On November 6th, At Tencent’s Techo Developer Conference, Xie Ming, Vice President of Tencent Cloud and General Manager of Cloud Architecture PlatformS, disclosed Tencent’s infrastructure evolution and innovation for the first time. According to him, behind the continuous development of WeChat, QQ and other national-level businesses, including servers, networks, IDC, computing, storage, databases and other infrastructure has been evolving, has entered a new stage of “born into the cloud.”
In terms of infrastructure construction, Tencent has more than 1 million servers, the first company in China with more than one million servers, and Tencent’s network bandwidth peak has exceeded 100T.
In the cloud era, Tencent’s infrastructure will continue to evolve in the direction of extreme resiliency and intelligence through self-study and customization, Mr Xie said.
In terms of servers, Xie Ming introduced, due to the higher demand for server performance, cost, security and so on, Tencent Cloud started the process of self-study server. Tencent’s self-research server Star Sea uses the hardware tamper-proof capability of trusted chain transmission, advanced hot siphon cooling technology and fineduct design and fan control to achieve a 30% reduction in TCO, a 50% increase in load energy efficiency, and a 35% improvement in the overall performance of examples.
Xie Ming mentioned that Tencent Cloud’s fourth-generation data center T-Block, in accordance with the building block approach, achieved the modular configuration and rapid construction of the whole data center, compared with the traditional large-scale data center, the site construction cycle is reduced by half, natural cooling technology can reduce PUE to less than 1.2 in the industry. With the T-Block solution, a campus with 300,000 servers could save 250 million degrees of electricity a year.
In terms of storage, Tencent developed distributed storage TFS for Q-zone album business scenarios as early as 2006, supporting the daily image upload of Qzone albums. At present, Tencent cloud object storage COS realizes the cluster all nodes to participate in the system throughput, disk utilization rate of more than 90%, while the system’s complete autonomy also greatly reduces the labor input of operation and operation.
With the rapid development of cloud computing, cloud computing users have higher requirements for the technology architecture of resource management granularity, resource scheduling and isolation capabilities, and compute/storage decoupling. Mr. Sherecallscalle calls this higher requirement “extreme resilience”.
Xie Ming believes that the ultimate elasticity is first of all performance and capacity to help large customers to carry the business flood peak, but also to allow small customers not to waste a penny, the need for computing, storage, network resources for fine-grained division. Second, we need to improve inventory management and scheduling capabilities and the corresponding product capabilities, really pay-per-use. Finally, the ability to deliver quickly, the need for computing, storage and other resources can be fully decoupled to meet the requirements of seconds up seconds down, fast up and down.
Tencent Cloud’s underlying products are also working toward extreme flexibility. Among them, Tencent Cloud built the third generation of universal computing platform service-free function computing can provide better flexibility, the real sense to achieve 100% resource utilization, the corresponding cost of users fell by more than 50%. It is worth mentioning that Tencent Cloud through the function concurrent monitoring data real-time calculation, early expansion, so that the function cold start rate reduced to less than one in 10,000. The self-developed lightweight virtualization technology, capable of starting a virtual machine in 35ms, has set a record for the fastest start-up of a virtual machine in the industry, and has been steadily applied to “small programs and cloud development”.
At the same time, Tencent cloud object storage COS also provides different geographical billing methods, users according to the use of the scene, on-demand choice of the most cost-effective solution. The database product CynosDB is based on the computational storage decoupling architecture, supporting large instances and enabling better resiliency. A single instance can reach a maximum of 100 TB levels, while storage sharing between multi-compute nodes can scale up to 15 compute nodes for the same instance.
Xie also believes that in the future, cloud computing will combine intelligence to provide better services, including intelligent operations and intelligent services, will better improve service quality and reduce service costs, for more users to provide VIP-level cloud services. Taking the database as an example, Tencent Cloud provides online performance optimization services based on CDBTune, a database performance optimization system for intensive learning, even better than a full-time DBA.