hotline：+86-0755-23762446
Server hardware architecture providerFocus on server hardware sales and various storage customization services
- High-performance storage solutions
发布时间：2018-05-28 阅览次数：7798 次
Application background
In recent years, in the area of server storage, the performance of components such as processors, memory, network cards, and switches has increased substantially. They are no longer the bottleneck that restricts the performance of the entire system. So where is this bottleneck? The answer is storage. For storage arrays, how to deal with application environments with extremely high performance requirements is a problem that high-performance storage needs to solve. This type of application environment is usually presented as a combination of data-intensive and computationally intensive applications, such as large-scale data calculations encountered in education and scientific research, as well as enterprise data warehouses and data analysis applications.
solution
For storage systems, the major bottleneck is the read/write speed, especially random read/write. For this situation, adding SSD as the Cache in the existing IP-SAN architecture is a very good choice. In addition, the unique RQR technology, reconstruction IO, and service IO conflict resolution technology can also solve the problem of long global reconstruction time.
Program advantages:
1. The random read/write performance of small files in the storage system is greatly improved. For applications such as database and data analysis, the overall system processing speed is greatly improved. Hao Xiang Technology has optimized the hotspot migration algorithm to automatically place data with high frequency in the SSD Cache. When the hotspot data is accessed again, the access delay will be greatly reduced.
2. No need to dial, rely on RQR technology, the emergence of bad fast disk can avoid global reconstruction, even if multiple disks appear bad blocks, RAID can continue to work, eliminating unnecessary rebuild time, extend the service life of the disk.
3. Fast reconstruction, relying on the lustre cluster file system to solve the conflict between RAID reconstruction I/O and service I/O, greatly ease the monitoring of dropped calls during the rebuild, and shorten the time required for global reconstruction.
In recent years, in the area of server storage, the performance of components such as processors, memory, network cards, and switches has increased substantially. They are no longer the bottleneck that restricts the performance of the entire system. So where is this bottleneck? The answer is storage. For storage arrays, how to deal with application environments with extremely high performance requirements is a problem that high-performance storage needs to solve. This type of application environment is usually presented as a combination of data-intensive and computationally intensive applications, such as large-scale data calculations encountered in education and scientific research, as well as enterprise data warehouses and data analysis applications.
solution
For storage systems, the major bottleneck is the read/write speed, especially random read/write. For this situation, adding SSD as the Cache in the existing IP-SAN architecture is a very good choice. In addition, the unique RQR technology, reconstruction IO, and service IO conflict resolution technology can also solve the problem of long global reconstruction time.
Program advantages:
1. The random read/write performance of small files in the storage system is greatly improved. For applications such as database and data analysis, the overall system processing speed is greatly improved. Hao Xiang Technology has optimized the hotspot migration algorithm to automatically place data with high frequency in the SSD Cache. When the hotspot data is accessed again, the access delay will be greatly reduced.
2. No need to dial, rely on RQR technology, the emergence of bad fast disk can avoid global reconstruction, even if multiple disks appear bad blocks, RAID can continue to work, eliminating unnecessary rebuild time, extend the service life of the disk.
3. Fast reconstruction, relying on the lustre cluster file system to solve the conflict between RAID reconstruction I/O and service I/O, greatly ease the monitoring of dropped calls during the rebuild, and shorten the time required for global reconstruction.
