Infinidat Introduces RAG AI Data Services

‘We work with [customers’ file data] to accelerate the process of learning and help avoid hallucination. We are targeting on-prem data. So this is not designed to go search the web or anything. That’s what ChatGPT and other products do. We are focused on corporate data for customers including the Global Fortune 2000,’ says Infinidat’s Chief Marketing Officer Eric Herzog.

Enterprise storage technology developer Infinidat Wednesday unveiled a new service it will provide to help customers prepare their data for use with GenAI.

Infinidat’s new retrieval-augmented generation workflow deployment architecture is aimed at helping users of the company’s InfiniBox and InfiniBox SSA enterprise storage systems to use them to optimize data for output to AI models, said Eric Herzog, chief marketing officer for the Waltham, Mass.-based company.

The new architecture is targeting AI workloads, anything that supports NFS file-based data, Herzog told CRN.

“Our professional services team will do the implementation,” he said. “No cost to the end user, no cost to the partner. Our professional services team works with our partners all over the world all the time. We help do migrations. We help we do custom programming for customers with our partner base.”

Retrieval-augmented generation, or RAG, will work with any vector databases such as Oracle or PostgreSQL, Herzog said.

“We work with that to accelerate the process of learning and help avoid hallucination,” he said. “We are targeting on-prem data. So this is not designed to go search the web or anything. That’s what ChatGPT and other products do. We are focused on corporate data for customers including the Global Fortune 2000.”

The services provided as part of the RAG workflow deployment architecture work with both new and old InfiniBox and InfiniBox SSA systems, including older generations, Herzog said. And since it works with NFS file data, it can be applied to heterogeneous storage environments, he said.

“If the customer has Infinidat with NetApp, or us with EMC file data, we could the RAG deployment to help pull that data in, not just from an Infinidat-only environment,” he said. “It’s really based on what LLM or SLM the enterprise supports, or their own. … Whatever applications they support, as long as it’s NFS file data. It could be a copy. Doesn't have to be the actual database. If they make a duplicate copy offline, or they've copied the database or the file system as snapshot, we could grab it from the snapshots as well.”

In addition to working with the InfiniBox and InfiniBox SSA storage systems, the new RAG workflow deployment architecture also works with Infinidat’s InfuzeOS Cloud Edition, the cloud version of the company’s core technology, Herzog said.

“If they put private data in the cloud, not going out to the internet to search everything, but some sort of data that they want that they've stuck in a cloud repository, and it's running our InfuzeOS operating system, we could pull data from any data that's in an on-prem private cloud, and any data that they put out in AWS or Microsoft as long as they're running the InfuzeOS operating system,” he said.

The new architecture works on live data as well, he said.

Once the preparation work, including designating which data sets are needed, what model is required, the source of the data, and so on, the actual service takes only 10 minutes to 15 minutes, Herzog said.

“We don’t charge for this, so the partner doesn't pay us a dime,” he said. “If the partner wants to charge, that’s up to the partner. We're not charging them.”

Eventually the services will be available for partners who want to do it themselves, Herz said.

People within Mainline Information Systems see the new Infinidat RAG workflow deployment architecture a big step in the right direction in the AI world, said Bob Elliott, vice president of storage sales for the Tallahassee, Fla.-based solution provider and Infinidat channel partner.

“As we speak with our customers, they are telling us that they are working to develop their AI strategies,” Elliott told CRN. “Infinidat is putting the tools in place to make it easier for them to create an AI environment and accomplish their goals.”

While Mainline Information Systems has its own professional services business, Elliott said the company appreciates working with Infinidat’s professional services team on AI projects.

“My experience with Infinidat services has been that it is absolutely a top-notch organization,” he said. “Infinidat stands by their solutions and services. When we sell anything Infinidat, we know it’s going to be a solid solution and the customer is not going to have problems. It’s a quality solution and that’s part of the reason why we so gung-ho with Infinidat.”

In any event, Elliott said, there are plenty of services opportunities in AI for everyone.

“If we can get a kick-start with Infinidat services to get a project going, to get it off the ground, there’s going to be plenty of opportunities to jump on and deliver our own services,” he said. “I look at this as seed for a much bigger environment.”