Datadog APM
Tail-based or head-based sampling kicks in at cardinality limits. The trace waterfall is often incomplete on the worst-case requests, exactly the ones you need to see during an incident. Custom tags are billed per dimension.
SOLUTIONS / APPLICATION PERFORMANCE MONITORING
Application performance monitoring on traces that aren't sampled.
Distributed tracing with zero sampling at ingest. Over a million spans rendered in a single trace view. Pivot from a span to surrounding logs in the same query language. OpenTelemetry-native. Fully managed.
WHY YOUR CURRENT APM STACK CAN'T KEEP UP
What you've tried, and why the trace isn't complete when it counts.
Every APM stack that controls cost by sampling drops exactly the worst-case requests you opened the trace to debug.
New Relic
Per-user pricing means restricting who can see traces. Per-transaction limits force teams to choose which services get full APM and which don't.
Honeycomb
Strong sampling-aware design, but sampling is still part of the contract. The team carries the cognitive load of "what survived" alongside the incident itself.
Jaeger / Tempo self-managed
Open-source baseline. The platform team owns the cluster, the storage tier, the indexing decisions, and the on-call rotation for the tracing stack itself.
Splunk Observability Cloud
Same data-platform constraints as Splunk. Tracing layered on top inherits the licensing economics that pushed teams off Splunk for logs in the first place.
HOW AXIOM SOLVES IT
APM on the same event storeas the rest of your observability.
EVERY SPAN, NO SAMPLING
Axiom is built to capture every span at production scale. There's no head-based or tail-based sampling at ingest. Every span lands on EventDB, the petabyte-scale event store with 95%+ compression. The Console renders over a million spans in a single trace view, so the waterfall is intact during the incident, not after sampling has shaped it.
SPANS NEXT TO LOGS
Service and operation dashboards roll up the same trace data into the aggregate APM view, with high-cardinality metrics sitting alongside on MetricsDB. APL queries logs, events, and traces (MPL for metrics), so engineers pivot from a slow span to the surrounding logs in one query, without a join. Virtual fields handle query-time transformation on span attributes.
ONE DIAL, NO SURCHARGE
The platform is OpenTelemetry-native, with no proprietary agent to install. Pricing is usage-based on the same dial as the rest of Axiom, so APM doesn't show up as a separate line item with its own per-host or per-dimension surprises. The whole thing is fully managed.
CUSTOMER HIGHLIGHT
How Plex gets real-time visibility intoa vast streaming service on Axiom.
“Axiom easily integrated into our project with their Vercel integration and allowed us to hit the ground running with dashboards and alerts.”
BACKGROUND
Global streaming on Vercel backend.
PROBLEM
Quarter-long observability rollout needed faster.
WHY AXIOM
Pre-built Vercel integration, no sampling.
IMPLEMENTATION
Vercel logs + OpenTelemetry traces.
RESULTS
Dashboards and alerts in days.
PARTNERSHIP
Ongoing as service scales.
WHAT CHANGES FOR EACH ROLE
What changes when you move APMand distributed tracing to Axiom.
FOR SRES AND ON-CALL ENGINEERS
Trust the trace waterfall is complete. No "did this span survive sampling" anxiety on the call you're paged into. Pivot from a slow span to surrounding logs in one query, in one language.
FOR ENGINEERING LEADERS
Cut the APM line item without losing diagnostic depth. One usage-based dial across traces, logs, metrics, and events instead of a per-host or per-dimension APM surcharge.
FOR PLATFORM TEAMS
Drop the self-managed tracing cluster (Jaeger / Tempo) and the sampling decisions that came with it. OpenTelemetry-native ingest keeps traces portable. Virtual fields preserve schema flexibility on span attributes.
WHAT YOU'LL USE
What you'll use.
Distributed tracing — zero sampling at ingest, over 1 million spans in a single trace view
Span waterfall view — complete on the worst-case requests, not just the average ones
Service and operation dashboards — aggregate APM rollups on the same trace data
High-cardinality metrics — MetricsDB handles dimensional complexity natively
Unified with logs, events, traces, metrics — pivot from a span to surrounding logs, no join
OpenTelemetry SDK — no proprietary agent to install or maintain
Integrations — Vercel, AWS, Cloudflare, Kubernetes, Cribl. Edge through backing services.
HOW THIS COMPARES
How this compares.
FAQ
Questions teams ask during APM and distributed tracing evaluation.
SCALE AND COST
No. Every span lands on EventDB. Sampling decisions don't happen at ingest, and the trace waterfall isn't shaped by them at query time either.
Over a million in production. The Axiom Console renders the full waterfall on the worst-case requests, not just the average ones.
No per-dimension surcharge. MetricsDB handles high-cardinality time-series natively for the rolled-up metrics; spans themselves are stored on EventDB without cardinality penalties.
Sub-linear. Per-unit rates drop at higher volume tiers, and there's no per-host APM SKU. See /pricing for current rates.
OPENTELEMETRY
No. OpenTelemetry-native ingest. Your existing OTel collectors and SDKs route directly into Axiom.
Yes, GA as of March 2026. Service-level metrics land alongside the traces on the same usage-based dial.
QUERY AND AI
Yes, in APL, without a join. Logs, events, and traces sit on the same event store; APL queries them together. MPL handles metrics.
Through the native MCP server, in APL. The SRE skill ships with the server: patterns that move agents from hypothesis to proof, lowering MTTR.
APM on traces that aren't sampled. One event store, one bill.
Always free to start. Usage-based scaling. Fully managed.