The rapid retrieval of data—at scale—required for many cognitive computing use cases depends on several factors, only one of which is the vaunted computational power many attribute to the current resurgence of Artificial Intelligence.
Cognitive computing systems must also have credible memory, which frequently takes the form of in-memory capabilities these days. Most of all, perhaps, they need quick, dependable storage systems to retrieve data for anything from Natural Language Processing to the most complicated machine learning deployments, Extended Reality use cases, and much more.
Competitive storage solutions underpinning modern cognitive computing efforts are characterized by a number of different traits, including:
- Performance: In terms of storage, the speed at which data is accessed for AI almost directly correlates to how performant a storage system or unit is—especially “with flash, performance is important,” acknowledged Nexsan CTO Surya Varanasi.
- Capacity: Capacious storage systems can accommodate the rapid growth of data organizations are contending with, conservatively estimated by Varanasi to occur at a growth rate of “10 to 20 percent every year.”
- Backups: With ransomware and other forms of malware attacks rising in incidence and severity, it’s critical to leverage storage that functions as targets for backups.
- Data Migration and Disaster Recovery: In today’s increasingly hybrid and multi-cloud reality, it’s necessary to leverage storage options that excel at migrating data between settings. Disaster recovery is just one of many use cases for swift and accurate data migration in which organizations can avail themselves of “a copy in the cloud in case you have a disaster and both [on-premise] sites are down, you’re able to get back your data,” Varanasi mentioned.
Fast storage with these qualities supports some of the most vital cognitive computing deployments today in verticals as varied as the Department of Defense and financial services.
Facial recognition is an application of computer vision in which advanced machine learning models (many of which involve deep neural networks) can identify people based upon their facial features. Although the individual use cases for facial recognition technologies widely differ, they’re used to authenticate people in certain military and civilian settings. In the former that authentication may be used “at points of entry: you want to quickly look and figure out who a person is and get that data to an…officer,” Varanasi commented.
That officer can then verify the identity of the prospective entrant or decline him or her based on the results of the analytics involved. In civilian settings in certain countries, facial recognition is leveraged to authenticate users onto networks. In the Department of Defense example Varanasi referenced, “time is really important.” Storage systems are responsible for not only housing “two petabytes worth of storage,” Varanasi said, but also for quickly retrieving that data for this time-sensitive use case. Another vital attribute of storage systems in this type of deployment is using options that are removable and portable to secure locations at a network’s edge.
Virtual Reality (VR) is a cognitive computing means of creating immersive data visualizations in which users can surround themselves with their data to experience them as close to firsthand as possible. Numerous organizations, “including financial services,” Varanasi revealed, access VR through Managed Service Providers (MSP). The most common example of a MSP is a cloud provider. According to Varanasi “very dense storage is used to serve a lot of applications,”—which is another vital requisite for provisioning VR capabilities in cloud settings.
In fact, the density of a storage system is an important factor relating to its performance, which is immensely improved when “you can put 10 petabytes in a very comprehensive and dense system,” Varanasi disclosed. With so many different users accessing various apps from MSPs, this type of storage is necessary to underpin “a separate cloud, if you will, that’s hosting VR applications,” Varanasi said. The density of the underlying storage solution is critical for enabling users to access the variety of apps MSPs host for consumers.
Bringing AI to Life
Ultimately, storage’s utility for cognitive computing is binary. Storage units are necessary to house the massive quantities of data AI systems need to consistently deliver enterprise value. Additionally, their speed and overall performance is an intrinsic part of leveraging such solutions in a timely manner.
As Varanasi alluded to in the DoD example, facial recognition systems must work expeditiously to preserve the integrity of the environments they’re deployed to secure. In-memory technology and the computational gains of modern IT systems are relatively worthless without credible storage units to quickly retrieve the data on which cognitive computing applications depend.
Featured Image: NeedPix
Jelani Harper is an editorial consultant servicing the information technology market. He specializes in data-driven applications focused on semantic technologies, data governance, and analytics.