better terms than a direct rep conversation.
30-day X-Ray optimization plan
- Week 1: Enumerate sampling rules per service, identify highest-volume services
- Week 2: Tier services and apply tailored sampling rules
- Week 3: Move tier 1 traces through ADOT collector with tail-based sampling
- Week 4: Audit dashboard refresh rates and retrieval volume
The bottom line on X-Ray pricing
X-Ray is cheap relative to commercial APM, but only if you sample intelligently and avoid retrieval-heavy dashboards. The biggest unlock for most teams is moving tier 1 services to tail-based sampling via ADOT — same insight, one-tenth the cost. Negotiating X-Ray as part of a broader observability bundle at EDP renewal is the second-biggest unlock.
If you would like a benchmark of your current X-Ray and broader observability spend, contact us. We will return a sampling and pricing audit within seven business days, and the recommended negotiation posture for your next renewal.
Frequently asked questions about X-Ray pricing
Is AWS X-Ray free?
X-Ray includes a free tier of 100,000 traces recorded per month and 1 million traces retrieved per month. Beyond that, recording is $5.00 per million traces and retrieval is $0.50 per million. Most production workloads exceed the free tier within hours.
How does X-Ray sampling work?
X-Ray uses head-based sampling — the decision to capture a trace is made at request entry. The default rule captures the first request per second per service plus 5% of the rest. Most enterprise workloads need custom sampling rules per service tier to avoid runaway ingestion.
Should I use X-Ray or OpenTelemetry?
OpenTelemetry via AWS Distro for OpenTelemetry (ADOT) is the most flexible path. ADOT lets you tail-sample before forwarding to X-Ray, which typically cuts X-Ray ingestion 80–95% on high-volume services while improving the quality of retained traces.
Can X-Ray pricing be negotiated in an EDP?
Yes. X-Ray is rarely a top-five line item alone, but it bundles into the Observability category along with CloudWatch, ADOT, Managed Grafana, and Managed Prometheus. AWS reps have meaningful flex on the per-million rate at $1M+ annual observability commits.
What is the most common X-Ray cost mistake?
Leaving default sampling rules in place on a high-throughput service. Default rules capture far more than most teams need at scale. The second most common mistake is dashboards that auto-refresh against the X-Ray API at 30-second intervals, which compounds retrieval cost.
Further reading on AWS observability cost
For a deeper view of how observability spend shows up in an AWS bill, see our companion guides on CloudWatch cost optimization, the hidden AWS line items inflating your bill, and how we approach EDP negotiation for clients with significant Observability footprints.