The Yunqi Conference ended successfully, and Jike Tiancheng made a wonderful appearance
With the theme of "Computing, Evolution and Future", the 2022 Cloud Conference was successfully held in Hangzhou Cloud Town on November 3th.
NVIDIA, the king of GPU, made a wonderful appearance. As a deep partner of NVIDIA and the only recognized domestic storage manufacturer, Jike Tiancheng was invited to attend this conference.
Jike Tiancheng and NVIDIA Bluefield DPU cloud native storage products appeared at booth NVIDIA and interact with live partners.
As a leading high-performance distributed storage manufacturer in the industry, Jike Tiancheng has made key technological breakthroughs in cloud computing, distributed storage and other fields through continuous independent innovation, redefines the future storage architecture, and provides users with the innovation of cloud native data infrastructure architecture.
The time is right for data centers to change
The current single-block NVMe SSD performance can reach 1 million IOPS, which is very different from the HDD mechanical hard disk. Traditional storage hardware and software architectures are no longer suitable for carrying high-speed flash memory media, and they fundamentally restrict the productivity of new storage media. Full flash arrays have gradually shifted to full NVMe hardware, but its scalability and flexibility have always been the biggest bottleneck in the cloud computing era. Based on the all-flash memory media configuration, the distributed full-flash storage system adopts the standard hardware platform, efficient emerging storage protocols, and the extreme software optimization, which can provide simpler, economical, elastic and efficient data storage solutions.
NVMatrix is a distributed fully flash storage fully independently developed by Jike Tiancheng for high IO intensive application scenarios. Based on the efficient distributed storage software design and high performance full flash memory hardware platform, NVMatrix can be multiple node NVMe SSD resources through RDMA ethernet or Infiniband network into a high performance, high availability, easy extension block storage resources pool, and through NVMeoF, NFS Over RDMA and other high-speed storage network interface, can be widely used in all kinds of high IO performance requirements of business scenarios.
Summary of NVMatrix distributed full flash storage product advantages
Storage expands horizontally, expands on demand, and performance capacity increases with the number of nodes. The core code architecture of the product is based on the industry-leading hash algorithm, with deep code optimization in key links, including the optimization of network, disk processing efficiency, data stratification and caching mechanism. It makes the storage system can meet the requirements of high concurrency and high input and output efficiency, combined with the mainstream NVMe flash memory technology, by optimizing the IO path and configuring the general hardware of different performance, seamlessly adapt to the performance requirements of various application scenarios such as hot, cold and cold data switching. By combining the RDMA high-speed network technology to handle the internal communication of the storage clusters, the performance of the network storage is much higher than that of the local NVMe SSD.