Qumulo added a predictive cache function of NeuralCache improving performance to its cloud of cloud data.
The Cloud Data Fabric (CDF) was launched in February and has an kernel standard of data and central objects with coherent covers at the edge. The nucleus is a cluster for storing data and distributed objects that runs on most systems, suppliers or public cloud infrastructure. The consistency between the nucleus and edges sites comes from the consciousness of the file system, from replication in the block, distributed locking, authentication of access control and journalization.
NeuralCache uses a set of supervised AI and automatic learning models to dynamically optimize reading / writing cache, with Qumulo saying that it “offers an unequaled efficiency and scalability in cloud and on -site environments”.
The CTO Kiran Bhageshpur declares: “The NEURALCACHE QUMULO redefines the way organizations manage and access massive data sets, dozens of petacts on the scale of exubytes, by adapting in real-time factors to multi-varied factors such as users, machines, applications, date / time of the network and the conditions of the clouds. “
Neuralcache, “says Qumulo:” Setting continuously depending on real -time data models. Each cache or lack of cache refines the model, improving efficiency and performance because more users, machines and AI agents interact with it. ”
He “stacks intelligently and combines object scriptures, minimizing API costs in public cloud environments while optimizing I / O reading / writing cycles for local deployments – offering significant cost savings without compromising sustainability or latency”.
Neuralcache software “automatically propagates the modified data blocks in response to any writing on cloud data fabric” and “users, machines and AI agents always access the most recent data.”
Bhageshpur claims that “improves applications performance and reduces latency while guaranteeing data consistency, which changes the situation for industries that are based on working data flows, including AI research, media production, health care, pharmaceutical discovery, exploratory geophysics, space and orbital telemetry financial services. “
Qumulo says that NeuralCache excels in overall data scales of 25 bp to several exubytes, “learning and improving the volume of data and the complexity of the workload develop”.
This predictive cache software was included in the version of the February CDF, but a spokesperson for Qumulo told us that “was not entirely live and we generately refer as” predictive cache “. Since then, we have done the customer test and provide comments like a beta test. And we officially appointed him NeuralCache.
Interestingly, the supplier of high -end storage tables Infinidat has a chatting function which is named in the same way, but based on the dram of its table controller. In June 2020, we wrote that its Table A “data taken from a data taken from a memory cover using a neurons cache engine with predictive algorithms … The neuronal cache engine monitors accessible data blocks and prefers adjacent blocks in DRAM.” It allows more than 90% of the reading data readings to be satisfied with memory rather than much slower storage players.
Despite the similarity of name, however, the technology of NeuralCache de Qumulo is distinct from Infinidat patented Neural cache technology
The Qumulo NeuralCache is available immediately as part of the latest version of the supplier software and is transparently integrated into the Qumulo Cloud data fabric. Existing customers can spend there without downtime. Learn more here.
