Analysis of HPC projects in Tsinghua University

Tsinghua University, the highest school in China, the dream of the university,1932, its mechanical Engineering Department (hereinafter referred to as the mechanical department) was established, is one of Tsinghua University's first engineering department. It has a number of national key laboratories, in recent years, and materials science, automation technology, information technology, laser technology and other disciplines to work together a number of scientific research projects, in the academic reputation, research level, research team, experimental equipment and so on have a top technology and facilities.

Mechanical Department of Daily Teaching and research projects need to be applied to fluent, Vasp, Gaussian, Ansys software to achieve a variety of high-performance computing, so will produce a large number of data, which also on the storage system in high concurrency, reliability, sharing will have high requirements. However, as far as existing mechanical system storage systems are concerned, their rapid development needs are not met.

Chuzhi A2000 cluster Storage solution

Chuzhi technology provides a Chuzhi A2000 cluster storage solution for high performance computing requirements in mechanical systems. The Chuzhi A2000 cluster storage solution consists of two parts of the metadata storage node and the intelligent storage node.

The Metadata storage node manages the metadata of the file system (including file directory tree organization, attribute maintenance, file operation logging, authorized access, etc.), manages the namespace of the entire storage system, and provides a unified system image storage pool externally.

The Intelligent storage node is used to store the user's actual data and realize the cluster storage function, mainly including storage device management, network equipment management, authority management, etc.

Chuzhi A2000 cluster storage system uses advanced distributed and parallel dual-group storage architecture, multiple metadata storage nodes and intelligent storage nodes separate and parallel work, greatly improve the reading and writing efficiency of mass files, and easily realize massive data management. At the same time, A2000 has outstanding advantages in high concurrency, high reliability, integration and expansibility, which satisfies the requirements of mechanical system for high performance computing applications.

The continuous operation of high-performance computing applications will produce more and more data, which requires the storage system to have sufficient capacity to meet rapid growth requirements. Chuzhi A2000 cluster Storage System single volume supports petabytes of storage space and helps users efficiently manage Bai or even more files. At the same time, the system can also implement the file global namespace, all compute nodes can see a consistent view of the file system (administrators can also set permissions to allow users to see only their own application data), the global sharing of such data can enhance collaboration between different servers in the same application and improve the service efficiency of the application.

High concurrency

In the mechanical system storage environment, the application type is many, the concurrent access degree is high, this brings the great pressure to the storage system. A2000 cluster storage System meets the requirement of high concurrent access by means of simultaneous access to data by multiple storage nodes. On the application server side, when the application writes files to the storage system, the files are partitioned onto multiple storage nodes, and when the file is read, the system reads the data from multiple servers concurrently. As a result, a large number of data I/O requests are dispersed across multiple storage nodes, the aggregation bandwidth is added by the bandwidth of multiple storage nodes, which makes the disk performance and network bandwidth of all storage nodes fully utilized, and overcomes the performance bottleneck caused by the single outlet point of the original system. The high concurrent access requirements of users are well implemented.

High reliability

The A2000 cluster storage System has a complete recovery process, which can automatically recover the data when it fails, and effectively eliminates the single point of failure. For users, any hardware failure requires no manual processing, and their only management is to add hardware and replace damaged hardware when capacity is insufficient to ensure that the A2000 cluster storage system provides stable storage services uninterrupted.

Easy to integrate and easy to expand

A2000 cluster storage systems support existing heterogeneous storage systems, using the latest management modules to seamlessly connect to traditional storage architectures. At the same time, it can be seamlessly compatible with many different operating systems. In addition, A2000 cluster storage System can support the dynamic expansion of storage capacity, that is, using configuration tools to dynamically add storage nodes to increase the capacity and scale of the system (no need to interrupt any application), and as the storage node data increase, the entire system's aggregation bandwidth will be linearly increased, to achieve automatic load balancing, Fully meet the capacity and performance requirements generated by the continuous development of the business.

Easy to manage

In addition, Chuzhi A2000 cluster storage System with easy-to-use graphical management and monitoring interface, to support users through a dedicated security channel for remote configuration and system-wide monitoring, intelligent detection of all kinds of hardware and software failures and unified summary, to ensure that users can deal with the failure in a timely manner, but also reduce their search for trouble in the workload.

Chuzhi A2000 cluster storage solution help Tsinghua University Machinery department to set up a set of perfect and efficient high-performance computing storage System, has been recognized by users and praise. By using the Chuzhi A2000 cluster storage solution, Tsinghua University Machinery department not only in the storage utilization, high concurrency, high reliability and other aspects have been greatly improved, with a more powerful computing power, and through the storage system strong integration and ease of management to promote, but also effectively reduce the procurement costs and it operating costs.

"Responsible editor: Iris Wei TEL: (010) 68476606"