LIVE · cybersecurity feed
Live wire
In Other News: Ransomware Developer Sentenced, Plugin4Shell AI Attack, Critical SAP FlawCisco alerts customers to second actively exploited zero-day in as many daysCisco warns of max severity ISE zero-day exploited in attacksCVE-2026-89026 · Attackers Exploit Issabel Framework Flaw Enabling Unauthenticated OS Command ExecutionCVE-2026-58704 · Google Patches Pixel Modem Flaw Amid Signs of Limited Targeted ExploitationAcronis warns of actively exploited flaw in its cPanel backup pluginOracle September 2026 Critical Security Patch Update addresses 672 CVEsCVE-2026-76461 · U.S. CISA adds Cisco Secure Email Gateway flaw to its Known Exploited Vulnerabilities catalogHackers target WordPress sites via third-party WooCommerce pluginCVE-2026-51990 · Hackers exploit Tencent app flaw to deploy GrayRabbit malware
cloud

Saving another 100TB of RAM with math (and Rust)

Cloudflare's global network is immense but not limitless. As we look for small ways to trim our resource usage, we sometimes get lucky and we can cut significantly more. Here’s how we reduced one of our Pingora-based service's RAM usage with statistics.

zeroday.news ·

Cloudflare has announced that it has successfully optimized its Pingora proxy, which is written in Rust, to save approximately 100 terabytes of RAM across its global network. This significant memory reduction was achieved through a series of optimizations, primarily focusing on how Pingora handles HTTP headers.

The core of the optimization involved a re-evaluation of how HTTP headers are stored and processed within Pingora. Initially, Pingora would parse incoming HTTP headers and store them in a `HeaderMap` structure, which is a hash map. While functional, this approach proved to be memory-intensive, especially given the high volume of traffic processed by Cloudflare's edge network. The `HeaderMap` would allocate memory for each header key-value pair, leading to considerable overhead.

To address this, Cloudflare engineers implemented a strategy that leverages the fact that many HTTP headers are common and frequently repeated across requests. Instead of storing full copies of these common headers for every request, Pingora now employs a system where it can reference a canonical, pre-allocated copy of these headers. This "deduplication" process significantly reduces the memory footprint by avoiding redundant storage.

Further memory savings were realized by optimizing the parsing process itself. The new approach minimizes temporary allocations during header parsing, ensuring that memory is used more efficiently from the outset. This involved fine-tuning the Rust code to reduce intermediate data structures and directly process header bytes into their final, optimized representation.

The impact of these changes is substantial. By reducing the memory consumption of each Pingora instance, Cloudflare can now handle a greater number of concurrent connections and requests with the same amount of hardware. This not only translates to direct cost savings in terms of RAM but also improves the overall efficiency and performance of their network infrastructure.

Pingora, Cloudflare's custom-built proxy, replaced its previous proxy infrastructure, which included NGINX. The decision to develop Pingora in Rust was driven by the language's strong performance characteristics, memory safety guarantees, and concurrency features, which are critical for a high-performance network proxy. This latest optimization further validates the choice of Rust for such demanding applications.

The 100TB RAM saving represents a cumulative figure across Cloudflare's entire global network of data centers, where Pingora is deployed at the edge to handle incoming web traffic. This optimization is part of Cloudflare's ongoing efforts to enhance the efficiency and scalability of its infrastructure, ensuring it can continue to support the growing demands of the internet.

cloud
ShareXLinkedInWhatsAppFacebook

More News

view all →
vulnerability

Microsoft Patches 18 Vulnerabilities in AI, Cloud Products

Microsoft fixed vulnerabilities across Azure and AI-branded products, with privilege escalation flaws accounting for the majority. The post Microsoft Patches 18 Vulnerabilities in AI, Cloud Products appeared first on SecurityWeek.

vulnerability

Researchers use AI to find widespread software decoder flaw

The bug, since patched, gave attackers remote code execution privileges and access to user accounts and production environments, including Meta’s core product suite and an OpenAI software repository. The post Researchers use AI to find widespread software decoder flaw appeared first on CyberScoop.

ai

Researchers used Claude to hack OpenAI employees' ChatGPT accounts

Agentic exploits for the win (again)

nation-state

Nations take action on North Korean IT workers after UN report

A report published Wednesday said that as of July, Vietnam, Laos, Pakistan and Argentina took meaningful steps to respond to allegations involving North Korea listed in an October study.

ransomwarecritical

In Other News: Ransomware Developer Sentenced, Plugin4Shell AI Attack, Critical SAP Flaw

Noteworthy stories that might have slipped under the radar: Mandiant's 2026 AI risk report, PhantomRaven malware used by bug bounty hunter, WordPress plugin bug exploited. The post In Other News: Ransomware Developer Sentenced, Plugin4Shell AI Attack, Critical SAP Flaw appeared first on SecurityWeek.

ai

Did an AI really try to break free from human control?

An unreleased OpenAI model wrote instructions telling itself to ignore developer controls. Here’s what actually happened.