In a recent Rackside Chat, Ruban, ASA/Racklive’s SVP of Systems and Solutions, sat down with Mark Pastor, Director of Product Management at Western Digital, to explore how AI is reshaping the role of storage—and how disaggregated infrastructure is driving better performance, flexibility, and scalability across industries.

Rethinking Storage for AI Workloads

AI workloads don’t operate like traditional enterprise applications. The stages of training, inference, and generative output each place unique demands on infrastructure, especially storage.

• Training requires massive data sets and favors high-capacity, cost-effective storage like HDD-based JBODs.
  
  
• Inference relies on real-time responsiveness, demanding high IOPS and low latency, which is best served by flash.
  
  
• Generative output, depending on the use case, often returns to long-term storage, again prioritizing capacity.

Western Digital is addressing these needs through purpose-built platforms like Ultrastar Data60/102 JBODs for capacity and OpenFlex for performance. This approach supports both ends of the AI spectrum—storing terabytes of raw training data and enabling fast, scalable inference at the edge or in the data center.

Why Disaggregated Storage Is the Future

Traditional server designs often tie compute and storage together, limiting flexibility. As compute components like GPUs evolve rapidly, replacing full server stacks becomes costly and inefficient.

Western Digital’s solution is a disaggregated architecture, where storage and compute scale independently. The OpenFlex platform, using NVMe-over-Fabrics, delivers performance comparable to locally attached storage while remaining fully modular. This model enables faster refresh cycles for compute while preserving long-term investment in storage.

For organizations managing evolving AI workflows, this architecture offers both immediate performance gains and long-term adaptability.

Ensuring Compatibility with the OCCL

To make deployment easier, Western Digital has launched the Open Composability Compatibility Lab (OCCL)—a testing initiative that validates how various third-party components perform within their composable storage ecosystem. This helps customers confidently build infrastructure using SSDs, DPUs, switches, and software tools that are already proven to work together.

This type of ecosystem validation is particularly valuable as enterprises embrace open standards and modular infrastructure at scale.

What’s Ahead in AI Infrastructure

Looking forward, both ASA/Racklive and Western Digital see increased demand for modularity, openness, and shared infrastructure. Storage systems must keep up with the explosive growth of AI—without locking customers into inflexible designs or short lifecycle investments.

Through ongoing collaboration, we’re bringing these next-generation solutions to life—whether it’s deploying OpenFlex flash at the edge or high-density JBODs in the data center.

Closing Thoughts

AI isn’t just changing the way businesses operate, it is transforming how infrastructure must be built. And storage is right at the center of that transformation.

If your organization is exploring AI infrastructure, you can reach out to us at info@asacomputers.com for free consultation or for more info on ASA’s WD storage solutions, please visit https://bit.ly/4lztQJc