May 2023 Summaries

In this blog post, we explored how Oxygen (Oxy), a new open-source framework developed by Cloudflare, enables developers to easily build custom proxies with minimal effort. The key feature of Oxy is its extensible architecture that leverages a concept called "hooks" for customization and integration of various features. Oxy's hooks are designed to be simple yet powerful, allowing developers to control different aspects of the proxy such as connection handling, packet inspection, and more. These hooks can be easily plugged into Oxy to add new features or modify existing ones. Furthermore, Oxy provides a wide range of pre-built plugins that developers can use out-of-the-box for common tasks like firewalls, load balancing, content filtering, etc. By leveraging the power of Oxy's hooks and plugins, developers can quickly build sophisticated proxies without having to worry about low-level networking details. This makes it easier for organizations to deploy customized proxies tailored to their specific needs, whether it be for security purposes, performance optimization, or other use cases. In conclusion, Oxygen (Oxy) is an innovative open-source framework that simplifies the process of building custom proxies by utilizing a modular and extensible architecture based on hooks. With its wide range of pre-built plugins and easy-to-use API, Oxy has the potential to become the go-to tool for developers looking to build robust and scalable proxies.

In this blog post, we discussed how TCP memory limits can be ignored in the Linux kernel and introduced a patch that fixes it. The issue arises when a sender's window is larger than what a receiver can handle due to limited memory resources. We found that this can result in excessive memory usage or even disconnections after reaching the memory limit set by autotuning. We explained how TCP works, including window scaling and its limitations. The Linux kernel patch we wrote enables TCP window shrinking when necessary to prevent memory overuse and ensures correct behavior during a "window full" situation. We also provided test results demonstrating the effectiveness of the patch in addressing these issues. Finally, we shared the results from deploying this patch on the Cloudflare production network, showing improvements in packet drop rates, memory usage, and other metrics. Overall, our findings highlight the importance of enforcing TCP memory limits to ensure optimal performance and reliability for users.

On May 22nd, 2023, Veronica Marin announced an updated version of Cloudflare's Open Source Software Sponsorships program. The new program will be more inclusive and address the needs of the OSS community better than before. Previously, sponsorship was limited to engineering tools; however, now, non-profit open source projects are also eligible. To qualify for this program, a project must operate on a non-profit basis, include a link back to Cloudflare's homepage, and not be intended for event sponsorships. Sponsored projects will receive benefits such as free Cloudflare Pro Plan zone upgrade, access to requested products (subject to approval), and additional resources through community forums and Discord server. Open source project owners can apply by filling out the application form on their website.

Cloudflare held its annual developer week from 17 October to 20 October, featuring a raft of new product features and updates focused on making cloud computing more accessible, efficient, and secure for developers. Key announcements include the integration with popular AI platforms LangChain JS and MosaicML, support for Neon, PlanetScale, and Supabase databases, the general availability of S3 to R2 Super Slurper, new developer experience features such as Quick Edit in Cloudflare Workers, and improvements to existing products. The company also announced partnerships with other database providers including Neon, PlanetScale, and Supabase.

Cloudflare has announced the second cohort of its Workers Launchpad, which aims to help startups leverage its developer platform for building and deploying apps at the edge. The 25 selected startups are tackling a diverse set of real-world problems using Cloudflare's technology, such as AI agents, secure productivity SaaS, mobile app testing, and more.

Global organizations aiming to provide consistent app experiences for users worldwide often turn to Cloudflare's services. In 2021, an upgraded China Network was introduced in partnership with JD Cloud to improve performance for users in China. However, requests to the origin still required travel outside of China. To address this issue, in early 2023, China Express was launched as a suite of connectivity and performance offerings in collaboration with partners like China Mobile International (CMI), CBC Tech, and Niaoyun. One key service is Private Link, an optimized, high-quality circuit for overseas connectivity provided by local partners. This offers more reliable and high-performance connections from China to the global internet. Acme Corp, a global online shopping platform, serves many direct-to-consumer brands primarily through e-commerce channels. Their web performance directly impacts customer engagement and revenue. With 90% of their suppliers based in mainland China, they utilized Cloudflare's China Network to enhance performance and improve suppliers' experience with store backend systems. Despite this, challenges persisted with dynamic content. To resolve these issues, Acme Corp opted for a Private Link service through Cloudflare's local partner CMI. The implementation of China Express resulted in the removal of timeout issues and a significant performance improvement during peak hours. This solution effectively tackles cross-border network congestion and ISP routing challenges faced by global organizations operating within China, leading to enhanced user experience and reduced administrative burdens for IT teams.

In this announcement, we introduced several new features for our Durable Objects Database (D1) alpha. These include SQLite 3.40.0 support, a JSON path syntax for querying JSON data stored in TEXT columns, location hints to influence database creation locations, and more documentation. We also provided details on D1's upcoming pricing model. In addition, we are planning to release an open beta, enable larger databases, improve metrics and observability features, and implement automatic read replication for faster query latency. To learn more about these updates or start building with D1, visit the Cloudflare blog.

1. Cloudflare Workers Queue Service allows developers to create message queues for asynchronous task processing in a serverless environment. It enables high-speed, reliable and cost-effective handling of large volumes of data and tasks that need to be processed independently. 2. Consumer Concurrency feature allows multiple instances of the consumer script to process messages concurrently, significantly improving the throughput and reducing the time taken to complete a batch of tasks. This is particularly useful in scenarios where individual tasks can take longer times to execute. 3. Explicit Acknowledgment feature provides more flexibility for handling task processing failures by enabling developers to explicitly acknowledge or retry individual messages within a batch, rather than re-processing entire batches on errors. It helps reduce message processing latency and costs associated with re-trying failed tasks. 4. Queues service supports higher throughput rates which have quadrupled recently from 100 to over 400 messages per second. This improvement enhances the overall performance of applications built using the service. 5. The planned features for future updates include R2 (Cloudflare's storage bucket) as a consumer, allowing developers to directly write data to an R2 bucket from a Queue without needing any additional code; and HTTP pull functionality, enabling consumers outside of Cloudflare's ecosystem to pull batches of messages for processing. 6. The service is currently in open beta and ready to be used by developers to build scalable applications. It offers a powerful toolset for managing asynchronous task processing workloads efficiently.

In a head-to-head comparison against Amazon S3, Cloudflare's R2 storage service emerged as the faster solution in both North America and Europe/Asia regions. The study tested 400 Catchpoint backbone nodes embedded in last mile ISPs around the world to retrieve a 1 MB file from both services in specific locations: Ashburn, Tokyo, and Frankfurt. Cloudflare's R2 service was found to be approximately two times faster than Amazon S3 for users accessing content hosted in the same region as them. Additionally, R2 outperformed S3 when users accessed data stored outside of their local regions. This performance advantage extends to workloads that combine storage with compute services like Cloudflare Workers. The combination of R2 and Workers provides developers with a simpler and faster end-to-end application experience compared to alternatives offered by AWS or Fastly.

Cloudflare Workers introduces Browser Rendering APIs that allows developers to programmatically control a headless Chrome browser instance running within their Worker. This enables developers to create applications like screenshotting websites, testing web software in CI, or gathering page load performance metrics of their applications with ease. The API provides low-level access to the DevTools Protocol and Puppeteer API allowing for complex manipulations on a live Chrome browser instance within the worker environment. It also supports other Cloudflare products like Durable Objects for reusing browser instances, Queues for handling jobs asynchronously, R2 for storing data, and KV store to cache data. To use Browser Rendering API, developers need to create a new project using wrangler generate, add the Puppeteer library dependency in their package.json file, configure the binding in wrangler.toml configuration file and write the worker code in src/index.ts file. After that, they can test their worker locally by running npx wrangler dev –remote command and publish it to production using npx wrangler publish command. The Workers Browser Rendering API is currently in open beta and will be enabled for customers in the wait list gradually with emails being sent out. Cloudflare also plans to upstream changes in their Puppeteer fork to the main project, support other browser instrumentalization libraries like Playwright, and keep an eye on WebDriver BiDi standard for future integration. Overall, this API provides a powerful toolset for developers to build sophisticated applications with ease using Workers and other Cloudflare products.

Cloudflare has announced support for three additional APIs from Node.js in their Workers platform, increasing compatibility with the existing ecosystem of open-source npm packages. The new APIs include Node.js Streams, PathStringDecoder, and StringDecoder. This expansion will allow developers to use their preferred libraries within Workers even if they depend on APIs from Node.js. Furthermore, a preview version of a new module type that mirrors the Node.js environment more closely has been shared in the open-source Workers runtime. These updates demonstrate Cloudflare's commitment to enhancing the capabilities and utility of their Workers platform for developers.

This post discusses Cloudflare's journey of rearchitecting their entire technical architecture for performance and security products from scratch while ensuring that it does not slow down the pace at which they develop new products. The author explains how they are adding capabilities to their developer platform for internal service development, and migrating existing systems over to this new platform. They also share insights on building a gradual deployment system, progressive health-mediated advancement, and modularity of code for easier reasoning about the system. The post concludes with an invitation to join their Discord server to learn more or provide feedback on their work.

Cloudflare and LangChain have joined forces to create a powerful new way for developers to build intelligent applications. By combining the capabilities of Cloudflare Workers with the language model capabilities of LangChain, developers can now build highly interactive and responsive applications that provide users with relevant information in real-time. One example of how this partnership can be used is by creating an intelligent Q&A system that allows users to ask questions about a specific topic or article and receive accurate answers based on the content of the text. This type of application would make it easy for users to access important information without having to sift through large amounts of text themselves. To build such an application, developers can use LangChain's powerful tools and capabilities, including document loaders, embeddings, vector stores, models, and chains. By using these tools in conjunction with Cloudflare Workers, developers can create highly scalable and efficient applications that can handle large volumes of data and provide fast response times to users. To get started with building your own intelligent application using LangChain and Cloudflare Workers, you can follow the example code provided in this article. This example demonstrates how to build a simple Q&A system that retrieves information from a specific Wikipedia article based on user queries. By leveraging the capabilities of both LangChain and Cloudflare Workers, developers have endless possibilities for building intelligent applications that provide users with valuable insights and information in real-time. So why wait? Start exploring what you can build with this powerful partnership today!

The advent of generative AI brings with it new challenges for security professionals. In this article, we've highlighted ten steps you can take to prepare your APIs and applications for the age of generative AI: 1. Conduct a thorough threat modeling exercise. 2. Implement strong API authentication and authorization measures. 3. Regularly review and update API access controls. 4. Use token-based authentication with short-lived tokens and long-lived refresh tokens. 5. Monitor your APIs for unusual or suspicious activity. 6. Ensure that all data transmitted over the network is encrypted using HTTPS or another secure protocol. 7. Limit the number of API requests allowed per user or IP address to prevent DoS attacks and conserve system resources. 8. Build and validate an API schema to prevent unauthorized access to sensitive data. 9. Analyze the depth and complexity of queries to identify potential security risks. 10. Communicate directly with your users about any security measures you implement and provide clear troubleshooting steps for resolving issues related to these measures. By following these recommendations, you can help protect your APIs and applications from malicious actors while still enabling the innovative use of generative AI technology by legitimate users.

Cloudflare has released three new features for Workers for Platforms: Outbound Workers, Dynamic Dispatch, and Tail Workers. These features provide more visibility, control, and customization options for platform developers using Cloudflare Workers. The company also plans to bring Workers for Platforms down to its pay-as-you-go plan later in the year.

Cloudflare has announced a new secrets management offering called Secrets Store. This service aims to provide a secure and centralized repository for storing sensitive data, such as API tokens, shared secrets, and other critical information. The need for such a solution arises from the growing concern over securing sensitive information within applications and systems. Cloudflare's Secrets Store will focus on ensuring ease of use, performance, reliability, and security to cater to both developers and security administrators. The service will also include measures like access control, auditing, and encryption key rotation for enhanced protection. Customers interested in using the Secrets Store can sign up for the beta program.

Cloudflare has announced an upgrade to its Pages platform, which is designed to simplify the process of building, deploying and hosting web applications. The new update will feature a modular architecture that enables developers to create more scalable builds for their projects. This change is expected to improve the performance and reliability of the platform while also reducing maintenance costs. Additionally, users can now opt-in to try out the updated beta build system with support for new languages and tools. In summary, Cloudflare has upgraded its Pages platform with a modular architecture that provides faster and more scalable builds for developers. This update is expected to improve the overall performance of the platform while also reducing maintenance costs.

The Cloudflare dashboard now supports in-browser development for its Workers, enabling developers to write, edit and preview changes to their workers without having to switch between multiple environments. This feature has been achieved by embedding Microsoft's Visual Studio Code (VSCode) within the Cloudflare dashboard using an iframe. The dashboard communicates with VSCode through a MessageChannel, allowing for seamless data exchange between them. The development process is as follows: When a developer opens a worker in the Cloudflare dashboard, VSCode is loaded into an iframe within the dashboard. The worker's code is then sent to VSCode via a MessagePort. Any changes made by the developer are automatically synced back to the dashboard, which then uploads them as previews of the worker on Cloudflare's global network. This enhancement significantly improves the development workflow for Workers, offering a more integrated and efficient experience. It reduces the need for context-switching between different environments while testing or making changes to workers. The new in-browser editor also has potential future developments planned, such as adding Typescript support and integrating with external Git providers like GitHub and GitLab.

In this post, we discussed how Miniflare v3 was developed to improve upon the previous version's limitations and offer more flexibility for data storage. We also introduced a new approach for running unit tests inside Workers using WebSockets for communication between Node.js host process and workerd "thread". Finally, we mentioned our efforts in porting workerd runtime to support macOS and Windows platforms.

# Cloudflare Workers DevOps improvements announced at developer conference At their annual developer's conference called "Workers Summit", the company announced several updates for its serverless computing platform, Cloudflare Workers. 1. Smart Placement: This feature optimizes where code is run, either on the edge (near the end user), or in a data center, based on factors such as latency and concurrency. The goal is to improve performance by reducing round trip time for network requests. 2. Development improvements: Cloudflare introduced new versions of its command-line tool Wrangler v3, and an upgraded online editor powered by VSCode in the dashboard. These updates provide a more efficient development environment with faster code reloads and improved debugging capabilities. 3. Increased resource limits: The company increased application size limits to 10MB (after gzip) and startup latency limit from 200ms to 400ms. They also announced plans to simplify pricing and increase other resource limits by introducing tiers. 4. Pages CI upgrade: Cloudflare upgraded the build image used for their Pages continuous integration system, allowing developers to use more recent versions of tools like Node.js and pnpm. 5. Framework adapters and C3: To support full-stack web applications, Cloudflare partnered with many web framework authors to create new or improve existing adapters for various popular JavaScript web frameworks. They also released a new sassy CLI called C3 that serves as an entry point into the Cloudflare universe for developers. Overall, these updates aim to make Cloudflare Workers the best place to run full-stack web applications by improving developer productivity and performance optimization capabilities.

Cloudflare Pages and Workers are converging into one singular platform for building serverless applications. This new development experience will bring together static site hosting, serverless functions, and bindings to resources in a single product with shared features like rollbacks, previews, environments, and more. The unified platform is built on the foundation of Cloudflare’s global network and performance capabilities.

"Connect" is a new function in Workers that enables applications built on the platform to connect directly to databases. It provides access to outbound TCP connections, UDP connections. The feature allows for seamless integration between Cloudflare's global network and your application running at one of the many locations worldwide."Connect" is a part of our larger commitment to providing developers with the tools they need to continue innovating building applications that are not just limited to HTTP requests only but also beyond the horizon goes on to describe other aspects related directly or indirectly all around this particular new function in Workers.

During Developer Week 2023, Cloudflare introduced Database Integrations, making it seamless for developers to connect their choice of database with Workers. Initially supporting Neon, PlanetScale, and Supabase, more databases like Prisma, Fauna, and MongoDB Atlas will be added soon. The integration process includes discoverability features such as a new "Integrations" tab in the Worker's settings where users can view supported databases and add them directly to their project. Authentication is also made easy with an OAuth2 flow that automatically sets up encrypted environment variables for database credentials. In addition, Cloudflare is working on an integrations platform for database providers and wrangler CLI support. This feature aims to help developers move faster in building applications by minimizing the time spent on configuration and focusing more on code development.

On May 16, 2023, Cloudflare announced its partnership with Snowflake to create an efficient and cost-effective solution for data storage and analysis. By integrating Cloudflare's R2 object storage platform with Snowflake's Data Cloud, users can now query data stored in R2 data lakes and load data from R2 into Snowflake without worrying about egress fees or vendor lock-in. This partnership aims to help organizations get valuable insights from their data while reducing costs associated with data transfer between different clouds and regions.

Training large language models (LLMs) and diffusion models requires massive infrastructure, including significant storage capacity for terabytes to petabytes of training datasets and model checkpoints. To manage storage costs and scalability, many machine learning teams have been moving to object storage providers like Cloudflare R2. However, these providers often charge high egress fees, making it difficult to leverage GPU capacity across multiple cloud providers or take advantage of lower pricing elsewhere. MosaicML's tools and Cloudflare R2 address these challenges by enabling efficient use of R2 as the durable storage backend for training LLMs on any compute provider with zero egress fees. This allows users to run training workloads on any compute provider, with total freedom and zero switching costs. The combination of MosaicML's platform and Cloudflare R2 provides maximum autonomy and control, allowing organizations to switch between cloud service providers as needed.

On May 16, 2023, Cloudflare announced updates to its Terms of Service to simplify and modernize them for users. The changes address confusion surrounding content-based restrictions, particularly regarding the use of their CDN service. They have clarified that customers can serve video and other large files using the CDN, provided the content is hosted by a Cloudflare service like Stream, Images, or R2. This allows developers to confidently innovate on Cloudflare's Developer Platform while leveraging its benefits. Furthermore, they streamlined their Self-Serve Subscription Agreement and removed an old restriction on benchmarking. These improvements aim to enhance the user experience and encourage innovation on their platform.

Cloudflare has announced Smart Placement for Workers and Pages Functions, a new feature that automatically moves compute resources to optimal locations in order to speed up applications. The feature is currently available in open beta for all Workers and Pages customers. Smart Placement aims to help developers build faster applications by reducing the number of network round-trips made between users and backend services such as databases or APIs, particularly when those services are not located near the end user. This will be achieved through automatic analysis of fetch requests made by Workers in real time, comparing them against latency data aggregated by Cloudflare's network to determine the best location for each request. The goal is to minimize overall request duration and improve application performance.

Cloudflare has announced the general availability of Super Slurper for one-time migration from S3 to their zero egress fee object storage platform, R2. The tool allows fast and reliable data transfer, with users reporting completion in less than half an hour for large migrations. Super Slurper also simplifies the process of copying objects and their associated metadata from S3 to R2 buckets. Additionally, Cloudflare is working on incremental migration capabilities that will allow files to be migrated from the origin bucket to R2 as they are requested, reducing egress fees over time.

Cloudflare has launched an experimental AI assistant called Cursor, designed to answer questions about its Developer Platform. The company aims to use AI to help developers build applications more quickly and efficiently. They envision a generative developer experience with three waves of advancement: wave 1 - AI-generated content; wave 2 - Q&A assistants; wave 3 - generative experiences. Cursor is an early milestone in this journey, aiming to streamline the developer experience by providing answers quickly and allowing developers to navigate Cloudflare's documentation in a conversational way.

As companies increasingly rely on Artificial Intelligence (AI) tools, it is critical that they adopt security measures and best practices to ensure the safe use of these technologies. In this blog post, Cloudflare offers tips for securely using AI services while protecting sensitive data. They recommend using a Zero Trust approach by enabling access controls at the network level, restricting data uploads, and controlling the use of APIs without a proxy.

Cloudflare has announced Constellation, a new platform that allows developers to run machine learning models in edge computing environments. The service is currently available as part of an early access program and aims to bring AI capabilities closer to end-users by executing them at the edge. Constellation runs on WebAssembly (Wasm) and enables developers to deploy machine learning models directly into Cloudflare's global network of over 250 data centers. This approach reduces latency, increases scalability, and improves privacy for users accessing AI-powered applications. The platform supports popular machine learning frameworks such as TensorFlow, PyTorch, ONNX Runtime, and Transformers.js. Developers can use these libraries to create custom models or leverage pre-trained models from repositories like Hugging Face's Model Hub. In addition to deploying models at the edge, Constellation also provides tools for integrating AI capabilities into existing applications. For example, developers can utilize Cloudflare Workers and other products in the company's developer stack to build complete AI-driven workflows that include model training, evaluation, and deployment stages. Some potential use cases for Constellation include image recognition, natural language processing (NLP), speech synthesis, recommendation systems, fraud detection, and predictive analytics. By making it easier for developers to deploy machine learning models closer to end-users, Cloudflare hopes that its new platform will help drive innovation in the field of edge AI. Overall, Constellation represents a significant step forward in bringing advanced AI capabilities to edge computing environments. As more organizations adopt edge strategies to improve performance and reduce costs, platforms like Constellation will play an increasingly important role in enabling developers to harness the power of machine learning at scale.

Cloudflare has developed plugins for OpenAI's ChatGPT, enabling users to access real-time data and insights from the company's products and services directly within the chat interface. Two of these plugins are currently available - Radar ChatGPT plugin and Cloudflare Docs plugin. The Radar plugin allows users to ask questions about Cloudflare's network traffic data, while the Cloudflare Docs plugin provides access to up-to-date information from their developer documentation. These plugins utilize Cloudflare Workers and OpenAI's API for embedding vectors, enabling efficient retrieval of relevant content. The use of plugins extends ChatGPT's capabilities by providing it with fresh insights based on recent data and contextualizes its responses with current best practices and information located within Cloudflare's developer documentation.

The author announces a program update for early-stage startups, expanding the eligibility criteria for their startup plan. Previously, only startups affiliated with an accelerator or having an employee referral could qualify, but now bootstrapped AI startups can apply without these requirements. To get on the startup plan, founders must create a Cloudflare account, add a domain, update name servers, and fill out a form specifying their startup's details. This program aims to provide valuable resources to promising AI startups that may not have access otherwise.

Developer Week has kicked off at Cloudflare with a focus on enhancing developer experience by addressing popular feature requests and refining existing tools. The company aims to enable developers to build innovative applications using AI technology. During the week-long event, new products will be introduced alongside tutorials, tips, and best practices for leveraging AI in development projects.

A 1947 paper introduced the concept of subroutines, a significant step in improving programmer productivity and widening the scope of who could be a programmer. Today's advances such as syntax highlighting, high-level languages, IDEs, libraries and frameworks, APIs, Visual Basic, code completion, refactoring tools, spreadsheets, and LLMs have further enhanced programming capabilities and productivity. The introduction of AI assistants like Cloudflare's new feature will continue this trend, making developers more productive and allowing more people to enter the field of programming. AI's potential is being compared to electricity's impact on work and life, as it has the power to revolutionize how programs are written, who writes them, and how humans interact with software. The future may bring even more exciting ways of programming through neural-type machines.

Pingora-limits is a Rust library for counting events efficiently using a probabilistic data structure called HyperLogLog++. It provides an estimator that can accurately count the occurrence of events with a small probability of overestimation, making it useful in scenarios where false positives are acceptable or can be mitigated by other means. The design is based on multiple hash tables and atomic counters to ensure lock-free and efficient performance. Pingora-limits is currently used in production at Pingora for the connection limit feature.

Between May 9-12, Pakistan experienced mobile internet shutdowns and blocks on several social media platforms following the arrest of former Prime Minister Imran Khan on corruption charges. This led to a significant drop in internet traffic nationwide, with peak levels falling by almost 30% within the first few days of the shutdowns. Mobile device request volume declined by 60%, while desktop traffic remained fairly consistent. The government's directive impacted four major mobile network providers, leading users to shift to fixed broadband networks as a backup. In addition to this, Pakistan's authorities ordered internet service providers to block access to social media platforms like Facebook, Instagram, YouTube, and Twitter. Despite these measures, there was an increase in the use of Cloudflare’s 1.1.1.1 resolver by users seeking ways to maintain access to the open internet. The shutdowns have had a significant impact on "gig workers", freelancers, and point-of-sale terminals that rely on mobile connectivity.

ChatGPT plugins are a powerful way to extend the capabilities of OpenAI’s conversational AI model, allowing users to access information and perform actions from a variety of external data sources. With Cloudflare Workers, developers can easily build, deploy, and scale custom plugins that connect with any API, database, or other data source. To help you get started building your own ChatGPT plugins, we've developed the ChatGPT Plugin Quickstarts – a collection of sample plugins and resources designed to streamline the development process and make it easier than ever to create custom plugins for ChatGPT. The quickstarts cover everything from setting up your plugin’s environment to defining its schema and implementing its endpoint logic, all with just a few lines of code. They also automatically generate an OpenAPI schema that helps ChatGPT understand how your code works, so you can focus on crafting custom behaviors, endpoints, and features for your plugins without getting caught up in the nitty-gritty. To demonstrate the capabilities of our quickstarts, we've created two example plugins: one that connects ChatGPT with the GitHub Repositories Search API, and another that uses the Pirate Weather API to retrieve up-to-date weather information. Both examples showcase how easy it is to create custom plugins that go beyond basic querying, providing more insightful and context-aware responses. With our ChatGPT Plugin Quickstarts, you can easily build, deploy, and scale custom ChatGPT plugins that connect with any API, database, or other data source, unlocking a world of endless possibilities for conversational AI applications. We're excited to provide this Quickstart, and would love to see what you build with it. Join us in our Discord community to share what you're working on!

The Linux kernel has its own crypto API that can be used to perform encryption and decryption operations directly within the kernel space. This is useful for applications requiring high performance and low latency, as it minimizes the number of expensive context switches between user space and kernel space. In this post, we reviewed the Linux Kernel Crypto API and its user space interface. We reiterated some security benefits of doing encryption through the Kernel vs using some sort of cryptographic library. We also measured the performance overhead of doing data encryption/decryption through the Kernel Crypto API, confirmed that in-kernel crypto is likely as good as in OpenSSL, but a better user space interface is needed to make Kernel Crypto API as fast as using a cryptographic library. We implemented AES-CTR with both the kernel's crypto API and OpenSSL for comparison. We found that while the Kernel Crypto API is slower than OpenSSL, the difference can be attributed largely to system call overhead rather than the actual encryption process itself. Using Crypto API is a subjective decision depending on your circumstances, it’s a trade-off in speed vs. security.

The coronation of King Charles III and Queen Camilla impacted Internet traffic patterns in the UK, with significant spikes and dips coinciding with key moments of the ceremony. On May 6, 2023, during the Coronation Procession, there was a clear drop in traffic by as much as 7%. The following day, during the Coronation Big Lunch event, Internet traffic dropped by up to 18%. Traffic patterns varied throughout the weekend, with significant drops observed during Prince William's speech and increases during the military flypast. Overall, daily traffic was 4% higher on Saturday compared to the previous week but dropped by 6% on Sunday. On Monday, a bank holiday in honor of the coronation, Internet traffic was noticeably higher than usual.

On May 10th, Cloudflare introduced Object Lifecycle Management for its R2 storage service, allowing developers to effectively manage object expiration and reduce costs through the use of rules. The feature is now generally available, with users able to define up to 1,000 rules from the R2 dashboard or via API. This announcement follows the launch of R2 last year, which provided developers with object storage while eliminating egregious fees. Tens of thousands of developers have utilized R2 since its debut for various applications. Object lifecycles enable users to define rules that determine how long objects uploaded to their bucket are kept. These rules could include deleting objects after a certain period or aborting unfinished multipart uploads. To implement object lifecycle management in R2, users can follow the steps provided in the Cloudflare dashboard or use the S3-compatible API with examples such as the one demonstrated using AWS SDK for JavaScript. As part of its ongoing commitment to improve R2, Cloudflare is also working on new, lower-cost ways to store infrequently accessed objects like long tail user-generated content and archive data. Developers interested in providing feedback and gaining early access can join the waitlist. Furthermore, developers are encouraged to share their experiences and provide feedback on R2 by joining Cloudflare's Discord community or filling out a documentation survey.

The introduction of network usage fees in Europe could negatively impact the internet experience for consumers and small businesses, according to a report by Cloudflare. The company argues that such fees would discourage content delivery networks (CDNs) from directly interconnecting with terminating internet service providers (ISPs), leading to worse performance for users. Instead of paid peering, Cloudflare recommends more settlement-free peering between CDNs and ISPs as it improves latency and reduces traffic congestion at regional internet hubs.

Cloudflare has the fastest Time to First Byte when compared with Zscaler and Netskope for connections from Central and South America to applications hosted in Brazil and the United States (South West Region). Specifically, Cloudflare's 95th percentile time to first byte times are up to 85% faster than Zscaler and up to 74% faster than Netskope.

In this project, a team of engineers rebuilt their network analytics system called Network Analytics from scratch, using ClickHouse and Apache Flink for data storage and processing respectively. The new system is highly scalable and fast, enabling real-time analysis of billions of network traffic events per day. The previous implementation was built on SQLite and Apache Kafka but suffered from performance issues as the amount of data increased. To address this issue, the team chose ClickHouse for its exceptional query speed, high availability features, and ability to handle large datasets efficiently. To implement real-time stream processing, the engineers used Apache Flink due to its fault tolerance capabilities, event time semantics support, and easy integration with other systems. They also implemented AggregatingMergeTree tables in ClickHouse for aggregated data analysis and pre-computation of minimum or maximum values using a single pass over the data. As a result of these changes, Network Analytics can now accurately identify different types of network attacks such as DDoS attacks with fixed destination IPs or ports, even if they span multiple days. The system has been deployed in production and is currently handling hundreds of terabytes of data per day across thousands of servers worldwide. In summary, the rebuilt Network Analytics system leverages ClickHouse for efficient storage and querying of large datasets, while Apache Flink enables real-time stream processing with fault tolerance and event time semantics support. This combination allows for accurate identification and analysis of various types of network attacks in near real-time.

An armed conflict between rival factions of Sudan's military government began on April 15, resulting in a significant disruption to Internet traffic. On May 2, internet traffic was still approximately 30% lower than pre-conflict levels. Major local internet providers experienced outages and disruptions during this time, with some ISPs reporting a nearly 90% drop in traffic. The most impacted ISPs were Sudatel, Zain, and Canar. Border Gateway Protocol (BGP) updates indicated consistent spikes with both the start of the outage and its return to normalcy. As of this writing, internet traffic continues to be much lower than before, showing some recovery in the past couple of days.

Cloudflare's focus on data protection, security measures such as Web Application Firewall (WAF), bot management, and encryption in transit are critical for protecting personal data globally. The company’s commitment to privacy is reflected in employee education, development of privacy-focused products, certifications, and response to government requests. In the future, Cloudflare plans to support Australia's privacy goals, including updates to its laws and regulatory structure.