Time to Simplify: A Fresh Look at Infrastructure and Operations for Artificial Intelligence

The hype triggered by the emergence of generative synthetic intelligence (AI) feels lots just like the early days of cloud, bringing the subject—and the necessity for a method—to the entrance of the IT chief agenda.

But whereas AI is poised to alter each facet of our lives, the complexity of AI infrastructure and operations is holding issues again. At Cisco, we imagine AI is usually a lot simpler once we discover methods to keep away from creating islands of operations and convey these workloads into the mainstream.

AI is driving massive adjustments in information heart expertise

AI workloads place new calls for on networks, storage, and computing. Networks have to deal with lots of knowledge in movement to gasoline mannequin coaching and tuning. Storage must scale effortlessly and be carefully coupled with compute. Plus, computing must be accelerated in an environment friendly method as a result of AI is seeping into each utility.

Consider video conferencing. In addition to the acquainted CPU-powered parts like chat, display sharing, and recording, we now see GPU-accelerated elements like AI inference for real-time transcription and generative AI for assembly minutes and actions. It’s now a combined workload. More broadly, the calls for of knowledge ingest and preparation, mannequin coaching, tuning, and inference all require totally different intensities of GPU acceleration.

Using confirmed architectures for operational simplicity

IT groups are being requested to face up and harden new infrastructure for AI, however they don’t want new islands of operations and infrastructure or the complexity that comes with them. Customers with long-standing working fashions constructed on options like FlexPod and FlashStack can deliver AI workloads into that very same area of simplicity, scalability, safety, and management.

The constituent applied sciences in these options are perfect for the duty:

• UCS X-Series Modular System with X-Fabric expertise permits for versatile CPU/GPU ratios and cloud-based administration for computing distributed anyplace throughout core and edge.
• The Cisco AI/ML enterprise networking blueprint exhibits how Cisco Nexus delivers the excessive efficiency, throughput, and lossless materials wanted for AI/ML workloads; we imagine Ethernet makes the perfect expertise for AI/ML networking attributable to its inherent cost-efficiency, scalability, and programmability.
• High-performance storage techniques from our companions at NetApp and Pure full these options with the scalability and effectivity that enormous, rising information units demand.

Introducing new validated designs and automation playbooks for widespread AI fashions and platforms

We’re working laborious with our ecosystem companions to pave a path for purchasers to mainstream AI. I’m happy to announce an expanded street map of Cisco Validated Solutions on confirmed trade platforms, together with new automation playbooks for widespread AI fashions.

These options span virtualized and containerized environments, a number of converged and hyperconverged infrastructure choices, and necessary platforms like NVIDIA AI Enterprise (NVAIE).

These answer frameworks depend on our three-part strategy:

1. Mainstreaming AI infrastructure to cut back complexity throughout core, cloud, and edge.
2. Operationalizing and automating AI deployments and life cycle with validated designs and automation playbooks.
3. Future-proofing for rising part applied sciences and securing AI infrastructure with proactive, automated resiliency, and in-depth safety.

“Building on a decade of collaboration, Cisco and Red Hat are working together to help organizations realize the value of AI through improved operational efficiencies, increased  productivity and faster time to market. Cisco’s AI-focused Cisco Validated Design can help simplify, accelerate and scale AI deployments using Red Hat OpenShift AI to provide data scientists with the ability to quickly develop, test and deploy models across the hybrid cloud.”
—Steven Huels, Senior Director and General Manager, Artificial Intelligence Business, Red Hat.

The momentum is actual; let’s construct for the longer term

AI’s infusion into each trade and utility will proceed to speed up, even because the part applied sciences every make their method by the hype cycle to adoption. Increased information assortment and computing energy, developments in AI frameworks and tooling, and the generative AI revolution—are all fueling change. Let us make it easier to construct on trusted architectures and take these workloads mainstream for max impact.

Join our December 5 webinar:

Share: