The Temporary End of Moore’s Law and the Revival of “Systems Performance”

The Chinese edition of "Systems Performance: Enterprise and the Cloud" (second edition) brings updated performance testing methods and tool recommendations from Brandon Gregg, a veteran of Solaris, Netflix, and Intel, while also incorporating the latest advances in eBPF and other observability technologies.

As transistor scaling reaches physical limits, the industry is shifting focus from raw compute growth to efficient use of existing resources, prompting developers to consider architectural, algorithmic, and data‑structural choices from a systems‑level perspective.

The article introduces the concept of the "golden 60 seconds," a rapid set of metrics—average load, context‑switch frequency, I/O, and network performance—that reveal recent resource utilization and help identify bottlenecks before they cause severe degradation.

Because such analysis traditionally relies on deep expert knowledge, the piece advocates for a "continuous performance watch" approach, embedding performance indicators into the development lifecycle much like test coverage or bug reports, thereby enabling proactive detection of regressions.

It recommends building a performance metric repository and tracking trends across product versions to avoid costly late‑stage optimizations and to inform hardware procurement decisions.

The rise of performance‑expert systems and knowledge‑graph‑based analysis methods is highlighted as a way to democratize expertise, allowing less‑experienced teams to benefit from accumulated best‑practice pathways.

Finally, the article envisions an integrated development environment for performance engineering that combines observability tools, automated alerts, and actionable recommendations, representing the next stage of non‑functional system engineering.