In the face of real-time recommendation and real-time risk control in the Internet and financial industry, which integrates streaming and batch processing, a high-performance CPU+24 LFF drives + 4 NVMe or 24 LFF drives + 8 NVMe (8U.2/8E1) configuration. s) The drive has been deployed. It offers the ultimate edge computing solutions power, supporting up to 120 computing cores and up to 100GB of network bandwidth. It can provide a maximum 3:1 HDD-SSD ratio, meeting the needs of tens of millions of IOPS while also providing hundreds of terabytes of data storage. The use of NVMe caches increases the speed of real-time data analysis by 26%.
Take autonomous driving as an example; its data processing includes the acquisition and annotation process. Self-driving cars with an automation level of L3 or above generate 5 to 8 terabytes of data per day. This requires distributed storage to support strong scalability. In addition, the development of autonomous driving features involves a lot of training. Distributed storage improves the efficiency of GPU data access by sending packets from multiple storage nodes to the central control system. The KR2266V2 edge computing solutions provide maximum data access bandwidth and higher I/O throughput while meeting storage capacity requirements meeting the application requirements in the field of autonomous driving.
For data lake library scenarios involving massive data storage, intelligent analysis, marketing trend analysis, inventory analysis, and AD hoc query, the configuration of CPUs LFF hard disks with moderate edge computing solutions power is adopted. The 24 hard disks can be installed in a RAID card to make full use of the disk performance and achieve more stable and efficient system data transmission. The storage performance of a single server is 116% higher than that of the general 2U product with 12 hard disks.
For a storage-compute separation scenario, the configuration of CPUs LFF hard drives with low edge computing solutions power can provide twice the storage density of general-purpose servers, save more than half of the data center space, and reduce storage power per TB by more than 45%.