HPE emphasized that Generative AI requires an architecture which is fundamentally different and more powerful than cloud-native architecture, and they unveiled it, along with related AI-native offerings.
Today, at their HPE Discover Barcelona 2023 event, Hewlett Packard Enterprise unveiled their next series of AI-native and hybrid cloud offerings for machine learning development, data analytics, AI tuning and inferencing, file storage and professional services. This all takes place through an open, full-stack AI-native architecture that utilizes a curated mix of software and infrastructure designed specifically for AI.
“We are announcing the development of what we are referring to as AI-native architecture,” said Evan Sparks, GM and Chief Product Officer for AI at HPE. “AI architecture is fundamentally different from cloud native architecture, reflecting the fact that the next decade will require full stack thinking around an extremely computationally intensive architecture. Our view is that this new architecture will require more thinking in terms of scale-up processing, as well as hybrid-by-design. To incorporate Generative AI will require moving from a cloud-native to AI-native architecture.”
HPE is emphasizing that their HPE GreenLake platform is strategically positioned to accomplish this with its open, full-stack AI-native architecture.
“Data is king, where enterprises can create differentiated experiences,” Sparks said. GreenLake provides a data-first pipeline to manage public and proprietary data using Generative AI, with AI lifecycle management software to accelerate workflows for training, tuning and inferencing.
“These kinds of AI solutions require world class lifecycle management software,” he stressed.
“You also need supercomputing in your DNA, and this is something at which HPE excels,” he added. He also emphasized the importance of hybrid by design to run AI anywhere from edge to cloud with data protection, as well as an open ecosystem.
“The goal is to accelerate time to value to deployment of AI in enterprises,” Sparks concluded.
New enhancements to the HPE Ezmeral Software container software platform further simplify and accelerate enterprise data, analytics and AI with a comprehensive, end-to-end platform that seamlessly operates across hybrid multi-cloud environments.
“These enhancements are about how we can design software to handle these new heterogeneous computer offerings,” said Mohan Rajagopalan, VP & GM, HPE Ezmeral Software. The platform now further simplifies and accelerates enterprise data, analytics and AI across hybrid multi-cloud environments.
The improvements include speeding up time to insight across the entire analytics lifecycle with the power of a hybrid data lakehouse that is now GPU- and CPU-optimized and can manage, access and analyze data across any NFS- or S3-compliant solution.
Model training and tuning in HPE Ezmeral Unified Analytics Software has been enhanced through deep integration with HPE Machine Learning Development Environment
NVIDIA GPU allocations across workloads and users have been optimized with GPU-aware capabilities in HPE Ezmeral Unified Analytics Software.
Finally ,third-party integrations with Whylogs for model observability and Voltron Data for GPU-accelerated queries have been expanded.
HPE GreenLake for File Storage, the company’s all-flash unstructured data platform, has also been improved to provide better AI capabilities.
“With HPE GreenLake for File Storage, we have significantly scaled up our workloads in the unstructured data space,” said Patrick Osborne, SVP and GM, Cloud Data Infrastructure, HPE Storage. “We now have significantly greater performance in throughput and density, and more support for very demanding workloads.” This translates into double the performance density, connectivity to new NVIDIA Quantum-2 InfiniBand for GPU-centric compute, quadruple the throughput, and quadruple the capacity density.
“We are evolving GreenLake for File Storage to address customers’ most challenging needs,” Osborne said.
New HPE Services around Generative AI include a broad portfolio of consulting services, workforce training and deployment solutions. The new AI services take customers through every step of the journey: from GenAI and LLM discovery to implementation, where customers develop the optimum operational models and hybrid cloud data strategies needed to build, deploy and scale solutions into transformative outcomes.
The new AI-optimized instances in HPE GreenLake will be rolled out throughout 2024, starting in January.