Cloud Hosting

Refresh

Thanks for following ITPro's coverage of the opening keynote at KubeCon 2024 in Paris. Stay tuned for more KubeCon news over the next few days!

The crowd is thanked for joining and, just like that, people begin to filter out the keynote theatre.

And now, as they leave the stage, it looks like it's already time for the opening keynotes closing remarks! The hosts announce some exciting events that are yet to come during the conference.

Managed model training is a topic of the pairs address, and they discuss using Buildpack and Kubeflow, before moving on to talking about multi-cluster scheduling. Knative and KServe are also platforms used at Bloomberg.

To support their features, they need a scalable infrastructure and they need to support a complex machine learning lifecycle. They leverage open source solutions to help with this process.

They talk about the companys flagship product, Bloomberg terminal. The company handles a massive amount of data every day. They talk about how Bloomberg terminal uses AI (unsurprisingly given the focus of the keynotes so far).

Now two more speakers for the next keynote, Yuzhui Liu Team Lead at Bloomberg, and Leon Zhou, Software Engineer at Bloomberg.

After another speaker jumps into the news room, Ibrahim Haddad (via video), its a wrap for the news show!

Now Kakodkar will speak to cloud native and AIs relationship more generally, talking about the relationship of infrastructure and cloud native as symbiotic.

The keynote now cuts to an imitated live report from somewhere outside the conference center.

This just inKubernetes enters its second decade. It learned how to walk and then run, and now will fly into the AI galaxy - it looks like this keynote session will feature several different news broadcasters, or just plain speakers, over its duration.

Next to the stage is the curiously named Cloud Native News Show (complete with faux-news show graphics) headed up by Nikhita Raghunath and Rajas Kakodar.

To close out, the panel host reaffirms Kubernetes as becoming the standard for AI platforms, while also going over some of the panels discussion on resource allocation, data loading, and the importance of choosing the right specialized compute for the right model.

The questions are flying: What can we do to optimize for sustainability? One panelist jumps in to tout lower usage compute systems that, in turn, create lower costs.

Compute power seems to be a running theme in KubeCon 2024s opening keynotes, and the panels attention is turned considerably towards concepts of GPU sharing and efficient GPU use.

The panel starts with a question about how Kubernetes might be simplified, followed by a question about the compute decisions people should make in regard to Kubernetes. GPU-free LLM inference is one response to the latter of the two questions.

Next to the stage are several people for a panel discussion: Lu Qiu from Alluxio, Susan Wu and Clayton Coleman from Google, Victor Jakubiuk from Ampere Computing, and Ricardo Rocha from CERN.

Thats why hes announcing the companys new KAITO architecture, which will deploy and run inference for models. Its a speedy run through of the product and then hes off!

Next to the stage is Jorge Palma, Principal PM Lead at Microsoft, who will be taking the lead for a sponsored keynote on building an open source platform for AI. There are a lot of issues along the way, he says, such as containerizing models.

Topology aware placement is key to avoiding problems, he says, as well as an understanding of fault tolerance and resiliency. Problems need to be diagnosed easily and observability needs to be built in.

Klues hands over to Chatterjee to talk about the issues around scaling out GPUs, particularly with regard to training. To satisfy the demands of generative AI, he says, a vast amount of GPUs are needed.

They talk about time-slicing, a technique used to split the demands of AI workloads, as well as GPU sharing. Kevin gives a little shout out to some other talks at the conference if people want to know more!

They talk about how GPUs are at the heart of the AI revolution and will talk about what it takes to enable GPU support in Kubernetes, support that can help in directing AI workloads.

Two new speakers hit the stage, Kevin Klues, Distinguished Engineer, and Sanjay Chatterjee, Engineering Manager at NVIDIA, to talk about AI workloads with GPUs in Kubernetes.

Sharma brings up the phrase open source washing, which turns the groups attention to the importance of open source in AI. Morgan talks about the open source nature of Kubernetes but then also comments that this is affected by whether a model is open source, while Lacroix says that it would be helpful for all the tooling for AI to be open source.

The group also talk about inference cases, the possibility of smaller model use, and evolution in the space in general. They talk about the different workloads for AI, such as training, fine-tuning, and inference.

Sharmas focusing on linking cloud native and AI together, now bringing to the stage the co-founder of Mistral AI, Timothe Lacroix, lead product manager at Google, Paiege Bailey, and founder of Ollama, Jeffery Morgan. They discuss the relationship between AI and cloud native.

Sharma develops a visual language model that impressively details a picture of the conference hall itself! Everyone was suitably pleased.

Sharma talks about how, together, cloud native has made platforms more resilient and stable, listing some of the big names associated with it - Adidas, Disney, Spotifythe list goes on.

This is apparently the largest KubeCon ever according to Sharma - so big, she jokes, that even her parents are here!

Priyanka Sharma, Executive Director of Cloud Native Computing Foundation (CNCF) has hit stage to wild applause from the audience.

2024-03-20T08:19:57.636Z

The keynote session has begun, kicked off with a promotional video!

Everyone has now been allowed into the keynote theatre and it's filling up quickly.

No one's allowed into the keynote theatre just yet so this is a good time to get acquainted with the conference space.

The weather's getting nice now! KubeCon's keynote talks are located on the third floor of the centre.

The registration area is filling up, although the Paris Expo Centre's coffee shops aren't open yet unfortunately!

It's a crisp but sunny day here in Paris as people make their way in to the Paris Expo Centre for the first day of KubeCon 2024.

Welcome to ITPro's live coverage of KubeCon 2024 at the Paris Expo conference center. We're on the ground in Paris waiting for the opening keynote to begin.

In today's session, we'll hear from the Cloud Native Computing Foundation (CNCF) on all the latest developments in the cloud native space, along with community updates and live presentations from a host of industry figures.

Receive our latest news, industry updates, featured resources and more. Sign up today to receive our FREE report on AI cyber crime & security - newly updated for 2023.

Receive our latest news, industry updates, featured resources and more. Sign up today to receive our FREE report on AI cyber crime & security - newly updated for 2023.

Read more from the original source:
KubeCon 2024: All the latest news and announcements - ITPro