50 Technology focus | Data storage while allowing efficient data sharing when needed. Private namespaces, combined with SR-IOV, ensure data can only be accessed by the virtual machine or host attached to the namespace, maintaining critical data privacy and security. The SR-IOV capability is combined with a PCIe Generation 4 interface and ruggedised automotive design, with random read and write speeds above 600,000 I/O operations per second (IOPS) and above 100,000 IOPS, respectively, for a 4 KB transfer. This allows the drive to efficiently manage data streams from multiple SoCs at once. For example, an automotive OEM can connect one port to an ADAS system and another to an IVI system, allowing each to store private data while accessing a common set of critical map data and reduce the cost per gigabyte of storage. This boosts performance and dissolves bottlenecks as both can access shared data simultaneously, and it alleviates the need to keep redundant copies of data. The quad ports also reduce the need for extra storage devices. As existing solutions can typically connect to only one SoC on their own, they tend to get placed locally with each automotive system, leading to unused capacity; alternatively, they require OEMs to use a costly, automotivegrade PCIe switch to connect a drive to multiple SoCs. By centralising storage for multiple systems, the SSD dramatically improves efficiency while streamlining architecture. The SSD is built with its triple-level cell (TLC) NAND, but it can be configured to support SLC and high-endurance (HE-SLC) data endurance groups, which offer 20 times and 50 times the endurance of TLC, respectively, to meet unique data requirements. For example, the HE-SLC endurance group can be used for heavy-write use cases, such as continuous blackbox data recording, where cars must constantly rerecord critical data from sensors, cameras and Lidar. In such a scenario, where data is being programmed and erased every few minutes, the HE-SLC mode offers the required endurance while eliminating the need for more expensive, volatile memory, such as DRAM. The SSD is designed to ASIL-B to support safety-system requirements, and it is offered in a ball-grid array package to help withstand the shock and vibrations typical of vehicles’ rugged environments, with capacities up to 1.8 TB to enable efficient storage of AI algorithms, large language models, advanced infotainment and telemetry data for the next generation of vehicles. Again, security is key, and the SSD includes hardware-based data encryption, device attestation, secure boot and cryptographically signed firmware. Conclusion The evolution of AV design is having a major impact on data storage. The move to a centralised architecture presents opportunities for multi-port storage, saving on costs and wiring weight, but is a challenge for the speed of retrieval and capacities. While NOR memory supports code storage, the latest NAND memory technologies and UFS standards are driving storage for AI and sensor-fusion data forwards. Acknowledgements With thanks to LJ Chen and Mehak Kalra at Synopsys, Tetsuya Yamamoto at Kioxia and Barbara Kolbl at Micron. JAPAN Kioxia +49 211 5382 7171 www.kioxia.com KOREA Samsung +1 800 726 7864 www.samsung.com TAIWAN Innodisk +886 2 7703 3555 www.innodisk.com US Collins Aerospace +1 319 265 5467 www.collinsaerospace.com Curtiss-Wright Defense Solutions +1 704 869 4600 www.curtisswrightds.com Kingston Technology +1 877 546 4786 www.kingston.com Lexar +1 877 747 4031 www.lexar.com Micron Technology +1 408 855 4000 www.micron.com Seagate +1 405 324 4714 www.seagate.com Synopsys +1 650 584 5000 www.synopsys.com Western Digital +1 408 801 1000 www.westerndigital.com Some examples of data suppliers June/July 2024 | Uncrewed Systems Technology
RkJQdWJsaXNoZXIy MjI2Mzk4