Shutterstock
Qumulo, an enterprise data management and storage company, has unveiled a new predictive caching solution called Qumulo NeuralCache. The new tool is designed to supercharge data performance for AI-driven enterprise applications and critical line-of-business workloads. It leverages AI and machine learning (ML) models to dynamically optimize read/write caching across cloud and on-premises environments.
The company claims Qumulo NeuralCache is the “world’s most advanced file system” and it “accelerates data access up to three times faster”, however, it did not provide any specific baseline or comparison point for this improvement.
According to Qumulo’s Chief Technology Officer Kiran Bhageshpur, NeuralCache has the potential to redefine how organizations manage and access massive volumes of datasets, operating at 25 petabytes and beyond.
One of the key capabilities of the tool is that it can adapt in real-time in response to various factors including users, machines, applications, date/time, system state, network state, and cloud conditions.
Every cache hit or miss feeds back into the system, allowing it to refine its behavior continuously. As more users, machines, and AI agents interact with data, NeuralCache becomes smarter and more efficient, according to Qumulo. This helps deliver performance improvements without manual intervention.
“This predictive model-based caching system enhances application performance, and reduces latency, while ensuring data consistency, making it a game-changer for industries relying on data-intensive workflows,” emphasized Bhageshpur.
With hybrid cloud environments in mind, NeuralCache is designed to optimize data operations by intelligently handling object writes and minimizing API charges in public cloud environments. This feature can be useful in cost savings without compromising on durability or latency.
As NeuralCache automatically updates data blocks when changes are made, it minimizes the risk of outdated or incorrect information in the decision-making process.
The tool is fully integrated into Qumulo Cloud Data Fabric, the company’s unified hybrid cloud storage solution, ensuring comprehensive data consistency and accuracy across its entire ecosystem. For existing Qumulo customers, NeuralCache is available immediately as part of the latest software release.
Traditional predictive caching systems depend on fixed rules or predefined policies to decide which data to store for faster access, making them rigid and less adaptable. NeuralCache leverages machine learning to make dynamic caching decisions, overcoming these limitations.
Carbon VFC, one of the early adopters of NeuralCache, uses the tool to support its data-intensive visual effects and motion graphics projects. “With NeuralCache, everything just works like it’s supposed to – no more additional configuration or monitoring of cache policies, and no more cold start delays,” said Carbon VFX Director of IT Jeff Drury.
(amiak/Shutterstock)
“Before NeuralCache, our artists would expect things to be slow every morning while the system cached the data. Now, they just sit down and get to the creative work. Qumulo allows us to scale to our needs versus completely overbuilding a solution in order to keep pace with the speed of our artists and the complexity of their workload, saving us time, costs, and frustration. It just works.”
Qumulo launched the Cloud Data Fabric file system earlier this year, with the aim of simplifying the management of unstructured data across cloud and on-premises environments. The Cloud Data Fabric provides a unified global file system and integrates edge caching technology to deliver fast and consistent access to data, even across dispersed networks.
With the introduction of NeuralCache, Qumulo builds on this foundation by enhancing caching capabilities specifically tailored for data-intensive workflows. This addition is particularly suited for AI and ML applications, where rapid data access is critical.
NeuralCache represents a significant advancement in enterprise data infrastructure. It aligns with the growing trend of intelligence storage, where instead of passively holding data, these systems learn how to use it and adjust performance accordingly.
