Here at Cloudflare we're constantly working on improving our service. Our engineers are looking at hundreds of parameters of our traffic, making sure that we get better all the time.
One of the core numbers we keep a close eye on is HTTP request latency, which is important for many of our products. We regard latency spikes as bugs to be fixed. One example is the 2017 story of "Why does one NGINX worker take all the load?", where we optimized our TCP Accept queues to improve overall latency of TCP sockets waiting for accept().
Performance tuning is a holistic endeavor, and we monitor and continuously improve a range of other performance metrics as well, including throughput. Sometimes, tradeoffs have to be made. Such a case occurred in 2015, when a latency spike was discovered in our processing of HTTP requests. The solution at the time was to set tcp_rmem to 4 MiB, which minimizes the amount of time the kernel spends on TCP collapse processing. It was this collapse processing that was causing the latency spikes. Later in this post we discuss TCP collapse processing in more detail.
The tradeoff is that using a low value for Continue reading
This post originally appeared on the Packet Pushers’ now-defunct Ignition site on October 1, 2019. Insurance companies that offer cyberinsurance policies are looking at ways to reduce their risk (and improve profit margins) by discounting for companies that deploy reviewed and approved technologies. Company executives will make decisions about the cost and value of […]
The post Analysis: Will Your Security Infrastructure Be Determined By Your Cyberinsurance? appeared first on Packet Pushers.
Cloud-native applications running on Kubernetes rely on container network plugins to establish workload communication. While Azure Kubernetes Service (AKS) provides several supported networking options (kubenet and Azure CNI) that address the needs of most deployments, Microsoft recently introduced the ability to bring your own networking solution, called BYOCNI, to help users address more advanced networking requirements. This new feature enables AKS customers to run Calico networking on AKS.
This blog will walk you through some exciting capabilities you can unlock with Calico running in your AKS deployments.
Calico is the most widely adopted container networking and security solution for Kubernetes. Powering more than 100M containers across 2M+ nodes in 166 countries, Calico is supported across all major cloud providers and Kubernetes distributions. Calico gives you a choice of data planes, including eBPF, standard Linux networking, and Windows HNS-based workloads running in public clouds and/or on-prem, on a single node, or across a multi-thousand-node cluster. Whether you need to scale to thousands of microservices with eBPF, or add Windows workloads to your Kubernetes deployments, Calico has you covered.
Calico’s core design principles leverage cloud-native design best practices, combined with proven, standards-based network protocols trusted by Continue reading
This post originally appeared on the Packet Pushers’ Ignition site on January 14, 2020. There is a slow but steady trend for Governements’ to take back control of internet in their countries. For China the “great firewall” is now a rigid access control on content. Russia has been progressing changes to to be isolate itself […]
The post Reading: The Case for a Mostly Open Internet appeared first on Packet Pushers.
This lesson walks through the basics of reaching an application running in a Kubernetes pod. Instructor Michael Levan brings his background in system administration, software development, and DevOps to this series. He has Kubernetes experience as both a developer and infrastructure engineer. He’s also a consultant and Pluralsight author, and host of the “Kubernetes Unpacked” […]
The post Kubernetes For Network Engineers: Lesson 2 – Services, Nodeports, And Load Balancers – Video appeared first on Packet Pushers.
In this IPv6 Buzz episode we talk about the benefit of IPv6 connectivity when IPv4 fails. We examine the types of IPv4 failures, how IPv6 behaves during IPv4 failure, application dependencies, and more.
The post IPv6 Buzz 104: IPv6 For Redundancy When IPv4 Fails appeared first on Packet Pushers.
Managed Transforms is the next step on a journey to make HTTP header modification a trivial task for our customers. In early 2021 the only way for Cloudflare customers to modify HTTP headers was by writing a Cloudflare Worker. We heard from numerous customers who wanted a simpler way.
In June 2021 we introduced Transform Rules, giving customers a simple UI letting them specify what the custom HTTP header’s name and value is—either a static string (i.e. X-My-CDN: Cloudflare
) or a dynamically populated value (i.e. X-Bot-Score: cf.bot_management.score
).
This made the job much simpler, however there is still a good amount of thought required—with a number of potential drop-off points on the user journey. For example, in order to dynamically populate the bot score into the value of an HTTP request header, the user needs to know the correct field name. To find that they'll need to go to the documentation site, find the correct section, etc.
When we analyzed how our customers use Transform Rules we found a set of very common use cases in the data. Four of the top eight fields used were relating to bot management; customers wanting to have the Continue reading
The post Learning and development initiatives at Noction appeared first on Noction.
AWS Identity and Access Management (IAM) provides access control for AWS users/applications to the various AWS services and resources. Features Available only in the Global region Globally resiliant service Always free service Account Types There are a number of account...continue reading
With this post, we kick off a new series based on Juniper Apstra. This post serves as an introduction to Apstra - we’ll look at what a true IBN system is, how relational and graph databases are different (and why graph databases are ideal for network infrastructure), concluding with some general workflows in Apstra.