The high-performance computing node cluster system should be equipped with the global shared memory system to centrally store and manage users’ data. The central memory is connected through an IO node and the central memory is shared via NFS in the traditional solution, but when the cluster scale is larger, tens of or hundreds of computing nodes simultaneously access IO node via the network, the network exit bandwidth of IO node will be competed. IO node can not process too many IO requests, so the load is too high and IO is blocked. The Sugon ParaStor distributed parallel memory system is recommended for a high-performance computing center. The parallel memory system can integrate the discrete memory resources and redirect the data access to multiple IO nodes, so it can avoid bottleneck of single IO node. The Sugon ParaStor distributed parallel memory system uses InfiniBand high-speed network as the sharing network to store data. The whole memory system can reach very high concurrent bandwidth.
Tiankuo I620-G30, developed by Sugon based on Intel ® xeon ® extensible processor platform, has extensive uses as a two-way server. I620-G30, which is fully self-developed, possesses superior computing performance and powerful IO expanding capacity, realizing high performance and high expansibility to the maximum extent, so it can meet the needs of all industries. I620-G30 server is very suitable for finance, securities, transportation, post, telecommunication, energy, Internet and other industries with data centers and remote business environment which is demanding on the server performance, extensibility and reliability.