IBM said on the 5th that STRADVISION, an AI vision recognition technology corporations for Autonomous Driving vehicles, adopted the company's storage deep archive to store large-scale research data long term and use it quickly.
The build includes the IBM Diamondback tape library. The company said it switched to a tape-based infrastructure to reduce the recovery delays and repeated access expense that occurred with its existing cloud archive.
STRADVISION has continuously accumulated video and sensor data to advance Autonomous Driving algorithms. However, in the previous environment, even with low storage unit costs, data recovery took more than 10 hours, and expense increased with each re-access, heavily burdening research schedules and operating costs. The new archive starts at an initial 10PB and can expand to about 30PB without downtime simply by adding tapes. It also secured the durability required for long-term storage and stable write performance.
Researchers also felt significant changes on the ground. According to the company, recovery times were drastically reduced compared with before, cutting procedures that took up to several hours to within 5 minutes. With model retraining and validation becoming effectively immediate, bottlenecks eased, it said. Support for the S3 interface also lowered adoption barriers by allowing data to be operated in a way similar to existing cloud environments.
Yoon Jae-min, head of IBM Storage, said, "In large-scale research environments, stability and expense predictability are important," and added, "We will help corporations reduce their data burden and focus on core research."
Founded in 2014, STRADVISION supplies object recognition solutions to 13 domestic and overseas OEMs and 50 vehicle models.