“So, was it ‘AI everywhere, all the time’ at the Super Computing Conference 2025 (SC25)?” That’s a question people frequently ask me when I share that I recently attended SC25 in St. Louis, MO.
Admittedly, every conversation that I had at SC25 certainly touched on artificial intelligence (AI). However, I primarily discussed the ability of solutions to address AI’s insatiable power demands and unstructured data management (UDM) requirements. Further, delivering on AI’s power requirements and bringing unstructured data under AI’s purview have cast a shadow on AI itself.
Power, or Lack Thereof, an Emerging AI Concern
Three technology providers DCIG met with at SC25 offer neither power nor cooling solutions. However, these providers (Hitachi Vantara, OSNexus, and Quobyte) acknowledged that the topic of power inevitably comes up in client engagements. They hear from their prospective clients that they simply cannot get more power for their data centers, at least in the near term.
Rarely do their prospective clients inquire about power when initially evaluating their respective solutions. This stems from the fact that few IT managers currently possess oversight or responsibility for their organization’s power costs or cooling requirements. As a result, these individuals do not deeply examine the solution’s power efficiency capabilities at the evaluation stage.
Rather, the topic of power, usually cloaked under “sustainability”, surfaces during the procurement process. At this stage, prospective clients examine more deeply how efficiently the respective solutions utilize available data center power.
The ways in which each solution utilizes available data center power depends on each organization’s use case. For instance, Quobyte utilizes a hyperscale architecture for its software-defined storage (SDS) solution to handle high-performance AI workloads. To specifically address power concerns, its SDS solution can run on any x86 or ARM servers.
Quobyte has found that its deployment model has become a differentiator for cost- and power-conscious clients for at least three reasons.
- First, organizations are still trying to get their arms around AI and its value proposition for them. Quobyte’s architecture gives them the flexibility to use existing servers while achieving high levels of performance.
- Second, they more efficiently use GPUs. The high cost of GPUs, coupled with the growing difficulty in obtaining them, requires organizations to optimize their available GPUs. To achieve this objective, Quobyte stripes files across multiple storage servers and disks. In this way, available GPU servers may read and write in parallel, to optimizing each GPU’s performance while reducing the number of GPU servers needed.
- Third, Quobyte’s efficient use of available resources limits future power demands. Quobyte shared that some of its current and prospective clients now struggle even to get more power to their data centers. Quobyte can negate or delay an organization’s need for more power. Quobyte can leverage available hardware and power to help organizations jumpstart their AI journey.
Managing Unstructured Data at Scale
Procuring the hardware and software that AI requires, plus the energy to power it, only represents part of the challenge. Organizations must still grant AI access to the data it needs to fully realize AI’s potential.
How organizations give AI access to their unstructured data stores represents an equally daunting challenge. Here, the size of the organization (number of employees) matters less than the amount of data the organization manages. Further, the location of the data and the media on which the organization stores it also becomes paramount to quantify.
While at SC25, we spoke at length with Arcitecta’s CEO, Jason Lohrey, about these and other challenges of managing data at scale. Arcitecta has, for some time, offered its Mediaflux Universal Data System to address this challenge.
Lohrey recognized long ago that, at some point, that the “just buy more” strategy of storage management would fail. AI is now bringing that to a head. However, while solutions now exist that can consolidate and store petabytes of data, barriers to adopting them still remain.
Arcitecta services many governmental agencies and universities that have political and financial constraints that make such holistic storage consolidations difficult. These constraints require these organizations to maintain siloed storage environments. However, these same organizations often want access to the data contained in these storage silos.
Mediaflux gives these organizations and others like them access to that data. To do so, Lohrey developed a proprietary algorithm that indexes files and objects across even the largest of environments. The Mediaflux index remains comparatively small given the number of files and objects it indexes.
This index facilitates fast search and access to files or objects located anywhere in the environment. Mediaflux’s fast search and access specifically apply to those organizations that must maintain storage silos and want AI’s benefits. Using Mediaflux, organizations can bring all their unstructured data under management – up to 100s of trillions of files.
Arcitecta demonstrates Mediaflux’s ability to reach almost anywhere, as evidenced by its integration with Wasabi Hot Cloud.
Power and UDM Requirements Cast Shadow on AI
Many of the exhibitors and solutions on display at SC25 represent the intersection of AI’s promises and challenges today. On one hand, attendees can witness and experience firsthand the technologies that have driven AI’s adoption in recent years. Conversely, one also gets insight into the next round of challenges that those adopting AI face.
Few, if any organizations have access to the seemingly unlimited demand that AI has for power. Even for those that have ample power, even fewer organizations can easily and efficiently access their unstructured data stores. Thankfully, as evidenced by these solutions that DCIG saw at SC25, answers to these emerging challenges exist.
KEEP UP-TO-DATE WITH DCIG
To be notified of new DCIG articles, reports, and webinars, sign up for DCIG’s free weekly Newsletter.
To learn about DCIG’s future research and publications, see the DCIG Editorial Calendar.
Technology providers interested in licensing DCIG TOP 5 reports or having DCIG produce custom reports on their behalf, please contact DCIG for more information.