The latest addition to Oracle'e Edge Cloud portfolio, Oracle’s second-generation Roving Edge Device (RED), delivers extreme processing power, seamless connectivity, and embedded security at the edge of networks and in disconnected locations. With RED, you can run multiple workloads, from enterprise apps to AI and select OCI services at the edge with simplified deployment, excellent price-for-performance, and superior security, including the ability to run isolated or air-gapped.
In the fast-paced AI age, milliseconds make a difference. Imagine a world where the boundaries of your network are limitless, and where the edge of your network becomes exponentially smarter. Oracle Roving Edge Device 2nd Generation (RED) introduces a new GPU optimized configuration with a compute- and storage-optimized configuration providing customers more flexibility with their deployment options.
Using the power of the Intel Xeon 8480+ processor at the edge, customers benefit from low-latency processing closer to the point of data generation and ingestion, resulting in more timely insights into their data. To put this power to the test, Oracle partnered with Intel to conduct a series of benchmarks over the first-generation RED. For the tests, we ran a Llama 2-7B, Yolov10 model, and Resnet50 convolutional neural network (CNN) solely on Intel Xeon processors. The following benchmarks compare the first-generation RED based on Intel Xeon 6230T against the second-generation with Intel Xeon 8480+:
Llama 2 is a family of pre-trained and fine-tuned text generation models based on an autoregressive, transformer architecture. Llama 2 comes with three models with 7 billion, 13 billion, and 70 billion parameters. For this simulation, Oracle benchmarked running the Llama 2 7 billion parameter model.
Latency Improvement
The second-gen RED can provide up to 13.6-times faster response rates than RED Gen 1 using the Llama 2-7B model, enabling lightning-fast performance for edge-based large language model (LLM) inferencing.
Throughput Enhancement Intel® Xeon® 8480+ processor
The RED Gen 2 can deliver up to 12.4-times better throughput using the Llama2-7B model, significantly boosting LLM data processing capabilities at the edge.
The YOLO series of models were designed for real-time object detection and accurate prediction of object categories and positions in images with low latency. In this series of benchmarks, Oracle compared running the YOLO v10 model on both generations of the Roving Edge Devices.
Latency Improvement
The new generation of RED can deliver up to 60% performance improvement over the previous generation.
67% improved throughput running YOLO v10
ResNet-50 is a convolutional neural network (CNN) architecture that belongs to the Residual Networks (ResNet) family, a series of models designed to address the challenges associated with training deep neural networks. Developed by researchers at Microsoft Research Asia, ResNet-50 is renowned for its depth and efficiency in image classification tasks. ResNet architectures come in various depths, such as ResNet-18 and ResNet-32, with ResNet-50 being a mid-sized variant.
Latency Improvement
The second generation delivers up to 3-times improved response rate over the previous generation, running the ResNet 50 CNN.
Up to 4x improved throughput
If you have application workloads that you need to deploy at the edge that require a scalable, secure, and flexible platform with the benefits of cloud technology and economics, Oracle RED is right for you. It’s a robust cloud-integrated service built to run mission-critical, time-sensitive applications at the edge in connected and disconnected locations.
With cost-effective, flexible configurations and the ability to support compute, storage, and GPU-intensive shapes, Oracle Roving Edge Device is the ideal infrastructure for anyone looking for a high-security, lowest-latency data processing and scalable environment right at the edge.. Learn more about RED, and give your organization the edge it needs today.
Tanmay Dhuri is a Senior Product Manager for OCI Edge Cloud suite of products spanning Oracle Compute Cloud@Customer, Roving Edge Infrastructure and Private Cloud Appliance.
Jeevan Sreenivas is Principal Technical Product Manager for OCI Edge Cloud Products.
Previous Post