Cloudian has enabled its Hyperscale AI Data Platform software to run on AWS Local Zones, giving AWS customers there single-digit millisecond latency access to object data.
An AWS Local Zone is a sub-region AWS service point, and has compute, including GPUs, storage, and select AWS services located closer to large population centers and industry hubs than an AWS region data center. A Local Zone AI Data Platform (AIDP) is primarily intended for enterprise customer document retrieval-augmented generation (RAG), with customers deploying “AI agents that understand and reason over their complete repository of documents, manuals, reports, and multimedia content stored in S3-compatible formats.“ It supports data sovereignty as the data is stored in the local zone.
Cloudian CTO Neil Stobart offered this thought: “By combining Cloudian’s high-performance S3-compatible storage with AWS’s GPU-based edge infrastructure, we’re enabling enterprises to run sophisticated RAG applications within milliseconds of their end users, with zero upfront cost.’ He reckons: “This dramatically accelerates AI adoption for organizations that previously couldn’t justify the infrastructure investment.”
Example uses could be customer service teams accessing product documentation faster for accurate responses, field technicians retrieving repair procedures in real-time, and employees finding answers to their problems without finding their way through nested file systems.
The AWS Local Zones extend AWS infrastructure closer to end-users in metropolitan areas, enabling ultra-low latency applications that could be critical for government and healthcare. Public sector agencies use them for real-time citizen services and emergency response systems, while health sciences leverage them for medical imaging, telemedicine, and patient monitoring requiring immediate data processing and regulatory compliance.
AWS says its Local Zones, with their pay-as-you-go pricing, can be used instead of building a local data center or having a CoLo contract deal. Trad edge AI data center deployments require substantial capital investment in data center building shells, GPU servers, networking equipment, storage, and power and cooling. This could take 6-12 months from planning to production. A Local Zone facility can be set up in less than a day, hours in fact.
AWS Local Zones are available in 35 metropolitan areas around the world, with GPU-accelerated instances for AI workloads available in select locations. For more details, visit the AWS Local Zones features page.
The AIDP local zone deployment features GPU-powered cloud servers purpose-built for AI/ML workloads, featuring up to eight Nvidia Hopper GPUs with 640 GB of GPU memory, 3rd Gen AMD EPYC processors, and 3,200 Gbps of Elastic Fabric Adapter (EFA) networking for scale-out performance. Cloudian says this enables real-time inferencing with sub-10 millisecond response times. It means organizations that need AI inferencing at their edge locations, for response time and/or data sovereignty reasons, but cannot afford a full edge data center, can now have inferencing at a lower cost.
Its software includes an integrated vector database and capabilities to automatically ingest, embed, and index multimodal content for RAG deployment.
Bootnote
Find specific AWS Local Zone locations here.
