Net Defender CloudFlare Goes Down, Taking Many Websites With It

The Internet was built with de-centralized infrastructures in mind. To scale globally, network providers like CloudFlare have emerged, to run decentralized infrastructures and offer them as a service. In general, keeping service independent of each others and maintaining heterogeneous networks have a proven track record of resilience, that is not necessarily inherent to the architecture of these providers. Just like Akamai had a bad day in 2004, CloudFlare today suffered from a global outage, that left many obvious collateral problems visible all over the Internet. Bloomberg, among others, reports:

CloudFlare Inc., an internet service meant to protect websites from going down, faced its own network issues on Tuesday, leading to several prominent sites — like blogging platform Medium and video game chat provider Discord — being unavailable for some time.

Source: Net Defender CloudFlare Goes Down, Taking Many Websites With It – Bloomberg

Behavioral advertising efficiency

Researchers from U Minnesota, UC Irvine and CMU took a look into “behavioural based advertisement”, a segment that requires heavy tracking of users across websites through cookies. A report of their findings is here: Online Tracking and Publishers’ Revenues: An Empirical Analysis. Money quote:

Empirical analysis of behavioral advertising finds that surveillance makes ads only 4% more profitable for media companies

They found that despite the 40% “ad-tech” premium charged by behavioral ad companies, the ads only added about 4% the media companies that published them, meaning that behavioral advertising is a losing proposition.

Source: Boing Boing

Twitter comes back to the Mac

The new Twitter app will be created using Apple’s Project Catalyst, that lets developers port apps from iOS to the recently announced macOS Version “Catalina”.

Google Cloud Is Down

Apparently Google was down the other day. Unfortunately, I was on a train in a really remote location and couldn’t verify what I picked up from Hacker News and tweeted.

Meanwhile, there is a bit of broader coverage on Slashdot and The Verge, that hint at the size of the outage.

Again, an outage like this reminds of the issues that remain with the Cloud, the difference to traditional infrastructure is merely scale at which an outage hits.

Source: Google Cloud Is Down | Hacker News

On Latency

Andrew Certain helps getting an understanding of percentile measurements and why it is important to look at certain things from specific perspectives. I found this useful.

GCP can run untrusted workloads

Google’s Kubernetes Engine (GKE) now supports node pools that are wrapped in gVisor to allow running untrusted workloads. The idea behind gVisors is to emulate all system calls in user space and provide a sandbox to processes that cannot be trusted. GKE allows to enable this with a configuration option now.

GKE on GCP

New GKE Sandbox brings added security to your containers running in Google Kubernetes Engine clusters.

Source: GKE Sandbox: Bring defense in depth to your pods | Google Cloud Blog