Weka today withdrew the cover from its latest product, Neuralmesh, which is a re-introduction of its distributed file system, which is designed to handle the expanding storage and serving needs-as the latency requirements and resistance-
Weka described Neuralmesh as “Fully containerized architecture based on a tap, which smoothly combines data, storage, calculation and service AI.” It is designed to support the data needs of extensive AI deployment, such as food and token warehouses, especially for the AI agent’s workload that uses the latest thinking techniques, the company said.
This agency load has different requirements than traditional AI systems, including the need for a faster response time and another overall workflow that is not based on data but on the service. Without the types of changes that Weka has built into Neuralmesh, traditional data architecture burden with organizations with slow and ineffective AI workflows.
Liran Zvibel is the CEO and co -founder of Weka
“This new generation of workload AI is difficult to different than anything we have seen before,” said Liran Zvibel, co -founder and CEO in Weka in a video published on the website of his company. “Traditional high -performance storage systems reach a turning point. What worked great in the old HPC now creates narrow places. The expensive GPU is sitting idle and waiting for data or unnecessarily counting the same chips over and over again.”
With Neuralmesh, Weka is developing a new layer of data infrastructure, which is oriented to services, modular and composible, Zvibel said. “Think about it as a software -defined production that connects data, calculates and AI services across any environment with extreme accuracy and efficiency.”
From an architectural point of view, Neuralmesh has five components. They include a core that provides basic software -defined storage environment; Accelete, which creates direct paths between data and applications and distributes metadata across the cluster; Deployment to ensure that the system can be operated anywhere, from virtual machines and bare metals to clouds and on-pm systems; Observe which provides manageability and monitoring of the system; and corporate services that provide security, access and protection management.
According to Wek, Neuralmesh receives computer clustering and data concepts. It uses more parallelized paths between applications and data and distributes data and metadata “intelligently”, the company said. It works with clusters with managed CPU, GPU and TPU, runs on prem, cloud or anywhere between them.
The times of accessing data on Neuralmesh are measured in microseconds rather than in milliseconds, the company claimed. The new offer “dynamically adapts to the variable needs of AI working procedures” using a microservis that processes various functions such as access to data, metadata, audit, observability and protocol communication. These microservisses run independently and are coordinated through the API.
Weka argued that Neuralmesh is now quickly and more durable, as the data and workload of AI increase, the company says. This performance is partly achieved due to routines for data striking it used to protect data. As the number of nodes in the NeuraSh cluster increases, the data is more flexible to multiple nodes, reducing the probability of data loss. When it comes to scalabibility, Neuralmesh can scalance up from petabytes to exabyt storage.
“Almost every layer of a modern data center has accepted services -oriented architecture,” wrote the Chief Product Director of Weka Ajay Singh on the blog. “Compute is supplied via containers and functions without servers. The network network is managed using software platforms and service networks. Observation, identity, security and even AI inference pipes run as modular scalable services. Databases and layers are offered as fully managed systems.
Related items:
Weka keeps GPU powered by fast new appliances
Legacy Data Architectures Holding Genai Back, Weka Report Finds
How to earn on the software defined storage, safely and suits
(Tagstotranslate) agent AI