The new AI infrastructure reality: Bring compute to data, not data to compute

AI transforms enterprise operations in various industries, critical problems continue to surface around data storage, without the right information storage infrastructure, even the most powerful AI systems can be dragged with slow, fragmented or inefficient information pipelines.

This topic took the center scene on the first day of the day Turn a vbAI updates aimed at medical footage in a session Peak: Aio and Soligm. Together, next to Medical open network for EU (monai) The project helps a framework for the development and placement of medical footage, how to increase the data infrastructure in the AI, the infrastructure in hospitals to strengthen the use of developed research and operation.

>> – siSee all transforms 2025 coverage of our entire transform<

Innovative storage on the edge of Clinical AI

Management of the partner managed by Michael Stewart M12 (Microsoft Enterprise Fund), SEANS Roger Cummings, CEO of the summit, Aio and Greg Matson, Products and Marketing Marketing and Marketing Solidm. The conversation examined how the Next Generation, high-capacity warehouse architecture, the speed, security and measurement needed to manage the mass data in clinical environments, how the EU has opened its new doors.

In addition, both companies have been deeply involved in Monai since the first day. King’s College is designed with London and others, MONAI is built to develop and place AI models in medical view. The tools of the open source frame, suitable for healthy requirements – DICOM support, 3D image processing and model allow to build high performance models for tasks such as researchers and tumor segmentation and body classification.

The purpose of Monai’s goal is to support the placement, while using it to achieve training and results to complete the sensitive patient information to hospitals. The performance of this framework requires a close support for the information infrastructure required, real-time clinic AI requirements, fast, expandable storage systems. This is the Solidigm and the summit: Aio enters the game: Solidigm brings high-density Flash storage to the summit, Peak: AIO specializes in the storage systems installed for AI workload.

“We are very lucky to work early with the Royal College and Professor Sebastien Orslund in London,” said Cummings. “Working with Orslund, on this framework, researchers, doctors and biologists have developed very quickly infrastructure.”

Binary Storage Requirements in Health EU

Matson noted that the AI ​​information saw a clear bifurcation for various solutions optimized for the specific stages of the pipeline. Cases such as Monai, similar edge AI placement, as well as the participation of training groups, these environments are often ground and electricity, but require local access for mass data.

For example, Monai was able to save more than two million full-body CT scans on a node within the existing IT infrastructure of the hospital. “Many space limited, electricity and a very high-capacity warehouse allowed some remarkable results,” Matson said. This is a game change for EDGE AI, which allows you to operate developed AI models without loss of efficiency, health, institutions, expansion or data security.

On the contrary, the workloads covering real-time mode and active model training are very different requirements in the system. These tasks require storage solutions that can deliver very high access / output / output operations (IOOS) per second to ensure the full use of the information you need by high bandwidth memory (HBM). Peak: The storage layer defined by AIO software is combined with high-performance firm and state drivers (SSD), the power, efficiency and speed required throughout the AI ​​pipeline.

A layer defined by a program for clinical AI workload

Cummings, Peak, AI storage technology defined by AIO software, when Solidig’s high-performance SSD allows you to read and archive mass databases in the Mass database in the MONALIs. This merger accelerates the model of model and increases the accuracy of medical scientist while working within an open source frame adapted to the medical environment.

“We can submit a software that can be placed on any commodity server, turn it into a high performance system for AI or HPC business load,” said cummings. “In the outside environment, we take the same ability, and we are scaling up to a Node by bringing approaching to where the data is lived.”

Key Ability Peak How to: AIO Memory helps you to overcome traditional memory problems by entering more AI infrastructure. “We accept us as the infrastructure itself – as something that often deviates. Our solution is not only storage, but also a memory workplace and metadata related.” Said cummings. It creates an important difference for customers that allow you to run the great models of space or cost over and over. By keeping the peak live and accessible in the memory resident, the AIO allows localized results permanently.

Brings intelligence closer to data

Cummings stressed that the enterprises should require a more strategic approach to managing AI business download. “You can’t be a place.

Some clear trends arise from a large-scale AI placement in newly established Greenfield information centers. These devices are developed with highly specialized hardware architecture, which is as close to GPU as possible. To achieve this, all the solid state storage and specifically rely on high-level SSDs.

“Now the same technology is mainly in the facility, in a microconcement,” Cumming explained. “So, it is critical to buyers of AI systems to determine how you choose your device and system seller, even if you want to get the most performance from your system in all solid, it allows you to bring a lot of information as a system.