AI at the Edge: HPE Apollo 4000 introduces support for NVIDIA A2 GPUs

[StorageExperts]

HPE introduces support for NVIDIA A2 GPUs on the Apollo 4000 data storage servers, enabling customers to accelerate AI and analytics workloads at the edge with up to 6x performance vs. CPU-based systems.

We're excited to introduce the latest updates to the HPE Apollo 4000 family. Last year, we released the HPE Apollo 4200 Gen10 Plus system, which provided expanded capabilities and performance to the existing Apollo 4000 portfolio.

Our customers continue to adapt to being data-driven organizations, and as such, desire highly scalable, performant, and intelligent infrastructure to support advanced data-centric workloads, often at the edge.

Edge AI is gaining traction

As we’ve discussed, enterprises today continue to make strides to becoming more data-driven organizations, and look for ways to find repeatable, sustainable value from their data. Improvements in AI and analytics, edge computing, and IoT have led to the emergence of edge AI. Simply defined, edge AI is the use of AI applications at the edge. This means the computation is done near the user at the edge of the network, close to where the data is located, rather than centrally in the cloud data center.

The underlying infrastructure supporting these efforts continues to evolve. And that’s where the updates to the HPE Apollo 4000 come in.  We are now announcing that the HPE Apollo 4200 Gen10 Plus system is capable of supporting NVIDIA’s A2 Tensor Core GPU. This means that customers can now run entry-level AI pipelines – typically found at the edge – on one system that can provide GPU-accelerated compute as well as ultra-dense storage capacity for large data sets. This further enhances your ability to accelerate data insights through AI and machine learning in edge locations. 

Additional highlights

• Ideal for AI the edge – NVIDIA A2 GPUs provide entry-level AI inference and low power requirements combined higher performance.
• Up to 6x performance gain – See up to 6x performance compared to CPU-based systems, dramatically speeding up time-to-insight.
• Unlock data insights faster – Ideal for key workloads such as video surveillance, natural language processing, Spark, converged AI and analytics, and more.

One of the most popular frameworks on which to run AI/ML models is Apache Spark. Apache Spark is a world-famous, open-source cluster computing framework used to process massive data sets by distributing these among multiple systemsHPE Apollo 4200 Gen10 Plus System in parallel. Read about the testing the HPE team has recently done for Spark on the HPE Apollo 4200 Gen10 Plus system. Among many reasons the Apollo 4200 Gen10 Plus is ideal for Spark, including its storage density, high throughput, and memory capacity, is the fact that it now supports the NVIDIA A2 GPU. As you can see, there are significant performance gains by running Spark on the A2 vs. a CPU-based system – over 3x faster for this specific workload. Overall, performance gains can be up to 10x faster depending on the workload.

Other HPE Apollo 4000 enhancements

The Apollo 4000 systems will now support 20TB SAS/SATA LFF drives, increasing the density and total capacity of the entire family of intelligent data storage servers, leading to more economic data lakes, archives, and data stores.  Check out the new total capacities for each model:

• Apollo 4510 Gen10: Up to 1.2PB in 4U
• Apollo 4200 Gen10: Up to 560TB in 2U
• Apollo 4200 Gen10 Plus: Up to 640TB in 2U

Learn more

• Read the technical white paper: Why the HPE Apollo 4200 Gen10 Plus server is an optimal system to run Spark
• More information on the HPE Apollo 4000 family portfolio
• More specific technical information and specifications for the updates to the HPE Apollo 4000 family

---

Meet Storage Experts blogger Matt Miller, HPE Storage

Matt leads Solutions Marketing for the HPE Storage business, covering such areas as file-and-object storage, scale-out storage, virtualization and containers, and cloud-native technologies. Connect with Matt on LinkedIn. You can also find him on Twitter @Matt_Miller22.

Storage Experts
Hewlett Packard Enterprise

twitter.com/HPE_Storage
linkedin.com/showcase/hpestorage/
hpe.com/storage