Traditional code reviews are inadequate for modern software production, necessitating new methods for knowledge transfer and accountability in an AI-driven environment.
About:
Galo Navarro is a software engineer with expertise in distributed systems and AI, currently a Principal Engineer at Midokura, with a strong background in technical leadership.
Website:
Specializations:
Interests:
Outgoing Links:
Subscribe to RSS:
AI workloads require a shift from the cattle model in Kubernetes to a more sophisticated approach that considers resource allocation and topology for optimal performance and cost management.
The post argues that treating Proof of Concept projects with the same standards as production work is essential to avoid technical debt and ensure long-term success.
Traditional code reviews are outdated and ineffective in modern software development, necessitating new methods for knowledge transfer and accountability in an industrial context.
AI can optimize testing workflows by intelligently selecting relevant integration and e2e tests, improving delivery speed while maintaining quality assurance.
The text discusses the impact of AI on the production and consumption of code, highlighting the potential negative effects on software delivery performance. It emphasizes the challenges of scaling delivery pipelines in the face of...
0Why aren't we all serverless yet?
2025-01-09 •
The text discusses the slow adoption of serverless computing despite its appealing value proposition. It explores the reasons for the friction against serverless, including the post-traumatic syndrome of microservices, the complex...
0Identifiers are better off without meaning
2024-05-01 •
The text discusses the inconvenience of having identifiers with semantic value, using examples from Last.fm, Tuenti, and New Relic. It highlights the challenges of changing classification of pages and the problems with semantic id...
0Alert on symptoms, not causes
2024-03-06 •
The text discusses the importance of focusing on symptoms rather than potential causes when defining health alerts for systems. It explains that causes do not guarantee an issue and add the burden of confirmation on the operator, ...
0How about we forget the concept of test types?
2024-02-06 •
The text discusses the concept of test types and its impact on testing software. It argues that the concept of test types does more harm than good, and suggests a systems perspective for analyzing testing pipelines. The author pla...
The text discusses the importance of engineers having a diverse set of skills from other disciplines such as product, project and people management, customer support, HR, finance, and UX. It highlights the potential negative conse...
The article discusses the migration of a large fleet of microservices from EC2 to Kubernetes. It explains the motivation, architectural decisions, outcomes, and future plans. The author also discusses the legacy architecture and t...
The article discusses the challenges and strategies of building a PaaS for 1500 engineers. It emphasizes the importance of balancing standardization and autonomy, the value proposition of platform teams, and the need for a clear a...
0Kubernetes made my latency 10x higher
2019-10-22 •
The post discusses the author's experience with migrating a microservice to Kubernetes and the unexpected increase in latency. The author explains the investigation process, which led to the discovery that the high latency was cau...
The text discusses the challenges of migrating JVM applications to Kubernetes, focusing on resource assignments and capacity planning. It explains the issues faced, the reasons behind OOM kills, and the steps taken to resolve them...
The post analyzes a Hotspot GC log exhibiting large GC pauses leading to allocation pressure and system load. It explains the details of the log and the issues faced, such as frequent Minor GCs, premature promotions, and poor GC t...