Clicky

Oracle Cloud Infrastructure Takes the High-Performance Computing Road
by Ron Westfall | September 28, 2020

The News: Oracle updated both its current capabilities and news related to the company’s Compute services roadmap. The announcements are designed to meet the high expectations of Oracle’s infrastructure customers, including for HPC solutions that deliver the performance they expect from on-premises systems and the pay-per-use, flexibility, and scalability of cloud, all at a competitive price point.

Included in Oracle’s announcements, Altair Engineering announced that it is selecting Oracle Cloud Infrastructure as the company’s preferred provider for internal use and for customer-facing SaaS products. Also, and as part of its HPC platform roadmap, Oracle is collaborating with Intel, including offering its next generation HPC Compute instances based on Intel’s Ice Lake processors early next year.

Continuing Oracle’s investments in HPC, its existing set of NVIDIA GPU offerings target fulfilling the unique demands of the public cloud space. Oracle announced earlier this year that it is working with NVIDIA on the next generation of GPU instances in Oracle Cloud Infrastructure with those instances becoming general available starting September 30 in the company’s U.S., EMEA, and JAPAC regions, at a competitive on-demand price of $3.05 per GPU per hour.

Oracle also announced an alliance with Rescale, an HPC provider. Availability of Rescale on Oracle Cloud Infrastructure makes it even easier for enterprises to onboard and get their jobs running in under a day. Rescale has more than 450 applications already preinstalled on Oracle’s high-performance computing instances, making it easy and giving enterprises the ability to bring their own licenses as well.

Moreover, Oracle is partnering with Ampere to bring an ARM offering to Oracle Cloud with the goal of using ARM-based compute instances to provide a compelling alternative for developers to diversify and take advantage of hardware innovation.

Finally, to get even further toward the company’s goal of responsive, performant, and cost-effective Compute services, early next year Oracle is going to offer customers E4 instances that build upon this platform with AMD’s next generation of CPUs instances. They are designed to allow for greater performance at the same economics, making this service well-suited for general-purpose workloads or custom applications. Read the full Oracle Software press release here.

Oracle Cloud Infrastructure Roadmap Update Centerpieces HPC

Analyst Take: Oracle Cloud Infrastructure (OCI) is betting massively on HPC in further differentiating and growing its business. In providing its hardware and compute roadmap update, Oracle unveiled a multitude of partnerships and customer deals that validate its strategic commitment to drive ecosystem adoption of cloud HPC applications.

In particular, Oracle stresses its alliances with top-tier semiconductor suppliers such as Intel, NVIDIA, Ampere, and AMD. These alliances are pivotal in enabling OCI to assert price/performance edges in the HPC domain over hyperscale cloud rivals such as Amazon Web Services (AWS) and Microsoft Azure.

These alliances are integral to OCI’s fundamental sales and marketing tenets consisting of:

  • Aiming to offer cutting edge hardware every new generation
  • Best performance per workload or use-case
  • Best price per performance per core
  • Driving differentiated and innovative capabilities such as Cluster Network
  • Easing migration with existing toolsets and instrumentation
  • Guaranteed performance SLAs

Oracle indicated that price/performance criteria are driving sales cycles and the key reason for winning new deals with enterprises such as Nissan Motors (competitive take away from AWS and Azure) and Altair Engineering (competitive take away from Azure). As such, Oracle is showcasing that OCI differentiators are fulfilling new customer demands and helping drive HPC innovation.

Oracle Cloud Infrastructure: Key Differentiators

Out of the semiconductor alliances OCI touted, the NVIDIA relationship is slated to have the most immediate HPC ecosystem impact as the general availability (GA) of the NVIDIA A100 GPU starts September 30, 2020. OCI is using the next generation NVIDIA Tesla A100 GPU based on 8 X A100 GPUs with fully connected NVLINK capabilities, 2 TB memory for massive graph workloads or GPU accelerated databases.

Additionally, Oracle is leveraging NVIDIA’s A100 GPU capacity to use over 25Tb non-volatile memory express (NVME) local storage and 1.6 Tb/s RDMA network capabilities along with 200G/s per GPU with ability to scale to 512x A100 GPUs per Cluster Network. For price differentiation, I see OCI’s claims of industry low pricing at $3.05/GPU in all regions as driving more enterprise adoption of OCI for HPC applications.

Of note, OCI GPUs combined with bare metal CPU and Cluster Network provides cloud-based performance that exceeds the on-premise performance of rival on-prem solutions such as Dell and HPE offerings. I believe Oracle’s ability to claim that OCI GPU performance exceeds even competitive on-premise offerings is a game changer toward driving accelerated, broader adoption of cloud HPC applications.

OCI is targeting H1 2021 for GA of its next generation HPC compute instances powered by Intel’s Ice Lake processors. The new offering promises the ability to build large clusters with over 20,000 cores using Cluster Network capabilities as well as 100G and remote direct memory access (RDMA) support with under 1.5 micro-second latency.

I believe these capabilities give Oracle sales and marketing credence to assert major differentiators such as enabling crash simulations, computational fluid dynamics (CFD), and electronic design automation (EDA) workloads that are expected to deliver more than 30% more performance compared to their existing X7 generation of instances at the same price per core.

Likewise, OCI is aiming at H1 2021 for the GA of its new ARM compute instances powered by Ampere’s Altra processors. The offering includes bare-metal or virtual machines (VMs) shapes that are available with up to 160 cores accompanied by single threaded performance at 3.3Ghz per core.

In sync with its upcoming HPC compute instance breakthroughs, I anticipate OCI gaining additional traction through ARM-based differentiators that include supporting general purpose workloads with up to 30% performance improvement over other x86 compute instances per core.

Finally, OCI announced that its next generation of AMD instances will be also GA in H1 2021. They are driven by AMD’s Milan generation of processors and are supported on bare-metal or VM shapes available through agile shapes allowing customers to pick and choose the core and memory split based on workload characteristics. The AMD instances are available through cloud native service such as Oracle Kubernetes Engine (OKE).

Oracle Cloud Infrastructure: Rescale Partnership

By announcing its strategic partnership with Rescale, OCI gives customers a holistic interface to migrate their on-prem HPC workloads and manage the entire process on Oracle Cloud. OCI needed to form an alliance with Rescale to counter the existing partnerships AWS, Azure, GCP, and IBM already have with Rescale and its ScaleX offering.

By enlisting Rescale’s hybrid and multi-cloud platform, which specifically targets running complex data processes and simulations in HPC environments, OCI strengthens its cloud HPC proposition and I anticipate the move will nudge OCI’s rivals to re-prioritize marketing their Cloud HPC specialist relations.

Oracle Cloud Infrastructure: Cloud HPC Key Takeaways

Oracle indicated that price/performance criteria is driving cloud HPC sales cycles over the last 12 months plus and its price/performance differentiators proved the key in winning new deals. However, I expect Oracle needs to heed security and energy considerations in order to further augment the impact of its OCI sales and marketing outreach across the HPC ecosystem. At least Oracle needs to blow kisses in the direction of cloud security and cloud energy considerations, as I anticipate these factors can spike in selection importance due to developments like a high-profile security breech or rival marketing campaigns. At the very least, Oracle needs to consider pre-empting rivals in their ability to use security and energy conservation criteria to work around any near-term price/performance shortfalls they have and dilute Oracle differentiators

Overall, I believe OCI significantly bolsters its competitive prospects in the cloud HPC domain, through its agile array of semiconductor and HPC partnerships and clear-cut ability to use price/performance differentiators to win new business. Now AWS, Azure, GCP, and IBM must directly counter the rising OCI threat or risk erosion in their cloud HPC market position.

Futurum Research provides industry research and analysis. These columns are for educational purposes only and should not be considered in any way investment advice.

Other insights from Futurum Research:

Oracle Commerce: Unpacking the Business in the Wake of Misinformation

Oracle Returns to Annualized Growth in its Fiscal Q1 as Cloud Grows

Meeting HPC Challenges is a Real Issue in Global Health Crisis

Image Credit: Oracle