{
  "id": 8357428,
  "title": "Saving another 100TB of RAM with math (and Rust)",
  "url": "https://urgent.news/2026/09/19/saving-another-100tb-of-ram-with-math-and-rust",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-09-19T00:28:07.000Z",
  "source": {
    "name": "Lobsters",
    "slug": "lobsters",
    "url": "https://blog.cloudflare.com/saving-100-tb-of-ram-with-math/"
  },
  "original_language": "en",
  "account": "In this post, Cloudflare engineers detail how they improved the memory efficiency of a service called Pingora Backend Router (PBR) by leveraging mathematical concepts and the Rust programming language. The issue arose from excessive memory usage in PBR, attributed to structures associated with pingora-ketama, an open-source library for consistent hashing.\n\nConsistent hashing is a method used to distribute tasks across multiple servers, ensuring that adding or removing servers doesn't significantly alter the distribution. It maps tasks and servers to a number line, with each server responsible for a range of hashes. The goal is to distribute tasks evenly among servers, with each server handling a proportionate share of requests based on the size of its range.\n\nHowever, the current implementation of consistent hashing resulted in uneven distribution, with some servers handling significantly more requests than others. To address this, the team aimed to minimize the coefficient of variation—the ratio of the standard deviation to the expected value—of the ranges each server handles. By increasing the number of hashes used to represent each server, they could reduce the standard deviation, leading to a more balanced workload distribution.\n\nThe engineers utilized mathematical formulas to calculate the expected range and standard deviation of the server ranges, ultimately demonstrating that adding multiple hashes per server led to a significant reduction in memory usage, reclaiming over 100TB of RAM globally. This optimization not only improved resource utilization but also showcased the power of applying mathematical principles and modern programming languages like Rust to tackle complex engineering challenges at scale.",
  "summary": null,
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 2,
    "also_reported_by": [
      {
        "outlet": "Hacker News",
        "title": "Saving another 100TB of RAM with math (and Rust)",
        "url": "https://urgent.news/2026/09/18/saving-another-100tb-of-ram-with-math-and-rust",
        "published": "2026-09-18T18:51:46.000Z"
      }
    ]
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}