Making ready for the way forward for AI and HPC with HPE GreenLake edge-to-cloud platform

The HPE GreenLake platform powers your initiatives with a cloud expertise that permits you to keep away from the prices, dangers, and latency of transferring your knowledge to and from the general public cloud, whereas assembly the compliance and regulatory necessities for privateness and knowledge sovereignty. Be part of us at GTC to be taught extra.

Immediately, organizations that begin their AI/ML pilots within the public cloud face a couple of essential questions when making ready to take their present mannequin and knowledge entry to manufacturing. How do I do know the true value of my service working bills on the subject of AI and ML? How correct is my mannequin? Are my predictions aligned to or drifting from what’s occurring? How do I run my knowledge science, AI and HPC workloads and different a number of workloads on the identical platform and at any time? How do I entry excessive efficiency compute, and parallelize GPUs and the reminiscence related to them to coach a mannequin at larger ranges of accuracy? How can we finest scale with value management in the long run?

As well as, for a lot of organizations, excessive efficiency computing may be expensive, in addition to advanced and time-consuming to handle. Public cloud is often not the most suitable choice, as a result of workloads rely upon low-latency knowledge motion, highly effective processing, and excessive availability of server sources. Fortunately, for IT and OT groups supporting AI/ML and HPC workloads, in addition to for groups managing the info and ML mannequin lifecycles, there’s an possibility with the flexibleness, pay per use, and scalability of public cloud on premises.

The HPE GreenLake edge-to-cloud platform is a totally managed on-premises resolution with a cloud expertise that permits you to keep away from the prices, dangers, and latency of transferring your knowledge to and from the general public cloud, whereas assembly the compliance and regulatory necessities for privateness and knowledge sovereignty.

HPE GreenLake for MLOps brings DevOps agility to each stage of the ML lifecycle, and frees up your knowledge scientists to focus on knowledge science by permitting them to simply entry open-source or third-party unbiased software program vendor (ISV) instruments, in addition to required knowledge sources throughout their group. The brand new launch of the answer runs a wide range of NVIDIA accelerators, together with the A100 Tensor Core GPU for AI/ML coaching and inferencing use circumstances in each HPE Apollo and HPE ProLiant servers, that are all NVIDIA-Licensed.

NVIDIA NGC catalog is a big software program ecosystem that accommodates a number of production-tested AI container libraries and continues to increase at a speedy fee. HPE and NVIDIA labored intently to combine two of those libraries, for deploying at scale, which might be examined and made out there within the HPE GreenLake market: NVIDIA RAPIDS Accelerator for Apache Spark and the NVIDIA Triton Inference Server. These containerized workloads ship sooner processing and response, sooner mannequin coaching and, in the end, sooner insights (when in comparison with CPUs). A part of HPE GreenLake MLOps, HPE Ezmeral software program simplifies the event, scaling, and administration of Kubernetes clusters throughout hybrid environments.

HPE GreenLake for Excessive Efficiency Computing now consists of help for NVIDIA NVLink and NVIDIA A100, A40, A30 Tensor Core GPUs in increments of 2-4-8 accelerators, able to quote, value, and ship. With HPE GreenLake, clients can select an on-premises HPC resolution working on NVIDIA GPUs with all the flexibleness, scalability, and utility-like consumption of the cloud. Clients can enhance their agility with pay-per-use pricing and a preinstalled buffer capability that is able to provision when wants develop. Whether or not demand for HPC sources spikes all of a sudden or grows steadily, clients are at all times prepared to satisfy their new wants.

Clients can design their very own HPC infrastructure resolution inside HPE GreenLake utilizing industry-leading HPE applied sciences. Alternatively, they’ll standardize their service with pre-sized configurations that provide self-service and are managed for them. With entry to the most recent know-how out there available in the market, HPE also can purchase out—and recycle—clients’ current infrastructure, serving to them meet sustainability targets.

The latest member of the NVIDIA GPU portfolio is the NVIDIA H100 Tensor Core GPU, powered by the NVIDIA Hopper structure, which delivers the following large leap in NVIDIA’s accelerated compute knowledge middle platform, securely accelerating numerous workloads in each knowledge middle, from small enterprise workloads to exascale HPC and trillion-parameter AI. The NVIDIA H100 delivers unprecedented efficiency, scalability, and safety to each knowledge middle, accelerating all the things from Exascale and trillion-parameter fashions all the way down to partitioned GPU-right-sized compute situations.

Higher collectively

With roughly 70% of apps and knowledge nonetheless residing on premises, the HPE GreenLake platform delivers the cloud expertise throughout your edges, colocations, and knowledge facilities. NVIDIA’s pioneering methods, apps, and fashions mixed with the experience and complete set of computing and infrastructure breakthroughs from HPE may also help you unlock the worth of AI/ML, VDI, and HPC and lead a data-first modernization. For instance, medical imaging pioneer Carestream will remodel X-ray methods throughout the globe utilizing HPE GreenLake for MLOps (learn extra), and Eni is upgrading its HPC4 system for its Inexperienced Knowledge Middle (learn extra).

Don’t miss these HPE periods at GTC

Attend HPE’s periods at GTC to find methods to speed up your AI implementation from edge to cloud with HPE GreenLake cloud companies and to community with a world group of builders, researchers, engineers, and innovators.

For those who haven’t already, make sure and register at: https://www.nvidia.com/gtc/?ncid=ref-spo-810249

SIMULIVE S42447: From Pilot to Manufacturing: Delivering Enterprise-Grade AI as a Service on an Edge-to-Cloud Platform

Glyn Bowden, CTO AI & Knowledge Advisory Providers HPE GreenLake, HPE

Wednesday, March 23      8:00AM PT / 4:00PM CET

SIMULIVE S41144: Reliable, Explainable AI in Monetary Providers With Knowledge-Centric Structure

Aileen Toleikis, Knowledge Analytics Guide, HPE

Jochen Papenbrock, Monetary Providers & Expertise Developer Relationship Lead EMEA, NVIDIA

Wednesday, March 23      10:00 AM CET / 1:00 AM PT

SIMULIVE S42645: What is the Buzz About Swarm Studying?

Steve Heibein, AI Lead, HPE

Wednesday, March 23       7:00 AM PT / 3:00 PM CET 

SIMULIVE S41141: The Underestimated Energy: AI & Knowledge Worth Creation Inside Firms

Thomas Meier, Chief Technologist Knowledge Follow, HPE

Wednesday, March 23        9:00 AM PT / 5:00 PM CET

Session S42455: Scaling Language Coaching to Trillion-Parameter Fashions on a GPU Cluster

Evan Sparks, Vice President, AI & HPC, HPE

Session S42462: Optimize Efficiency and Administration of Apache Spark Containerized Workloads

Matthew Morris, Technical Advertising Engineer, HPE

For extra data, go to us at HPE GreenLake for AI/ML and Analytics

Flynn Maloy
Hewlett Packard Enterprise

Twitter: @FlynnMaloy

LinkedIn: Flynn Maloy