How “latency numbers everybody should know” decreased from 1990–2020 (colin-scott.github.io)

Article summary

The article discusses how latency numbers have decreased from 1990 to 2020, with a focus on various technologies such as storage, networking, and CPU performance. The article appears to include an interactive visualization with a slider to compare latency numbers across different years. However, the exact content of the article is not available. The discussion around the article reveals that latency improvements have been significant in some areas, such as storage and networking, but less so in others, like CPU performance.

Main themes

• latency numbers
• performance optimization
• UX design
• Amdahl's Law vs Moore's Law
• parallelism and multicore processing
• storage and networking improvements

What commenters say

• The article's interactive visualization is difficult to use due to poor UX design, with the slider being hard to notice.
• Knowing the exact latency numbers is not as important as understanding their relative values and orders of magnitude.
• The improvement in latency numbers has been significant in some areas, such as storage and networking, but less so in others, like CPU performance.
• The article's title is misleading, as not every programmer needs to know these latency numbers, and they are more relevant to specific use cases like Google's development.
• The importance of Amdahl's Law over Moore's Law is increasing, as performance gains are now more dependent on parallelism than raw processing power.
• Optimizing UI/UX performance requires a different approach than optimizing computational performance, with some arguing that dedicated cores for UI/UX could be beneficial.
• The article's data and projections may be inaccurate or outdated, which could affect the conclusions drawn from it.