|
[51CTO.com original article] The application of information networks has gradually penetrated into various fields of social development. The rapid expansion of network business volume has put forward higher requirements for data transmission capacity and high-throughput cross-capacity. On November 22, Tencent Cloud officially released the Supermind intelligent network product. Compared with the characteristics of previous network products, the Supermind intelligent network will have three major characteristics: high performance, global interconnection, and intelligence.
[[210946]]
Gao Hang, Senior Product Manager at Tencent Cloud
Higher-performance network capabilities to fully meet high-real-time scenarios
Tencent Cloud has been committed to providing users with high-quality network products. In 2013, Tencent Cloud built the prototype of cloud network through basic network, elastic public IP, internal and external network load balancing, security group and other technologies; in 2015, Tencent Cloud's second-generation network products, through the overlay model, realized VPC, VPN, dedicated line access, elastic network card, elastic expansion and contraction and other functions, promoting the development of hybrid cloud.
Tencent Cloud's supermind intelligent network will have major breakthroughs in cloud host and load balancing performance.
At present, the physical network card configuration of Tencent Cloud Server has been upgraded to a 25Gbps bonding architecture. At the same time, by using the separation of the network action layer (fastpath) and the policy layer (slowpath) of the smart network card SDN module, the network bandwidth throughput of Tencent Cloud Host has increased from 8Gbps to 25Gbps, an increase of more than 3 times; the network packet forwarding capacity has increased from 500,000 pps to 4.5 million pps, an increase of 9 times; the network packet virtualization latency has been reduced by 50%, which can fully support scenarios with high network performance requirements such as financial analysis, gaming business, and video live broadcast. At the same time, by using DPDK optimization, the network packet forwarding performance of Tencent Cloud's four-layer load balancing has been upgraded from 2.2 million to 2.1 million, an increase of nearly 10 times.
In terms of load balancing, Tencent Cloud is the first cloud computing service provider in China to support the quic network transmission protocol. The quic network transmission protocol improves the functions of the TCP protocol, reconstructs the TLS protocol, and also absorbs most of the features of HTTP2, which has greatly improved the network's transmission, stability, security, concurrency, and speed. At present, based on this protocol, when performing network file transfer, the access speed can be greatly increased by more than 15%. Previously, Tencent Cloud used the QUIC protocol on some pages of QQ Space, and the first byte time was reduced by an average of 326ms compared to http2, and the performance was improved by about 25%. It can greatly improve the loading progress of the page and effectively reduce the problem of poor experience caused by network delays.
Inject AI power to better unleash network performance
The biggest highlight of Tencent Cloud Supermind network products is the use of artificial intelligence to achieve the entire process of network planning, construction, and operation and maintenance. It provides coverage planning, resource management, network monitoring, fault location, troubleshooting, security protection and other capabilities, forming an intelligent closed loop from network system design, management, operation, optimization, security and other aspects.
Tencent Cloud Supermind intelligent network can combine the characteristics of tens of thousands of lines for intelligent planning. Through historical usage analysis, network path analysis, user feedback analysis, node planning suggestions, dedicated line expansion suggestions, etc., it can reasonably plan the network construction route, greatly shorten the planning time, and save time costs.
By monitoring CVM, load balancing, public opinion, etc., analyzing events, operation logs, network logs, etc., locating faults with the help of multi-layer abnormal alarm convergence aggregation algorithm, and outputting alarm information after relevant alarm aggregation and simplification. By providing positioning suggestions through AI, the time of the whole process was shortened to less than 5 minutes, and the tie processing time was reduced by 75%; the time from manual alarm identification to completion of processing was controlled within 20 minutes, and the overall system SLA was improved by 20%.
In terms of fault self-healing, known network faults are graded through a pre-set fault evaluation mechanism. For faults with high fault prediction confidence intervals and that are relatively controllable, the corresponding fault isolation services are automatically triggered. For example, the external network will schedule DNS traffic, BGP traffic, etc., the internal network will isolate faulty devices or ports, and the underlying cluster will be automatically expanded or reduced in capacity, etc.
Tencent Cloud Supermind Network also has functions such as AI model disassembly, comprehensive information protection, semantic analysis and fraud protection prompts, public opinion analysis + key time point protection, anti-robot simulation of faces, fingerprints, voiceprints, anti-systemic DDoS attacks, and anti-AI blasting password definition and verification.
10+Tb backbone network, cloud-centric global network acceleration
In recent years, Tencent Cloud has been committed to the construction of network infrastructure. As of now, Tencent Cloud has deployed 36 availability zone nodes in 21 geographical regions around the world. It is the Chinese Internet cloud service provider with the most extensive global cloud computing infrastructure layout, covering Asia, Europe, North America, South America, and Oceania.
At the same time, Tencent Cloud's global network routes can provide customers with 35 domestic and nearly 100 global operator BGP access, and provide TB-level total export bandwidth capabilities, providing customers with high-quality, large bandwidth, and low-latency network interconnection capabilities. In 2017 alone, Tencent Cloud has expanded its bandwidth capacity by nearly 300%, and has built a total of 10+Tb-level backbone networks around the world. Based on this, Tencent Cloud has achieved cross-regional traffic scheduling, IP/LB cross-regional binding and other capabilities through "IP release control + cloud backbone network + BGP interconnection", helping users achieve network optimization and global service.
[[210947]]
With 15 years of BGP network operation experience, a powerful global interconnected node network, and a cross-regional network scheduling controller based on SDN (Software Defined Network), Tencent Cloud is also the first in China to launch global dynamic acceleration capabilities based on IPAnycast to fully help customers optimize their networks and implement cross-regional services. This will help users have a multi-path, multi-export disaster recovery system; reduce the packet loss, acceptance, and jitter caused by network congestion; reduce the impact of DDOS attacks on users; and enable global business services, and achieve the best quality through IP anycast in appropriate regions, fully helping customers optimize their networks and implement cross-regional services.
Gao Hang, senior product manager of Tencent Cloud, said that the network is one of the decisive factors in user experience. Tencent Cloud Supermind intelligent network will further narrow the distance between the world and better help user businesses achieve intelligent + transformation and upgrading in the cloud with intelligent and fast networks.
[51CTO original article, please indicate the original author and source as 51CTO.com when reprinting on partner sites] |
