Milan processor, corresponding to the instance star sea SA3 on the cloud, can provide up to 232 core single node computing power, which can further meet the user’s needs for computing performance in video coding and decoding, offline rendering, database, AI and other scenarios. Among them, in AI scene, the measured performance of star sea SA3 has been improved by more than 220.
Compared with the previous generation SA2 servers, SA3 has better performance in security, heat dissipation and so on. The new sev-snp feature can protect the data security more comprehensively; in terms of heat dissipation, the air cooling cooling capacity is improved by 22, and compatible with the liquid cooling design, which can reduce carbon emission by 8.
In terms of storage, SA3 adopts the latest generation CBS block storage engine of Tencent cloud, with the storage capacity of the whole machine up to 8GBs and IO delay shortened by 63, which can better support the application with low delay and high stability.
In terms of network performance, the bottom layer of star sea SA3 is based on the new 100g network architecture, providing up to 100g of super large network bandwidth. Compared with the previous generation, the bandwidth of single node network is up to 400. Combined with the new optimization of network virtualization platform, the network packet forwarding performance has increased by more than 260. In the case of high traffic and high forwarding network application scenarios, it is not in the future.
It is understood that Tencent has more than one million servers running in the whole network, and it is the first and fifth company in China and the world to operate more than one million servers. At the same time, Tencent cloud peak bandwidth has also exceeded 100t, and service and traffic have entered the first tier in the world.