technologyreseller.co.uk 39 DATA storage How the channel is helping VAST Data to extend its reach that is already known, to AI-assisted discovery when machines can recreate the process of discovery by capturing, synthesising and learning from data. You can find out more about its vision at https://buildbeyond.ai/special-event. VAST also has applications for businesses with less ambitious goals as Technology Reseller discovered when we caught up with the company’s recently appointed Vice President of Channels Greg Machon. Although Machon had been at VAST for less than three months when we spoke, he said he had been following the company since 2019 when it was in stealth mode and ready to announce itself to the world. “I was fascinated with the idea that VAST was going to create one massive pool or volume where people could store their unstructured data, whether it was NFS (Network File System) or SMB (Server Message Block) or S3 (Amazon Simple Server Storage), and dramatically simplify things and drive down costs for companies that would normally store their data on a combination of flash and/or hard disk drives. “I was fascinated by how you could create this highly performant platform on all-flash storage class memory with the economics (or better) of a high density disk environment or at least a hybrid environment. At the time, I said it sounded almost too good to be true.” Four years later and Machon reels off a list of the benefits of VAST’s platform, including a massive acceleration in performance and a 25% saving in compute and power costs due to the use of all-flash storage-class memory, plus cost savings of 66% from new storage algorithms, including VAST’s own patented similarity-based data reduction technology. “What's interesting from a data reduction standpoint is that we are seeing environments that have already been deduplicated and compressed and on our platform we are still able to get anywhere from 1.5X to 4X data reduction savings. We take a lot of data and compress that to the smallest amount so a customer doesn’t need to pay for more software licensing or more hardware to be able to store that data.” Machon adds that having a simplified, unified and higher performing platform enables companies to do interesting things with unstructured data using training and inference models and language learning models (LLMs). Hence VAST’s claim, even before its most recent announcement, to be the data platform for the AI era. “No longer do customers have to have silos for their tape environment or their archive environment, for their low-cost hard disk drives or for their upper tier performance or flash media. They can put all that into one big volume and allow these models and algorithms and AI machines to look at and gain insights from that entire pool of data far more quickly.” This message clearly resonates with On August 1, VAST Data, self-styled data platform for the AI era, revealed how it was bringing its vision to life with the launch of a ‘transformative’ data computing platform that it believes will accelerate AI’s ability to find solutions to humankind’s greatest challenges in healthcare, climate science, agriculture and other areas. Designed for the deep learning era, the VAST Data Platform builds on and expands the company’s existing Disaggregated Shared-Everything (DASE) distributed systems architecture to enable companies that have built large-scale AI clusters to automate the discovery of new ideas and solutions by interacting with data, including text, images, sound, video and instrument data, in a more intelligent way. Key advances include the creation of one global namespace, the VAST DataSpace, spanning public cloud, private cloud, edge and devices (and all geographies); the ability to run SQL queries on file and object metadata and cross-correlate between different data sources; and the ability automatically to run functions on unstructured data such as data preparation, data cleaning, data anonymisation, data inference and data training. VAST expects these new capabilities to ease the transition from generative AI and Large Language Models (LLMs), which it says are limited to performing routine tasks or reciting information Quality before quantity Continued... Greg Machon Having a simplified, unified and higher performing platform enables companies to do interesting things with unstructured data using training and inference models and language learning models
RkJQdWJsaXNoZXIy NDUxNDM=