Tens of trillions of daily feature flag evaluations. Scaling at that magnitude requires absolute objectivity. That’s why LaunchDarkly standardized on New Relic. Bundling feature flags and monitoring together creates a conflict of interest. LaunchDarkly chose neutral telemetry to isolate variables and ensure unbiased, high-cardinality tracking. Read more about it from our CTO Michael Frendo here: https://lnkd.in/gGq_qRMX
LaunchDarkly Chooses New Relic for Unbiased Feature Flag Evaluations
More Relevant Posts
-
At scale, observability is not just about visibility—it is about objectivity. New Relic’s point about neutral telemetry is important: when the system being measured also controls the feature flags, teams risk blurring cause and effect. Independent, high-cardinality signals create a cleaner basis for operational decisions. #Observability #EngineeringLeadership #DevOps #SoftwareReliability
Tens of trillions of daily feature flag evaluations. Scaling at that magnitude requires absolute objectivity. That’s why LaunchDarkly standardized on New Relic. Bundling feature flags and monitoring together creates a conflict of interest. LaunchDarkly chose neutral telemetry to isolate variables and ensure unbiased, high-cardinality tracking. Read more about it from our CTO Michael Frendo here: https://lnkd.in/gGq_qRMX
To view or add a comment, sign in
-
-
What if observability is creating more work instead of less? Learn why preserving context, not collecting more telemetry, is the key to faster investigations and better decisions. https://bit.ly/4vsRL1h
To view or add a comment, sign in
-
-
In observability, we’re often tempted to collect as much data as possible. But without a strategy guiding data collection, even the best intentions can lead to higher costs and more waste. Join Diana Todea (VictoriaMetrics) and host Kate Goldenring to get practical insights on reducing telemetry bloat and building greener observability pipelines, and learn why teams rarely query more than 50% of the metrics they store. Listen to the full episode now ➡️ https://lnkd.in/d9NJHBd2
To view or add a comment, sign in
-
What if observability is creating more work instead of less? Learn why preserving context, not collecting more telemetry, is the key to faster investigations and better decisions. https://bit.ly/4p4xnlx
To view or add a comment, sign in
-
-
Everyone unifies telemetry at query time. That's already too late, and too expensive. By the time MELT data hits storage, it's three separate piles. You pay twice: once to store them apart, and again to join them back together on every single query. Under load, that join is exactly what breaks. We flipped it. We compile the causal graph at deploy time and correlate in-flight, before storage. Think of it like a materialized view for causality: the joins are already done. What lands in storage is compact and natively connected. Correlation should be bounded by your service topology, not your event volume. It needs to hold as you scale, instead of degrading when you need it most.
To view or add a comment, sign in
-
-
Most teams bolt governance onto data after it lands in the warehouse. By the time you catch something sensitive, it's already been replicated three times. In this episode of What's New in Data, Striim's Alok Pareek breaks down how CDC-driven architectures are evolving, and why in-flight detection is becoming essential for modern governance teams. https://lnkd.in/gCSze_JP
To view or add a comment, sign in
-
Your pre-launch benchmark is a liar. 🤥 I watched it pass Monday and silently nuke our agents Tuesday. The provider tweaked routing. Shifted serving conditions. Your prompts didn't change, but your system broke. 📉 Evaluate only before launch? You have zero baseline for the task your agents actually face today. Task-level evaluation is not a pre-flight checklist. It is a production feature. Rerun it continuously. Know exactly when a dependency shifted versus when your code actually broke. 🛠️
To view or add a comment, sign in
-
A RAG pipeline that worked at launch will not necessarily work six months later. The data sources it was built on keep changing. Nodes update their schemas. Records get rewritten. Fields get renamed or dropped. The embedding model that made sense at build time gets superseded. None of these events trigger an alert. The pipeline keeps running. Latency stays flat. Queries return results. The results are just increasingly wrong. At thousands of nodes, the gap between what the index contains and what the sources currently say is not a monitoring problem you can solve by checking more often. It is an architectural problem that compounds from the moment the system ships. What does your pipeline do when a source node changes its schema mid-flight?
To view or add a comment, sign in
-
The Rapidemo June update is out! 🎉 I've published my first case study 🎬 The scenes feature is coming along nicely 🐛 Many bug fixes and stability improvements Read the full update here 👇 https://lnkd.in/d5sM4Mur
To view or add a comment, sign in
-
Planet Labs PBC,(NYSE:PL) a provider of satellite imagery and geospatial intelligence services, is drawing renewed market attention as improving earnings, stronger free cash flow trends, and a substantial contract backlog reshape its financial narrative. https://zurl.co/gxjfb
To view or add a comment, sign in
-
Really interesting point. As systems scale, having clear and trusted insights becomes even more important Danisue Gilpin Brian Peck Pegasus: IT Value Acceleration Services