Red Hat marches full steam ahead towards AI
At its summit, Red Hat also shows that there is no way around AI. But you won't find a chatbot or another OpenAI partnership.
(Bild: iX)
- Udo Seidel
Every year in spring, Red Hat invites you to its in-house conference. And of course, there can be no IT conference in 2024 that is not permeated by AI. The Summit is no different.
AI for OpenShift and Ansible
In principle, the Red Hat world has revolved around two things for a few years now: OpenShift as the in-house Kubernetes version and Ansible for automation. Linking the two with AI is therefore the basis and even the core of the new announcements during the summit.
OpenShift AI has been around for some time. With the new version, there are several new features. For example, AI models in the area of edge computing can be provided using a single node. Users can now also use various so-called model servers in a network. Support for Kserve is also included. This is known from the general Kubernetes world as a platform for model inference.
OpenShift AI can now also use Ray technology to accelerate data processing. KubeRay and CloudFlare make administration easier. The former focuses more on the actual workload in the Kubernetes cluster, while the latter is more concerned with the orchestration of individual tasks. Numerous acceleration profiles can now be created and configured. This makes life easier for users in particular: they can now simply select the acceleration platform of their choice. Model development and monitoring have also been improved.
The above-mentioned acceleration profiles are also complemented by the fruits of Red Hat's collaboration with Nvidia and Intel. Intel CEO Pat Gelsinger – unfortunately only represented virtually at the summit – announced the integration of OpenShift with Intel's AI technology. This applies to hardware such as the Gaudi chips or Arc GPUs, as well as the associated software platforms. Nvidia presented NIM (Nvidia Inference Microservices) – a set of microservices for GenAI on the in-house hardware platform CUDA – at its in-house trade fair GTC in March. They will soon support OpenShift; here too, integration will take place via the aforementioned Kserve. The collaboration with AMD is also still at an early stage. The declared aim is to provide a corresponding OpenShift operator, which is already available and installable as a technical preview for developers.
In this four-day online workshop, you will get to know Ansible and its areas of application from the ground up with the help of many practical exercises. You will set up your own Ansible environment and complete complex tasks with the help of playbooks. We will also show you the Ansible role concept in detail. The workshop will take place online, further information and tickets are available at heise.de/s/wNL8A.
The latest announcement in this area is the integration of Lightspeed. This applies to both OpenShift and the company's own Enterprise Linux. Those in the know have been familiar with this from the Ansible environment since last year. The AI behind it is intended to support users in their day-to-day business, for example with scaling tasks or the detection of an impending capacity bottleneck. On the RHEL side, Red Hat wants to support the administrator in distributing security updates: The AI is to bring aspects such as maintenance windows or service agreements into the patch organization. However, this is still a long way off.
Once again on Ansible: automation has long extended to much more than just infrastructure. Regulations, compliance, and other legally relevant conditions have become increasingly important recently. With the AI hype and the numerous applications in this environment, complexity has once again risen sharply. With Ansible Policy as Code, Red Hat aims to simplify this area for administrators and users. An AI is once again working in the background, which according to the provider is a natural evolution of classic automation for Ansible.
Better LLMs with IBM
Red Hat is also releasing InstructLab with the aim of significantly improving LLMs. It uses LAB technology, which originates from the IBM research department. The name is derived from "Large-Scale Alignment for ChatBots". LAB generates synthetic data according to a standardized procedure and adapts it in a multi-stage process. IBM has documented the exact details in a white paper. Big Blue is publishing several selected language models from the Granit family under the Apache license to support this.
If Red Hat keeps up the pace of announcements at the Summit, further innovations can be expected in the coming days. Incidentally, this time the Summit is not taking place in San Francisco or Boston, but in Denver. This will be the center of the Red Hat universe from May 7 to 9.
(olb)
