글로벌 퍼블릭 클라우드, 엣지 컴퓨팅 및 엣지 AI 전문 기업 지코어(Gcore)는 AI 애플리케이션에 초저지연 경험을 실시간 제공하는 ‘인퍼런스 앳더 엣지(Inference at the Edge)’ 솔루션을 출시한다고 10일 밝혔다.
Cost-effectiveness, security, model auto-scaling, unlimited storage
Deploying ML Model Edge Inference Nodes…Ensuring Smooth Real-Time Inference
Running on NVIDIA L40S GPUs… Sub-30ms response time
Zicore responds at the nearest edge based on ML models, ensuring smooth real-time inference.
Gcore, a global public cloud, edge computing, and edge AI company, announced on the 10th that it is launching its 'Inference at the Edge' solution that provides real-time, ultra-low latency experiences for AI applications.
Zcore Inference at the Edge runs on Zcore’s expansive global network of over 180 edge nodes (PoPs), all interconnected by Zcore’s sophisticated low-latency smart routing technology. Each high-performance node is strategically placed at the edge of Zcore’s network, close to end users.
AI inference at each edge of ZigCor runs on NVIDIA L40S GPUs. When a user sends a request, the edge node routes it to the edge inference node with the closest latency, guaranteeing a response time of typically less than 30 milliseconds (ms, 1/1000th of a second). The maximum bandwidth between networks is 200Tbps, enabling the highest level of learning and inference capabilities.
Inference at the Edge also supports a wide range of basic machine learning and custom models. Open source based models available on the ZigCor Machine Learning Model Hub include LLaMA Pro 8B, Mistral 7B, Stable-Diffusion XL, and more. You can receive data from multiple formats or sources, select a model for your use case, train it, and then deploy it to inference-at-the-edge nodes located around the world. This solves one of the major problems that development teams commonly face: poor performance caused by re-running AI models on the same servers where they were trained.
Zigcore Inference at the Edge offers a flexible pricing structure that allows customers to pay only for the resources they use, enabling ‘cost-effective deployment.’ It also features built-in DDoS protection, where each machine learning endpoint is automatically protected by Zigcore’s infrastructure against DDoS attacks. It ensures superior data privacy and security that complies with industry standards such as GDPR, PCI DSS, and ISO/IEC 27001.
Additionally, model auto-scaling ensures that your models always support peak demand as well as unexpected load spikes. Finally, it provides unlimited S3-compatible cloud object storage to accommodate rapidly evolving needs.
Through these strengths, ZigCor Inference at the Edge helps enterprises across a range of industries, including automotive, manufacturing, distribution, and technology, to enhance their capabilities through cost-effective, scalable, and secure AI model deployment. This will enable enterprises to rapidly realize use cases such as generative AI, object recognition, real-time behavioral analytics, virtual assistants, and production monitoring on a global scale.
“Zcore Inference at the Edge enables customers to focus on training their machine learning models without worrying about the cost, technology and infrastructure required to deploy AI applications globally,” said Andre Reitenbach, CEO of Zcore. “At Zcore, we believe that the edge is where performance and end-user experiences are at their best, and we are continuously innovating to enable all of our customers to achieve superior scale and performance. Inference at the Edge will provide a modern, effective and efficient AI inference environment across all industries.”