Conference Presentation, Other
Shaun O’Meara, CTO @ Mirantis: All Infrastructure is AI Infrastructure
- AI is expected to become the fundamental infrastructure for all future systems, replacing traditional application workloads and hardcoded business logic with dynamic AI-driven decision-making.
- While older workloads and AI interactions with the real world will persist, the industry faces resource scarcity regarding GPUs, power, and data center space, necessitating a shift toward using smaller models for unique workload parts to reduce costs.
- Computational demands will increasingly rely on a full-stack approach where many data pipeline jobs run more efficiently on CPUs, requiring separation of the management and observability layer from the data plane to ensure safety and prevent single-vendor lock-in.
- Organizations must prioritize rapid time to value and composable infrastructure to avoid becoming legacy systems, utilizing templates for repeatability and adopting a hybrid future strategy that mixes strategic vendor solutions with pure-play open source.
- The vision involves moving from resource abstraction to formal contracts where applications specify performance requirements, supported by the "Cordant" platform which is designed for multi-cloud, multi-cluster, and bare-metal environments using a declarative, pattern-based approach.
- Management and observability tools must enable flexibility to change stack layers as vendors mature, with specific attention to ensuring management plane failures do not impact workloads, while the AI Factory reference material remains subject to regular updates.