Update: Another source at Tesla says that the new Inifibox is not for the Autopilot program, but another source says that the new Autopilot is increasing Tesla’s need for data storage
Tesla’s next generation Autopilot/self-driving hardware suite has a lot more sensors gathering data than its previous generation. And with now ~25,000 new vehicles per quarter contributing to its growing fleet and gathering even more data, the company needs a solid data storage infrastructure to support its fleet learning capability.
In an effort to do just that, Electrek has learned that Tesla recently invested in a new state-of-the-art data storage system to support its new Autopilot program and the significant amount data expected to be collected through it.
Companies who acquire, store and analyze the most data, achieve the greatest competitive advantage. That’s actually the tagline of INFINIDAT, a data storage company and Tesla’s latest supplier, and it’s a statement generally accepted as a truth in the world of big data.
Tesla is certainly betting on it. The automaker has an aggressive timeline to bring self-driving vehicles to market, ~2018 vs ~2020 for most competitors, and the fact that it already collected hundreds of millions of miles of data through its first generation Autopilot program is seen as an important factor in Tesla’s confidence in being the first to market.
Its next generation hardware suite with 8 cameras, 1 radar, and 360 ultrasonics, is a significant step up and more vision based than the previous system, but Sterling Anderson, Director of Autopilot Programs at Tesla, confirmed that the data from the first program is still useful:
Tesla uses the data collected through the sensors of all the cars in its fleet to crowdsource high-precision maps of the driving environment and feed them back to vehicles in “map tile” when in the right area. The vehicles can then more easily navigate the environment while crosschecking with real-time data with the sensors and onboard processing power.
In order to store and analyze its growing fleet data, a source familiar with the transaction told Electrek that Tesla recently bought a new InfiniBox system (pictured above) from INFINIDAT. The company founded by Moshe Yanai, the inventor of EMC’s Symmetrix and IBM’s XIV storage systems, describes the InfiniBox:
InfiniBox, high-performance enterprise data storage, eliminates performance, availability and scalability issues to accelerate critical business applications. Based upon a fully abstracted set of software-driven storage functions layered on top of industry standard hardware, INFINIDAT delivers a fast, highly available, and easy-to-deploy storage system.
The source says that Tesla bought a single box, which starts at a minimum capacity of 2 petabytes of data and cost on average $1 million USD. The system is being marketed for applications in need of top reliability, like fleet-learning systems, and it features triple-redundant power and data paths, as well as triple redundant hardware, which enables 99.99999% uptime or less than 3 seconds of downtime per year, according to the company.
We asked Tesla for a comment and to confirm the new system, but the company didn’t respond to our inquiry. We will update if it changes.
The system is also energy efficient, which might not seem like a big deal since it’s not in the car making a dent on its efficiency and range, but big data has become so important that data centres now represent about 3% of global electricity consumption, which makes companies investing in new hardware more cautious about their decisions.
Tesla’s new InfiniBox could be put to use rather quickly as the automaker could deliver as many as ~15,000 vehicles with the new Autopilot/self-driving hardware suite by the end of the year and start feeding Tesla’s new data storage system. With Tesla’s production set to scale by 10x over the coming years and its data gathering/car also scaling, this is likely the tip of the big data iceberg.
Subscribe to Electrek on YouTube for exclusive videos and subscribe to the podcast.