Among those who have argued about how well a product is suitable for AI workloads, Hitachi Vantara has a unique backstory to support that debate. This means that Japanese array makers are part of a huge manufacturing conglomerate that manufactures everything from nuclear power plants and high-speed trains to air conditioners and home appliances, and uses Hitachi Vantara products to process data.
The key to the story is that the company will provide a convergent infrastructure portfolio (Hitach IQ) that combines NVIDIA GPUs and Enterprise AI software with Hitachi Vantara's VSP One storage array, hammer space storage, data orchestration, Hitachi Vantara Server products, and Cisco NetWorking equipment.
“Our group uses Nvidia's Omniverse Digital Twin Ecosystem, which provides AI training data that enables the development and expansion of robotics capabilities in manufacturing.”
AI convergent infrastructure
Meanwhile, Hitachi IQ, Hitachi Vantara's family of AI convergence products, is a fully converged infrastructure with one to sixteen super microservers with eight NVIDIA GPUs for AI processing using NVIDIA HGX configurations.
Next, there are multiple Hitachi HA G3 servers that share the (object storage) content of a VSP one array node. Some of these servers run the NVIDIA AI Enterprise Software Layer in a Kubernetes container. Others run HammerSpace storage software that allows parallel access between the GPU and storage.
Finally, the Cisco Nexus Switches connect the whole thing. Regarding the role of VSP One Array (the flagship of the Hitachi Vantara Array family), these servers are connected to HammerSpace servers to provide object storage for most of the data they distribute in file mode.
IQ Time Machine: Gives memory to LLM to VSP One
“There are some advantages based on the VSP array,” Hardy said. “Among them is the new Hitachi IQ Time Machine feature, which allows for the submission of previous versions of documents and data to LLM.
The point of Hardy here is that such documents on other systems are updated, and thus the previous version is lost to LLMS interrogating that dataset. The rag function is based on the retention of historical data in object storage in VSP One Array, and IQ Studio (a chatbot that provides Hitachi IQ infrastructure to Hitachi Vantara) provides this via the interface's timeline.
For example, if a member of the finance team wants to ask AI about an event, they can hover over the date and see the details notified via the document ingested at that time. This allows customers to access data for various periods via LLM.
Data storage is an important component of your AI project. This is because three constraints need to be successful. The array must communicate very quickly with the GPU. For RAG, the data must be in a format compatible with NVIDIA software modules that build AI applications. And finally, they need to help companies prepare and test the data they submit to AI.
With Hitachi IQ, which goes far beyond the capabilities of Storage, Hitachi aims to tackle these three challenges simultaneously.
