974 articles on Performance.
We just made Workers AI inference faster with speculative decoding & prefix caching. Use our new batch inference for handling large request volumes seamlessly. Build tailored AI apps with more LoRA options. Lastly, new models and a refreshed dashboard round out this Developer Week update for Workers AI.
Explore the iterative development journey of GitHub’s sub-issues feature. Learn how we leveraged sub-issues to build and refine sub-issues, breaking down larger tasks into smaller, manageable ones.
We re-architected Super Slurper from the ground up using our Developer Platform — leveraging Cloudflare Workers, Durable Objects, and Queues — and improved transfer speeds by up to 5x.
Llama 4 Scout 17B Instruct is now available on Workers AI: use this multimodal, Mixture of Experts AI model on Cloudflare's serverless AI platform to build next-gen AI applications. Build products that can understand images & text, accept large context, with high performance on model quality and inference latency.
Cloudflare’s commitment to the CISA pledge reflects our dedication to transparency and accountability to our customers. This blog post outlines how we deliver newly patched kernels across our infrastructure in order to keep Cloudflare secure.
HTML `` is the de facto element we turn to for embedding video content, but it comes with constraints. For example, it downloads the video file linearly over HTTP, which leads to performance hiccups, especially for large videos consumed on slower connections. But with adaptive bitrate streaming, we can split the video into multiple segments at different bitrates and resolutions.
It's becoming increasingly difficult to tell the difference between web content produced by humans and web content produced by AI. In this post, we're going to look at a promising new approach to making AI content distinguishable as such without impacting the performance of the model.
Some critics question the agnostic nature of Web Components, with some even arguing that they are not real components. Gabriel Shoyomboa explores this topic in-depth, comparing Web Components and framework components, highlighting their strengths and trade-offs, and evaluating their performance.
In February, we experienced two incidents that resulted in degraded performance across GitHub services.
Struggling with slow Largest Contentful Paint (LCP)? Newly introduced by Google, LCP subparts help you pinpoint where page load delays come from. Now, in the Chrome UX Report, this data provides real visitor insights to speed up your site and boost rankings. Matt Zeunert unpacks what LCP subparts are, what they mean for your website speed, and how you can measure them.
Modern frameworks are supposed to help speed up development while providing modern tools and a developer-friendly workflow. In reality, Kevin Leary has found that they cause far more problems than they solve. This ultimately leads to the big question: why are modern theme frameworks so popular, and do they really benefit developers in the long run?
See how Fluid combines server efficiency and serverless flexibility by reusing compute before creating new resources, reducing cold starts, and running in unconstrained environments while staying secure and fast. Cut compute costs by up to 85%.
In January, we experienced two incidents that resulted in degraded performance across GitHub services.
Optimizing web performance means looking beyond surface-level metrics. Time to First Byte (TTFB) is crucial, but improving it requires more than tweaking server response time. Matt Zeunert breaks down what TTFB is, what causes its poor score, and why reducing server response time alone isn’t enough for optimization and often won’t be the most impactful change you can make to your website.
We've optimized how ISR cache updates are managed on Vercel, making them faster and more cost-efficient.
We created a usage-based billing system with ultrahigh throughput; reliable, real-time events processing; and support for complex pricing models. Doing so required reimagining how to build a highly scaled, highly reliable event streaming platform.
With our slice monitoring program, we use machine learning to spot small pockets of performance declines amid the sea of global payments.
We’re sharing details about Strobelight, Meta’s profiling orchestrator. Strobelight combines several technologies, many open source, into a single service that helps engineers at Meta improve efficiency and utilization across our fleet. Using Strobelight, we’ve seen significant efficiency wins, including one that has resulted in an estimated 15,000 servers’ worth of annual capacity savings. Strobe
Shouldn’t there be a way to keep your apps or project data private and improve performance by reducing server latency? This is what on-device AI is designed to solve. It handles AI processing locally, right on your device, without connecting to the internet and sending data to the cloud. In this article, Joas Pambou explains what on-device AI is, why it’s important, the tools to build this type of
In December, we experienced two incidents that resulted in degraded performance across GitHub services.
I wrote a post for Smashing Magazine that was published today about this thing that Chrome and Safari have called "Tight Mode" and how it impacts page
We know that browsers do all sorts of different things under the hood. One of those things is the way they not only *fetch* resources like images and scripts from the server but how they [prioritize those resources](https://www.debugbear.com/blog/request-priorities?utm_campaign=sm-7). Chrome and Safari have implemented a “Tight Mode” that constrains which resources are loaded and in what order, bu
Product drops and sales are a great way to increase revenue, but these events can result in traffic spikes that affect a site’s availability and performance. To prevent website crashes, you’ll have to make sure that the sites you design can handle large numbers of server requests at once. Let’s discuss how!
In November, we experienced one incident that resulted in degraded performance across GitHub services.
Design tokens are design decisions as data and serve as a single source of truth for design and engineering. Utilizing deployment pipelines, they enable automated code generation across platforms, allowing for faster updates and improved consistency in design. Organizing tokens in layers—progressing from available options to tokens that capture how they are applied—ensures scalability and a better
See what we've learned from hundreds of real-world audits—what to look for and what you can do to improve your applications.
What are Slack Audit Logs? Like many Software as a Service (SaaS) offerings, Slack provides audit logs to Enterprise Grid customers that record when entities take an action on the platform. For example, when a user logs in, when a user updates their profile, when an app downloads a file, etc. The actual list of…
The Performance Engineering team at GitHub assessed how CPU performance degrades as utilization increases and how this relates to capacity.
Announcing the acquisition of Grep to further our mission of helping developers work and ship faster.
In October, we experienced one incident that resulted in degraded performance across GitHub services.
Feeling good with your Lighthouse score of 100%? You should! But you should also know that you’re only looking at part of the performance picture. Learn how Lighthouse scores are measured differently than other tools, the impact that has on measuring performance metrics, and why you need real-user monitoring for a complete picture.
Post-acquisition, we migrated Baselime from AWS to the Cloudflare Developer Platform and in the process, we improved query times, simplified data ingestion, and now handle far more events, all while cutting costs. Here’s how we built a modern, high-performing observability platform on Cloudflare’s network.
Every year for Black Friday Cyber Monday (BFCM), we put together a real-time visualization of purchases made through Shopify-powered merchants worldwide. This year we're cooking up something big, and in anticipation we wanted to show you how we built the globe last year. Video of BFCM 2023 Globe Besides going for better visuals, a big focus was performance. We've put together an interactive deep d
Learn how we built Cloudflare Queues using our own Developer Platform and how it evolved to a geographically-distributed, horizontally-scalable architecture built on Durable Objects. Our new architecture supports over 10x more throughput and over 3x lower latency compared to the previous version.
Hono is a web framework that is fast, lightweight, and built using the Web Standards API. Hear the story of Hono by the creator of Hono.
Learn how eval-driven development helps you build better AI faster. Discover a new testing paradigm for AI-native development and unlock continuous improvement.
Cloudflare’s Gen 12 server is the most powerful and power efficient server that we have deployed to date. Through sensitivity analysis, we found that Cloudflare workloads continue to scale with higher core count and higher CPU frequency, as well as achieving a significant boost in performance with larger L3 cache per core.
This article describes the problem of endless trigger recursion in PostgreSQL and shows how to deal with it and get good performance.
With special thanks to our stunning colleagues Amer Ather, Itay Dafna, Luca Pozzi, Matheus Leão, and Ye Ji. At Netflix, the Analytics and Developer Experience organization, part of the Data Platform…
Link’s newest feature, Instant Bank Payments, is priced at 2.6% + 30¢ per successful charge (a discount to cards), allowing you to benefit from the cost savings of bank payments while getting the speed of card transactions.
Our months-long investigation into slow search speeds led to a solution that not only improved performance but also set the foundation for future scale.
In September, we experienced three incidents that resulted in degraded performance across GitHub services.
Great thermal solutions play a crucial role in hardware reliability and performance. Gen 12 servers have implemented an exhaustive thermal analysis to ensure optimal operations within a wide variety of temperature conditions and use cases. By implementing new design and control features for improved power efficiency on the compute nodes we also enabled the support of powerful accelerators to serve
To celebrate Builder Day 2024, we’re shipping 18 updates inspired by direct feedback from developers building on Cloudflare. This includes new capabilities, like running evals with AI Gateway, beta products like Queues and Vectorize graduating to GA, massive speed improvements to KV and Workers AI, pricing simplifications, and bringing key features from Pages into Workers.
Today, we’re announcing up to 3x faster KV hot reads, with all KV operations faster by up to 20ms. And we want to pull back the curtain and show you how we did it.
With a new generation of data center accelerator hardware and using optimization techniques such as KV cache compression and speculative decoding, we’ve made large language model (LLM) inference lightning-fast on the Cloudflare Workers AI platform.
Since June 2021, we’ve been measuring and ranking our network performance against the top global networks in the world. We use this data to improve our performance, and to share the results of those initiatives. In this post, we’re going to share with you how network performance has changed since our last post in March 2024, and discuss the tools and processes we are using to assess network perfor
Explore how Incremental Static Regeneration (ISR) enhances your caching strategy. Learn its benefits, implementation across frameworks, and real-world applications. Optimize performance with this hybrid approach to dynamic content delivery.
In August, we experienced one incident that resulted in degraded performance across GitHub services.
Explore how Arm’s optimized performance and cost-efficient architecture, coupled with PyTorch, can enhance machine learning operations, from model training to deployment and learn how to leverage CI/CD for machine learning workflows, while reducing time, cost, and errors in the process.
Pingora handles 35M+ requests per second, so saving a few microseconds per request can translate to thousands of dollars saved on computing costs. In this post, we share how we freed up over 500 CPU cores by optimizing one function and announce trie-hard, the open source crate that we created to do it.
Google Analytics is often on a “need to know” basis, but why not flip the script? Paul Scanlon shares how he wrote a GitHub Action that queries Google Analytics to automatically generate and post a top ten page views report to Slack, making it incredibly easy to track page performance and share insights with your team.
You ever find yourself in bumper-to-bumper traffic? I did this morning on the way to work (read: whatever cafe I fancy). There's a pattern to it, right? Stop,
Using Git in the CLI can improve your development speed and power. Here are our top eight commands for using GitHub via your command line.
Ever felt pressure to speed up your design process? Here’s how to address unrealistic expectations and foster a shared understanding with stakeholders, ensuring everyone is aligned on the path to a successful delivery. Part of [Smart Interface Design Patterns](https://smart-interface-design-patterns.com) by yours truly.
