Stop scaling and start tuning - Matt Burke - NDC Oslo 2026

Stop scaling and start tuning - Matt Burke - NDC Oslo 2026

Title history
  • Stop scaling and start tuning - Matt Burke - NDC Oslo 2026 — observed on 20261026-2337 UTC (YouTube)

by Matt Burke · View the channel on PSI

Applied Sciences & Engineering Computing & Cybersecurity UYComputer scienceUYFComputer architecture and logic design
🎙 Matt Burke 👥 230K 📅 October 7, 2026 ⏱ 55 min 👁 2K 📄 expert opinion 🧭 2026-10-26
Available in: English (current) Français

Keywords

N+1 querylazy loadingeager loadingbatchingrecursive CTE

Summary

The talk begins with a common scenario: a web application becomes slow as the database grows, and teams often choose to scale up resources rather than fix the root cause. Matt Burke shares his experience of hitting this point where scaling no longer buys enough time due to a steeper growth curve. He explains that after repeated scaling, costs can balloon while the problem returns. The decision is made to invest in performance tuning instead. The first step is setting up comprehensive telemetry to gain insight into where time is spent. Burke discusses using tools like Application Insights and Zipkin to obtain call traces, spans, and raw SQL queries. He identifies three main patterns of poor performance: many small database queries (N+1), a single dominant slow query, and mysterious idle pauses. He then dives into the most common issue: N+1 queries, showing a demo where a dashboard page issues hundreds of queries to load a list with details. He attributes this to architectures that separate logic into repositories and response builders, and to ORM lazy loading which triggers extra queries when navigation properties are accessed. Solutions include eager loading with joins, batching multiple queries into fewer round trips, and disabling lazy loading to force explicit loading. For hierarchical data, he demonstrates using PostgreSQL recursive common table expressions to replace recursive application loops. The talk concludes by emphasizing that tuning can flatten the growth curve and that practical fixes—like adding indexes, reviewing query plans, and optimizing data access patterns—are often more effective than scaling. Burke provides concrete metrics showing the same page going from 7.8 seconds to under 400 milliseconds.

269 words

Critical Evaluation

Value of the Information & Strength of the Argument

The value of the talk lies in its practical, experience-based approach. Burke does not merely list theoretical problems; he shows a real-world scenario with measurable improvements. He argues convincingly that scaling is a short-term fix that becomes unsustainable, and that investing in tuning can yield dramatic performance gains. His reasoning is pragmatic: he weighs the cost of development time against the cost of cloud resources and the risk of breaking features. He also emphasizes the importance of telemetry and baselines, demonstrating that without good monitoring, optimization is guesswork. The argumentation is solid, supported by concrete examples and before/after metrics. However, the talk is anecdotal and does not include controlled experiments or comparative studies, so the claims are based on the speaker’s experience rather than systematic analysis.

Scientific Rigor, Source Quality, Title Accuracy

The talk is scientifically rigorous in the sense that it follows a methodical approach: identifying symptoms, setting up monitoring, gathering data, and testing solutions. Burke references specific tools (Zipkin, App Insights) and ORM features (lazy loading, eager loading) but does not provide external citations or peer-reviewed sources. The sources cited in the description are conference links (NDC Conferences, NDC Oslo), which are not technical references but organizational. The title is appropriate because the core message is to stop scaling and instead tune the system, which matches the content precisely. No comments were provided for analysis, so public reception cannot be assessed.

242 words

Title / Content Match

The title accurately reflects the content: the talk advocates for tuning database and application performance instead of scaling cloud resources, and demonstrates specific techniques to achieve this.

Quality & Reliability

8/10

The talk is grounded in real-world experience and provides concrete metrics and actionable solutions. The speaker clearly explains the reasoning behind each fix and offers a practical blueprint. However, it is anecdotal and not peer-reviewed, and some claims rely on personal observation rather than formal testing.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

  • NDC Conferences — Conference website for NDC events.
  • NDC Oslo — The specific conference where this talk was given.

External References

Contribution & Novelties

The talk provides a practical, hands-on blueprint for database performance tuning that emphasizes solving root causes rather than scaling resources. It systematically addresses common pitfalls like N+1 queries, lazy loading, and inefficient data access patterns, offering concrete fixes such as eager loading, batching, and recursive CTEs. The speaker shares real-world metrics to demonstrate the impact of these changes, making the content immediately actionable.

Pour aller plus loin :

  • N+1 query problem — Overview of the N+1 query pattern and its impact on database performance.
  • Eager loading in EF Core — Official documentation on eager loading techniques to avoid lazy loading overhead.
  • PostgreSQL recursive queries — Documentation on common table expressions including recursive ones, used for hierarchical data.
  • Application Insights — Microsoft’s telemetry and monitoring service, mentioned in the talk for performance diagnostics.

132 words

Radar Profile

The radar profile shows strong scores in information quantity and quality, indicating a content-rich and well-structured talk. The technical level is moderately high, suitable for developers familiar with ORMs and SQL. The overall reliability is good, supported by concrete examples and measurements, though it relies on anecdotal evidence.

Reliability 8/10