Fujitsu – Vexata Partnership for AI/ML
Fujitsu started cooperation with emerging startup Vexata Corporation. The San Jose company’s solution will allow the Japanese manufacturer to reach customers using the latest technologies.
Fujitsu starts cooperation with the Vexata
During a press meeting in Sunnyvale, where Fujitsu Laboratories of America is located, representatives of both companies talked about cooperation plans and presented Vexata all-flash arrays. So far, this solution is addressed to a small group of recipients, including organizations using artificial intelligence or machine learning mechanisms on a wider scale. In recent years, the environment of the cognitive era is stepping up more and more often, in which self-learning IT systems play a prominent role. While there are quite a few discussions about the challenges related to computing power, the issues of data storage are in the further place. Meanwhile, Vexata believes that mass storage solutions available on the market are not adapted to work in a cognitive environment.
– Working with cognitive systems, handling high frequency transactions or risk analysis and fraud require higher performance than traditional all-flash arrays. – explains Ashish Gupta, CMO Vexata.
Vendors do not stand still and are constantly looking for ways to improve the technical parameters of the devices. One of the methods is installing in NVMe storage arrays, although Ashish Gupta thinks that this is not a perfect concept.
– NVMe improves the internal performance of the system, but can burden traditional storage controllers, which in most cases processes both input / output operations and is responsible for traffic control. – explains Ashish Gupta.
Vexata has separated the data and control paths that entrust the implementation of the I/O FPGA (Field Programmable Gate Array). Each storage unit has an FPGA system, the manufacturer calls it a VX-OS router, increasing the system throughput.
– The VX-100M matrix with Intel Optane 3D XPoint media reaches a capacity of 7 million IOPS, the delay is 40 microseconds, and the bandwidth is 80 GB / s. These parameters are better than the traditional all-flash arrays. – adds Ashish Gupta.
Vexata can use a standard X86 processor to manage all VX-OS control functions, maintaining the right performance while reducing costs. This is an interesting option, for example for Oracle users, where the license fee depends on the number of cores on the device.
– 70 percent mass storage systems are media costs. All producers buy the same disks, paying the same price for them. We stand out with innovative architecture. – says Ashish Gupta.
Not without reason, Fujitsu has invited journalists to his laboratory in Silicon Valley. The newly-launched center is to be a place where a Japanese producer plans to build a technological and business ecosystem together with startups. The projects are to be related to a hybrid cloud, the Internet of Things or artificial intelligence.
– Cooperation with Vexata will allow customers to go the way from traditional IT solutions BI to AI. With the passage of time, we will observe the migration from the current architecture, in which arrays support four layers of data for a model typically limited to three: cognitive, traditional, and backup and archiving – explains Alex Lam Vice President & Head of the North America Strategy Office at Fujitsu.
During the meeting, Fujitsu presented a reference architecture consisting of three Primergy RX2540 servers, two Mellanox SN2700 100GbE switches and the Vexata VX-100F NVMe system.