Cloud computing has thoroughly transformed the business model dominated by local computing. It rids business departments in an enterprise of the limitations of various factors such as hardware, software, region and time, and enables them to devote to their own business innovation more flexibly and passionately. Acceleration of the business process and improvement of the service quality of these business departments render the enterprise more competitive.
In a super-large-scale cloud data center, normally thousands or even tens of thousands of servers need to be deployed and maintained. Cloud service providers need to comprehensively monitor the throughput capacity and health status of the entire data center. Higher requirements are thus proposed for not only the performance, but also compatibility, stability and maintainability of the storage products.
A large domestic e-commerce company has adopted the high-capacity NVMe SSD of PBlaze Series to replace the traditional HDD and SATA SSD with a hardware ratio of 1:5, which greatly improves the performance and capacity of a single disk/card, drastically reduces the space occupied by storage devices and energy consumption per unit capacity, and saves operation and maintenance costs. Additionally, it is easy for operation and maintenance personnel to carry out management and maintenance.
In the CDN application scenario dominated by read requests, the read performance is enhanced by 6 times.
The traditional SATA SSD is replaced with a ratio of 1:5 to reduce the space occupied by devices and give full play to the efficiency of CPU, network and energy.
The application speed of random small I/O is accelerated, which enables the Docker to respond within several seconds—a significant improvement.
Separate storage is conducted for hot and cold data, which shortens the response latency of data and drastically betters user experience.
Efficient compaction algorithms are achieved based on the all-flash storage system developed by the customer. Memblaze provides data storage services with high bandwidth, high IOPS, low latency and high capacity for core databases such as Oracle RAC and RadonDB.
Relying on all-flash configuration based on PBlaze NVMe SSD and RDMA network, a single disk (volume) offers up to 100K read/write IOPS (4K random read/write), and presents a latency of 90μs (4K random write).
PBlaze Series adopts the NVMe standard protocol, which effectively reduces the complexity of NeonSAN infrastructure and the risk of faults due to standard difference. In addition, PBlaze Series supports SPDK technology, which further lowers the access latency and improves the user experience.
Eight U.2 PBlaze NVMe SSDs not only provide superb performance, but also offer an available capacity up to 100 TB. Relying on such advantages, NeonSAN is able to provide highly stable and easy-to-use storage services with high performance and high capacity for the applications such as upper-layer databases.
The customer, by adopting the VMware vSAN solution, provides a scalable storage resource pool composed of standard components for its cloud platform services. It configures a dual-disk group consisting of HDDs and PBlaze Series NVMe SSDs at each vSAN node to optimize the overall I/O performance of the vSAN system.
The solution integrating PBlaze Series NVMe SSD + VMware vSAN facilitates unified configuration and management of hosts and storage resources, and optimizes the resource ratio, thereby significantly lowering the hardware and software purchase costs of the solution.
The solution offers excellent I/O performance, three times that of the original storage solution, and it also supports flexible horizontal and vertical performance extension.
Old servers are re-organized into distributed object storage clusters to make full use of the hardware resources, improve the return on investment of equipment purchased, and further save costs.
By adopting the ServerSAN distributed storage architecture, each server node is configured with the PBlaze Series NVMe SSD and 12 SATA SSDs (4TB), greatly boosting system reliability and storage performance.
In the simulation test of hard drive failure, the time needed for the reconfiguration of each TB of data is reduced to 30 minutes, only 1/25 that of the traditional FC SAN.
The test on the same virtual machine corresponding to the same physical machine shows that data processing with the Server SAN architecture is more than four times faster than that with the FC SAN.
A local tax bureau customer creates virtual machines at the nodes with the vSphere, and protects the application with Virtual SAN disaster recovery strategies. Finally, it adopts the VMware virtualized cluster as the platform basis of the Third Phase of the Golden Tax Project and builds the data center with general hardware. Eight PBlaze Series NVMe SSDs are deployed (one for each node).
The performance of the Virtual SAN platform based on PBlaze Series NVMe SSD grows linearly as the number of nodes increases. The 8-node Virtual SAN provides a storage space of 32TB, outperforming the existing storage systems with the performance nearly three times that of the previous system.
The solution, while improving the system performance, no longer requires purchasing new centralized storage systems and servers, reducing the overall purchase and maintenance costs by 50%.
The existing local tax services are smoothly moved to the Virtual SAN, and new services are successfully deployed as well. Thus, the services of the local tax system are upgraded.
