OpenEBS’ Post

Just published a guide on the @openai docs covering seven principles to reduce the latency of an llm-based application – go check it out!

Here’s a super helpful post on “Optimizing LLM’s for higher efficiency”!

LLMs (especially GPT4 and Gemini Pro) get expensive and are usually slow but my friend (Ilan Bigio) at OpenAI put together some thoughtful tips on how to reduce both. My favorite is the seventh bullet; you might not need an LLM. Usually it’s complete overkill to use a model such as GPT4 to parse something like semi-structured data. It would similar to hiring an individual human to manually switch each traffic light on and off. Give it a read and let me know what you think!

Not only why but how to apply is crucial when dealing with AI Exploration of optimizing API latency on OpenAI's platform not only enhances performance but also reduces computing power and energy use, supporting sustainability and lessening environmental impact. #Strive #ZaiStrive #سعي #زكاة_علم

The full recording of last week's Service Mesh Academy workshop with Flynn and Tom Dean is now live! Watch Linkerd Major Feature Deep Dive: Mesh Expansion, HAZL, and Native Sidecar Support below. Every Linkerd major release brings major new features, and in this Service Mesh Academy, Flynn fires up Buoyant Enterprise for Linkerd 2.15 and takes it for a spin, focusing on three new advanced features: 🔹 Mesh expansion lets you add security, reliability, and observability to non-Kubernetes workloads, a new feature of Linkerd 2.15 🔹 The High Availability Zonal Load balancer (HAZL) is a new load balancer that can eliminate expensive cross-zone Kubernetes traffic without sacrificing overall reliability, even for multi-cluster traffic. 🔹 Support for Kubernetes' native sidecar containers eases some of the long-standing annoyances of the sidecar model in Kubernetes Watch this to get your hands dirty and put all three of these new features to work!  #linkerd #servicemesh #opensource #cloudnative

LIVE TOMORROW on Service Mesh Academy: Linkerd major feature deep dive: Mesh expansion, HAZL, and native sidecar support. Register today! Every Linkerd major release brings major new features, and in this Service Mesh Academy, we’ll fire up Buoyant Enterprise for Linkerd 2.15 and take it for a spin, focusing on three new advanced features: 🔹 Mesh expansion lets you add security, reliability, and observability to non-Kubernetes workloads, a new feature of Linkerd 2.15. 🔹 The High Availability Zonal Load balancer (HAZL) is a new load balancer that can eliminate expensive cross-zone Kubernetes traffic without sacrificing overall reliability, even for multi-cluster traffic. 🔹 Support for Kubernetes' native sidecar containers eases some of the long-standing annoyances of the sidecar model in Kubernetes. Join us to get your hands dirty and put all three of these new features to work! #linkerd #servicemesh #opensource #cloudnative

Guide on optimizing llm accuracy for production deployment.

https://lnkd.in/gE7d8ujR "Chick-Fil-A in their most groundbreaking talk details how they viewed and used Kubernetes to scale and enable us all to eat more Chikin."

11 million times the official curl docker image was pulled from https://lnkd.in/e9mJ9Z49, a few million from docker hub ... just on wed last week... makes a one time daily record of ~14m usages of the latest, most secure version of curl in container land. Imagine if instead all those invokes were done by an older (sometimes much older) version of curl with all the attendant cves. Guessing this curl container is in top 10-20 of most used containers world wide ... real security should be real boring.

1. Select GPU runtime 2. %load_ext cudf.pandas 3. import pandas yes. it's that easy. https://lnkd.in/e4qet2Rr