Jumping Off the Deep End of AI Data Storage, an Edge Conversation

Image generated by AI

When I think about data storage at the edge, it makes me think of the lemming. A well-known myth of them jumping off cliffs during migration has created quite a name for these little creatures. In the IT world there are cases where people can act like the lemming: following blindly and, in some cases as we approach the edge AI, even go off the edge. 

Much like the 1991 video game Lemmings, where players guide creatures by changing a single lemming's status (e.g., turning it into a "blocker" or "digger"), IT professionals can now leverage the insights of others more effectively than ever. Maybe even a genie could be a guiding solution, but more on that later. Leveraging these insights is contingent on all of us speaking the same language, though. As one might guess, the term edge is not a simply defined, always-accurate description of what is really being developed in the storage market today. Here at Solidigm, we have been working with customers and partners to better portray the definition of edge storage.  We provided some guidance in our article Top 5 Benefits of Solidigm D5-Series SSDs for Edge Storage Solutions early in this understanding of what truly defines “the edge” in data storage. Think of Solidigm as your hero figure in the game of Edge Definition. 

Image generated by AI

What is data storage at the edge?

As we continue down the path of what it means to be at the edge, here are some indicators:

• Distributed storage architecture
• Stored data at edge of network, not centralized cloud or on-prem
• Close to point of origin for data
• Remote/decentralized location

These are all considerations that come into play when trying to understand what is needed to deploy an edge architecture. One unique aspect of this is storage use cases. Unlike the ability to mix and match platform architectures in a data center, designs at the edge require more alignment with consolidated solutions, especially when it comes to storage devices. One critical consideration is to pick the right product at the outset, since edge solutions are not as fungible as data center and enterprise deployments. It is considerably more difficult to realign a workstream to new hardware, since there is no room for extra hardware, given space, power, and accessibility constraints. Solidigm looked at this in our earlier article, Emerging Use Cases for Data Center SSDs from Core to Edge. 

Continuing to deepen understanding of edge storage

As time evolves, so does the understanding of data storage at the edge. Recently Solidigm presented at the Tech Field Day Event, AI Infrastructure Field Day 2, where we got into the nuts and bolts of what it could mean to be an edge product.

This effort dove deep into the specs as it gave insight into what current IT developers are dealing with at many levels. We also showed how to deploy solutions at the edge, regardless of how far removed from the core data center the systems may be. To accomplish this we must have a better understanding of what customer solutions need and how customers are deploying those solutions. We have worked with several collaborators in this space, from a local zoo, to astrophotography, and even film and TV. There is so much more to this view of engagements, as we haven’t even touched on the impact of AI that is part of this amazing IT ecosystem change. 

The Solidigm advantage at the edge

Solidigm is the storage expert in AI, focused on solving real-world problems and forging deep relationships with the world’s most innovative companies. Together we can offer products that optimize efficiency and unlock new possibilities in constrained edge environments. No other storage provider can match our combination of technologies and expertise to help customers realize the value of AI, every time, everywhere. One might ask why? Let’s dive into that. 

80% of AI leaders are now saying their organizations will do more edge training and inference in the near future, according to Dell.¹ This means you need to have the right balance of compute and storage in a smaller, more restricted space. To manage large amounts of data, you must also have the capacity to store it easily, which is where the Solidigm™ D5-P5430 SSD in its E3.S form factor comes into play, as described by Gestalt IT.  

Another key factor to address is the efficiency of these edge AI systems, where key measurements are not in IOPS or GB/s but more important measurements like TBs/Watt, and TB/in^3. These new measurements highlight the importance of managing space that is a premium in any edge environment. We can’t have rows of racks in these solutions. In fact, in many cases, deployments are constrained to a single rack, a half rack, or even a car trunk!  Having a versatile set of solutions available is key. There are no one-off solutions, but these deployments need a coherent, thought-out list of options. This very thing is illustrated in the article “Ruggedized” Edge Server Uses Solidigm™ SSDs to Handle Extreme Conditions by Jordan Ranous at Storage Review.

What’s next?

How do we solve that next big problem? How do we manage expectations for ideas yet to come? At Solidigm, we look at designs, solutions, technology, and expertise of our team members. We have created a family of products that excel in the data center and are adaptable to edge platform needs. Think of the story of Aladdin and the genie, where IT pros of today are Aladdin looking to get ahead and find their place, and with a bit of guidance from a genie, can enable a happy ending. Going back to a favorite movie line with a twist, “Phenomenal Cosmic Capacity, Itty Bitty SSD Living Space” as a prompt to AI showcases just what Solidigm can do. Building products that solve customer needs at the core, edge, and in between with capacities that seemed so unattainable in recent past.  

In the Lemmings game, the goal is to get as many of the lemmings into the safe house as possible. As success at the edge depends on storing data where that data is created, Solidigm is committed to overcoming challenges to your deployments at the edge which, unlike those lemmings, does not follow the crowd, but creates individual solutions to keep your data accessible in whatever edge deployments your storage strategy requires.