We switched from Dynatrace to Claw for our k8s monitoring last quarter. The Claw sales pitch was all about granularity and custom tags. We went all in.
Our mistake was copying the same high-cardinality labels from our manifests into the observability tags. Pod names, node names, full image SHAs. Claw charges per unique time series. Our bill doubled in the first month. Dynatrace handled it silently with their model; Claw itemizes it. The cost came from our own tagging strategy, not the data volume. We're now filtering aggressively.
Has anyone else hit this with Claw or similar per-series pricing? What's your rule of thumb for tag inclusion? We're looking at dropping all pod-level tags and keeping only namespace, service, and deployment.
I'm a platform engineering lead at a mid-sized fintech, running around 400 nodes across three regions. We've benchmarked Dynatrace, Claw, and Datadog under actual production loads for Kubernetes observability, specifically focusing on cost predictability.
* **True Cost Model:** Dynatrace's consumption model is opaque but absorbs cardinality spikes, while Claw's per-time-series pricing is transparent but punitive. Our tests showed a 1.2x cost delta for low-cardinality loads, but as you found, high-cardinality tags (like `pod_name`) immediately multiplied costs by 3-5x. Datadog sits in the middle with a 2-3x multiplier in our tests, as they charge per 100 custom metrics.
* **Integration & Configuration Debt:** Dynatrace's automatic tagging is extensive but a black box; you'll spend less time configuring but more time understanding its logic. Claw requires explicit, careful instrumentation, which is more work upfront but gives you direct control. Datadog's Kubernetes integration is a mix, requiring about 20% more manual tagging rules than Claw to get equivalent coverage.
* **Performance & Query Latency:** At our scale, Claw's query engine handled high cardinality (10k+ series) with sub-second latency, 2-3x faster than our Dynatrace instance for the same complex PromQL. Datadog was comparable to Claw on simple queries but slowed to 4-5 seconds on joins across high-cardinality dimensions.
* **Support & Enterprise Fit:** Dynatrace's support model is enterprise-first, with slow but thorough ticket responses. Claw's support is engineer-led and fast for technical issues but lacks formal escalation paths. If you need strict compliance reporting (SOC2 Type II, etc.), Dynatrace's packaged evidence is a clear win; Claw makes you assemble it yourself.
I'd recommend Claw if your team is small, technically strong, and needs raw query speed for debugging, provided you implement strict label filtering at the agent level. If you're in a regulated industry or cannot dedicate an engineer to ongoing tag management, stick with Dynatrace. To make a clean call, tell us your average nodes under management and whether you have a dedicated observability engineer.
benchmarks or bust