Sign up for our newsletter and get the latest HPC news and analysis.
Send me information from insideHPC:


Oracle Cloud Speeds HPC & Ai Workloads at GTC 2019

In this video from the GPU Technology Conference, Karan Batta from Oracle describes how the company provides HPC and Machine Learning in the Cloud with Bare Metal speed.

Oracle Cloud Infrastructure was built from the ground up to be the most powerful public cloud for customers running artificial intelligence (AI), high-performance computing (HPC), and big data applications. At GTC in Silicon Valley, the Oracle Cloud Infrastructure team is showcasing our work with these game-changing technologies. We’re also talking about the many ways that NVIDIA’s industry-leading graphics processing units (GPUs) can greatly accelerate these and other compute-intensive workloads in the cloud.

Oracle Bare Metal Cloud Services (BMCS) public cloud infrastructure. Oracle BMCS is a new generation of scalable, inexpensive and performant compute, network and storage infrastructure that combines internet cloud scale architecture with enterprise scale-up bare metal capabilities, providing the ideal platform for demanding High Performance Computing workloads.

Advantages include:

  • Performance. Bare metal GPU servers, without the hypervisor overhead, deliver uncompromising and predictable performance.
  • Elastic and On-Demand Compute. True on-demand bare metal instances deliver elasticity for your workloads. Start, terminate, and monitor bare metal instances as needed, using the console, APIs, or SDKs.
  • Control. Bare metal instance isolation and dedicated access provide unparalleled visibility and control.

Oracle Cloud Infrastructure was designed to deliver the best performance for GPU-accelerated computing, and the Oracle team accomplished this in several ways:

  • There’s no resource over-subscription, so performance doesn’t get compromised when networks are busy.
  • flat network topology reduces the number of hops and the associated latency between any two devices.
  • Off-box network virtualization offloads effort from the server, significantly reducing the typical performance tax.

Additionally, Oracle Cloud Infrastructure offers wide-ranging support for NVIDIA GPUs, including the high-performance NVIDIA Tesla P100 and V100 GPU instances that provide the highest ratio of CPU cores and RAM per GPU available. With a maximum of 52 physical CPU cores, 8 NVIDIA Volta V100 units per bare metal server, 768 GB of memory, and two 25 Gbps interfaces, these are the most powerful GPU instances on the market.

HPC Workloads

In addition to supporting the most powerful GPUs on the market, Oracle Cloud Infrastructure offers on-demand HPC infrastructure based on the most advanced compute, storage, and networking technologies—all at a fraction of the cost of building it yourself.

In the past, organizations were forced to run HPC applications on premises because first-generation clouds simply couldn’t handle the load. That’s not the case with Oracle’s second-generation cloud. With a wide variety of compute options, networking innovations, unmatched security, and industry partnerships, Oracle Cloud Infrastructure can handle any and all HPC workloads.

For example, Oracle’s bare metal cloud instances take advantage of clustered networking. As a result, enterprises don’t need to run specialized networking equipment on premises to connect with HPC workloads in the cloud. Additionally, Oracle Cloud Infrastructure is the first public cloud with a high-bandwidth, low-latency, remote direct memory access (RDMA) network.

Another benefit of running HPC in the cloud is more efficient utilization of server resources. Many organizations invest heavily in on-premises HPC infrastructure only to find that those server farms rest idle for long periods of time until actual HPC activities occur. By deploying HPC on Oracle Cloud Infrastructure, organizations pay only for what they use, and performance is consistent.

Big Data

Big data applications require clouds that offer elastic, high-performance compute, and infinitely scalable, independent storage. With its high-performing network, flexible compute shapes from virtual machines to bare metal instances, and multiple classes of storage, Oracle Cloud Infrastructure is built for big data.

Oracle enables customers to effectively analyze and gain valuable business insights from data across Apache Hadoop, Apache Kafka, NoSQL, and Oracle Databases. It’s all available on demand with consistent, predictable performance, and cost-effective pricing.

About the author: Kyle York is VP of Product Strategy, Oracle Cloud Infrastructure and GM, Oracle DYN Global Business Unit.

Check out our insideHPC Events Calendar

Leave a Comment

*

Resource Links: