COMPARE / DATADOG-AND-A-SPREADSHEET
The real incumbent isn’t a vendor.
It’s a spreadsheet.
Datadog is excellent at what it's for: services, hosts, containers, APM traces, logs. If your web tier is on it, keep it there — this page is not an argument to replace it, and telemetria ingests OpenTelemetry precisely so you don't have to.
The seam: Datadog sees your Spark executors as anonymous pods and your Databricks bill not at all. So every month, someone on your team exports the CUR, exports system tables, and joins them in a spreadsheet by hand — the actual incumbent we replace. It's three days late, row-level wrong, and nobody trusts it enough to act on.
| dimension | Datadog + spreadsheet | telemetria |
|---|---|---|
| services & APM | world-class, keep it | — not an APM, by design |
| k8s infrastructure | pods, nodes, events — anonymous | same layer, joined to Spark apps, jobs, teams |
| spark visibility | JVM metrics at best | stage/task/executor, skew, spill, mid-run |
| platform spend (DBUs, credits) | — invisible | attributed to job → table → team, daily |
| the monthly join | a spreadsheet, ~3 engineer-days/mo, always stale | continuous, with the math shown |
| actioning findings | manual tickets from the spreadsheet | modeled delta → PR → verified |
| cost of the tool itself | Datadog you already pay + hidden engineer time | published bands on monitored spend |
choose Datadog + spreadsheet if
- Your data workloads are small enough that one engineer-day a month of spreadsheet work genuinely covers it.
- You have no Spark/Databricks footprint — telemetria's depth would be wasted on generic containers.
choose telemetria if
- The monthly cost spreadsheet exists, someone you like maintains it, and it's always three days late.
- You've been paged for node pressure that turned out to be someone's skewed join — and it took two tools and four hours to find out.
- You want Datadog to stay exactly where it is, with the workload layer covered by something that speaks Spark.
Coexistence is the design: keep Datadog for services. telemetria ingests OpenTelemetry, correlates its traces with the workload layers it can't see, and never asks you to migrate a dashboard.