DEVOPS WEEKLY ISSUE #663 - 10th September 2023

Reliability, secure AI, service mesh adoption and prompt engineering testing tools amongst other hopefully interesting topics this week.And if you’re in London, make sure to head to Devopsdays London in a few short weeks time.

StackHawk sponsors Devops Weekly

Bridge the gap between security and development teams by sharing valuable context on common web application threats. Check out 5 common threats to address together here:
https://sthwk.com/common-web-threats

News

A talk from the GoTo conference, all about real world reliability and tools like SLOs that can be applied to improve things.
https://www.youtube.com/watch?v=es3DNd5Qipg

The adoption of new technologies and techniques changes the thread model for an organisation. With AI, that change is both broad and happening very quickly. SAIF is a framework for helping adopt best practices as we know them today to secure AI usage and development.
https://blog.google/technology/safety-security/introducing-googles-secure-ai-framework/

A detailed case study looking at the adoption of service mesh at scale, including on-demand cluster discovery and what it took to migrate a large existing system.
https://netflixtechblog.com/zero-configuration-service-mesh-with-on-demand-cluster-discovery-ac6483b52a51

Understanding edge networking is increasingly useful if you find yourself scaling up a system or working on something with lots of users and traffic. A good technical introduction post about packet flow.
https://slack.engineering/traffic-101-packets-mostly-flow/

An attempt to demystify local balancers, reverse proxies and API gateways via analogy.
https://dev.to/aws-builders/load-balancer-reverse-proxy-and-api-gateway-analogies-to-real-life-scenarios-54el

Events

Devopsdays London is coming up soon, on September 21st and 22nd. The usual mix of talks and open spaces, covering the cost of observability, psychological safety, infrastructure migration and more.
https://devopsdays.org/events/2023-london/welcome/

Tools

Promtfoo is an interesting looking tool for introducing testing into prompt engineering. Useful for testing alternative models, or avoiding regressions as models evolve.
https://github.com/promptfoo/promptfoo
https://www.promptfoo.dev/

Triton is a language and compiler for writing highly efficient custom Deep-Learning primitives. The aim of Triton is to provide an open-source environment to write fast code at higher productivity than CUDA, but also with higher flexibility than other existing DSLs.
https://openai.com/research/triton
https://triton-lang.org/main/index.html
https://github.com/openai/triton